8 posts tagged with "Claude"

Most guides about "Claude and Sales Navigator" jump straight to browser bots, Playwright scripts, and API keys. That is one valid path — we wrote the deep technical version in Automate LinkedIn Sales Navigator with Claude Code — but it is not where most SDRs should start, and it is not what most of you are searching for.

If you are a rep who lives inside Sales Navigator every day, you do not need to build a scraper. You need a repeatable, manual workflow where Claude does the research and writing while you stay in control of the account. No code. No automation tools that get your profile restricted. Nothing that violates LinkedIn's terms.

This is that workflow. Copy the prompts, run it on your real saved searches this week, and you will cut the research-and-writing half of your day down to a fraction of it.

The rule that keeps your account safe​

Before any workflow, one hard line: Claude never touches LinkedIn directly in this method. You do the searching, the profile reading, and the sending inside Sales Navigator like a normal human. Claude works on the text you paste to it.

Why this matters: LinkedIn detects and restricts automated browsing. Tools like Dux-Soup, LinkedHelper, and Expandi live in a permanent cat-and-mouse game with LinkedIn's detection, and when they lose, your account — your book of business — gets locked. The no-code workflow sidesteps all of that because there is no bot. You are just a rep who happens to write faster and research deeper than everyone else on the floor.

If you later decide the volume justifies real automation, go read the technical automation guide and make that call deliberately. Until then, manual is safer and, for most reps, plenty fast.

The four-step workflow​

Here is the whole loop. Each step has a prompt you can lift verbatim.

1. Segment — Turn a saved search into a prioritized worklist.
2. Research — Turn each profile into a one-paragraph angle.
3. Write — Turn the angle into an email, a connection note, and a DM.
4. Reply — Turn inbound responses into fast, on-voice follow-ups.

Step 1: Segment a saved search into a worklist​

Open your saved search in Sales Navigator. Select a page of results and copy the visible rows — name, title, company, and any snippet Navigator shows you. Paste that block into Claude with this prompt:

You are helping me prioritize outbound. Here is a list of prospects pulled from a LinkedIn Sales Navigator search. My ICP is [describe your ICP — e.g., "VP or Director of Sales at B2B SaaS companies, 50 to 500 employees, that run an outbound SDR team"]. Rank these prospects from most to least worth contacting today. For each, give a one-line reason tied to their title, company, or any signal in the row. Flag anyone who is clearly out of ICP so I can skip them.

You now have a ranked list instead of a wall of names. This is the same prioritization logic that anchors the morning block in the Claude SDR daily routine — start every session by deciding who before you touch how.

Step 2: Research each prospect into an angle​

For your top prospects, open the profile in Sales Navigator, then copy the parts that matter: the About section, current role, a recent post if there is one, and the company's tagline or recent news. Paste it in and ask:

Here is a LinkedIn profile and some company context for a prospect I want to reach. Write me a three-sentence briefing: (1) what this person likely cares about right now based on their role and recent activity, (2) the single most credible reason my product could matter to them, and (3) one specific detail I can reference in an opener so it does not read as templated. My product: [one-line description]. Do not invent facts — only use what is in the text I gave you.

That last sentence — do not invent facts — is the whole game. Claude will happily hallucinate a funding round if you let it. Constrain it to the source material and it becomes a research assistant instead of a liability. This is the same discipline we cover in depth in Prospect Research with Claude Code and Automate Lead Research with Claude Code.

Step 3: Write the first touch​

Now you have an angle. Turn it into copy:

Using this briefing, write three things for me in my voice: (1) a cold email under 90 words with a specific opener, one clear value sentence, and a soft ask for 15 minutes; (2) a LinkedIn connection note under 300 characters that references the same detail; (3) a short follow-up DM to send after they accept. Keep it plain and human — no buzzwords, no "I hope this email finds you well," no fake urgency. Here is my briefing: [paste].

Read every draft before it goes out. Edit one line in each so it sounds like you and not like a model. The reps who get flagged as "AI slop" are the ones who send raw output; the reps who win are the ones who use Claude to get to a strong 80% draft in ten seconds and spend their judgment on the last 20%. The mechanics of doing this at volume without sounding robotic are in Personalized Cold Emails at Scale.

Step 4: Reply and follow up​

When responses come in — including the "not right now" and "who are you" replies — paste the thread into Claude:

Here is a reply from a prospect. Draft a response in my voice that moves toward a meeting without being pushy. If they raised an objection, address it honestly in one or two sentences. Keep it short. Thread: [paste].

Keep a one-page "voice doc" — three of your best real emails — and paste it in alongside these prompts. The more you feed Claude examples of how you write, the less editing you do over time.

What this replaces (and what it does not)​

This workflow eats the two biggest time sinks in an SDR's day: research and first-draft writing. It does not replace judgment, relationships, or the actual conversation. Claude is a prep and drafting engine, not a rep.

Do not use it for:

• Sending. Your sequencer sends; Claude drafts. Keep those separate for deliverability.
• Live calls. Claude preps you before the call — see Meeting Prep with Claude Code — but it should never be on the call.
• Anything relational. Referral asks, exec sponsorship, expansion talks. A Claude-written DM to a CFO reads as Claude-written, and they clock it instantly.

We made the full argument for where the human line sits in Why General AI Won't Replace the SDR Stack.

Claude, ChatGPT, or something else for this?​

For the Sales Navigator workflow specifically, Claude's long context is the edge: you can paste a full profile, a company page, and three of your past emails all at once, and it holds the whole picture while it writes. If you want the honest head-to-head on model choice for sales work, we broke it down in Claude vs ChatGPT for Sales Teams.

If your goal is broader than Sales Navigator — sourcing net-new accounts, not just working a saved search — pair this with How to Use Claude for Lead Generation.

Where this fits in the bigger picture​

Sales Navigator tells you who exists. It does not tell you who is in-market right now, and that is the difference between a cold list and a warm one. The workflow above makes you faster at working any list; the leverage compounds when the list itself is prioritized by real buying signals.

That is the layer MarketBetter sits in: it surfaces the accounts showing intent, routes them, and tracks what happens after the touch — so your Claude-powered outreach lands on the prospects most likely to reply. See how the pieces fit in the AI SDR tech stack and the way a signal becomes a booked meeting in From Buying Signal to Booked Meeting in 24 Hours.

For the full map of everything Claude can do across the SDR role — research, email, CRM cleanup, pipeline reporting — start with the pillar: Claude for SDRs: The Complete Guide.

Start this week​

Do not automate anything yet. This afternoon:

1. Open one saved search and run the Step 1 segmentation prompt on a single page of results.
2. Take your top three prospects through Steps 2 and 3.
3. Send three genuinely researched touches before you log off.

Get the manual loop working on real prospects first. If the time savings are obvious — and they will be — then decide whether the volume justifies going technical.

If you want the signal layer that decides which prospects belong in your Claude pipeline in the first place, that is what we built MarketBetter for. Book a demo and we will show you the whole loop end to end.

Let's start with the honest answer, because most articles on this topic won't give it to you: Claude cannot generate leads by itself. It has no built-in contact database, it can't scrape LinkedIn at scale, and if you ask it for "50 CMOs at Series B fintechs," it will happily hallucinate 50 names, half of which don't exist.

So why is "how to use Claude for lead generation" one of the fastest-growing searches in B2B sales? Because the people asking it have figured out something real: Claude isn't the source of leads — it's the reasoning layer that turns raw, messy, low-quality lists into a prioritized worklist of accounts actually worth your time. That's where 80% of a lead-gen team's hours disappear, and it's exactly the part Claude is world-class at.

This is the step-by-step playbook for doing it right. Five stages, the exact prompts, and a clear line on what Claude can and can't do — so you don't waste a week discovering the limits the hard way.

If you want the broader role-level picture, the complete Claude-for-SDRs pillar guide covers the full SDR job. This post is narrower and deeper: it's specifically about generating and qualifying net-new leads.

Can Claude generate leads? What it actually can and can't do​

Set expectations first. This one table saves you the most common mistake.

The pattern: Claude is the judgment and synthesis engine. You still need a source of raw leads and, usually, an enrichment step for verified contact data. Get those two things feeding Claude and the middle of the funnel — the qualification grind that eats your reps' mornings — collapses from hours to minutes.

For a head-to-head on which model handles this best, see Claude vs ChatGPT for sales teams. Short version: Claude's long context and consistent reasoning across a 2,000-row list is the deciding factor for lead gen specifically.

The 5-stage Claude lead generation workflow​

Here's the full pipeline. Each stage feeds the next.

1. Encode your ICP — turn "our best customers" into a machine-readable rubric
2. Source raw leads — get companies and contacts from a real source
3. Qualify at scale — score the raw list against the rubric
4. Enrich the winners — find the right person and their context
5. Prioritize into a worklist — a ranked queue your reps actually work

Skip stage 1 and everything downstream is garbage. Let's build it.

Stage 1 — Encode your ICP as a machine-readable filter​

Most teams "know" their ICP but have never written it down in a way a machine can apply consistently. That's the highest-leverage 20 minutes in this entire process.

Prompt:

You are helping me build a lead qualification rubric.

Here are 8 of our best current customers and why they're great fits:
[paste 8 accounts + one line each on why they closed and stuck]

Here are 4 accounts that looked good but churned or never closed:
[paste 4 + why they failed]

Produce a scoring rubric with:
- 5-7 firmographic criteria (industry, size, tech, funding stage, etc.)
- 2-3 disqualifiers (auto-reject signals)
- A 0-100 scoring formula weighting each criterion
Output it as something I can reuse to score new companies.

The output is a reusable rubric grounded in your real wins and losses — not a generic "50-500 employees, B2B SaaS" guess. Save it. You'll paste it into every qualification run from now on.

For a deeper treatment of turning fit into a repeatable score, see Claude Code SDR Part 6: Lead Scoring.

Stage 2 — Source your raw leads (this is the part Claude can't fake)​

Claude needs raw material. You have three honest options for where it comes from:

Option A — LinkedIn Sales Navigator. Build a search that roughly matches your ICP, export or copy the results, and hand them to Claude to qualify. The Sales Navigator + Claude workflow walks through this end to end. Sales Nav gives you breadth; Claude gives you the filtering Sales Nav can't.

Option B — Website visitor identification. This is the highest-intent source that most teams ignore. The companies already researching you are worth ten cold ICP matches. Tools that de-anonymize your traffic turn "someone from a mid-market logistics firm read your pricing page twice" into a named account you can act on today. That's the source we care most about — more on it below.

Option C — Free and low-cost tools. If you're bootstrapping, there's a real stack of free options. We break them down in the best free AI lead generation tools for B2B and the best B2B lead generation tools.

Whatever the source, the output of this stage is a raw list — messy, unqualified, full of noise. That's fine. Stage 3 is where Claude earns its keep.

Stage 3 — Qualify the raw list at scale​

This is the magic step. You have 300 raw companies and a rubric from Stage 1. Feed both to Claude.

Prompt:

Here is my ICP scoring rubric:
[paste rubric from Stage 1]

Here is a raw list of 300 companies with the fields I have
(name, industry, employee count, website, any notes):
[paste CSV/list]

For each company:
1. Score it 0-100 against the rubric.
2. Give a one-line reason for the score.
3. Flag any auto-disqualifiers.
Return the top 40 by score as a table, sorted high to low.
Be conservative — if you lack evidence a company fits, score it lower,
don't guess.

That last line matters. Telling Claude to penalize missing evidence instead of inventing it is the single most important instruction for keeping lead-gen output trustworthy. A rep who can trust the top-40 list works it; a rep who's been burned by hallucinated fits ignores the whole thing.

Two minutes of Claude replaces an afternoon of a rep eyeballing a spreadsheet — and it's more consistent, because Claude applies the same rubric to row 300 as it did to row 1. For the underlying research mechanics, see automate lead research with Claude Code and Claude Code SDR Part 2: Prospect Research.

Stage 4 — Enrich the winners​

Now you have 40 qualified companies. You need the right person at each and enough context to open a real conversation. Claude can't pull verified emails on its own, but once you feed it enrichment data (from your provider) plus public signals, it synthesizes a briefing no rep has time to write by hand.

Prompt:

For each of these 40 companies, I've pasted the LinkedIn profile of the
most likely buyer plus their company's recent news:
[paste enrichment data]

For each, produce:
- Confirmed best-fit contact + title + why them
- A one-paragraph "why now" briefing (trigger event, pain, angle)
- One specific, non-generic opening line I could actually send
Keep each under 80 words. No filler, no "I hope this finds you well."

You now have 40 fully-briefed, ready-to-work leads. The full breakdown of turning research into first-touch lives in AI for sales prospecting and, for the outreach itself, LinkedIn outreach automation with Claude Code.

Stage 5 — Prioritize into a daily worklist​

Forty leads is still too many to work well at once. The last step is ranking them into the order a rep should actually attack — fit plus intent, not fit alone.

Prompt:

Here are my 40 enriched leads with fit scores.
I'm also pasting intent signals where I have them
(site visits, content downloads, job changes, funding):
[paste]

Re-rank all 40 into a single prioritized worklist. Weight recent,
high-intent signals heavily — a medium-fit account that just visited
our pricing page outranks a perfect-fit account that's gone quiet.
Group into: Call today / Sequence this week / Nurture.

That's a lead-generation pipeline that runs in an afternoon and outputs a worklist your reps trust. To wire this into a daily cadence, the Claude SDR daily routine shows the exact 90-minute block. And if deliverability is a concern as you scale outreach, read how to build a prospecting engine without burning your domain first.

The honest limits (and how to work around them)​

Because no one else will say it plainly:

• Claude will confidently invent contacts. Never let it be the source. Always give it a real list to work on, never ask it to produce one from nothing.
• It doesn't have live data. "Recent funding" or "current headcount" needs to come from your source or enrichment tool. Claude reasons over data; it doesn't fetch it.
• Verified emails require a real provider. Claude can guess an email pattern; it can't confirm one is deliverable.
• Scoring is only as good as your rubric. Garbage ICP in, garbage worklist out. Stage 1 is not optional.

Work within those lines and Claude is the best qualification-and-synthesis engine your team has ever had. Ignore them and you'll generate a list of ghosts.

Where the leads should really come from​

Here's the strategic point most "Claude for lead gen" advice misses. The best raw source isn't a bigger cold list — it's the people already showing intent. Companies visiting your site are further down the buying journey than any cold ICP match, and they've told you what they care about by which pages they read.

That's the gap MarketBetter fills. We de-anonymize your website traffic into named accounts, layer on the buying signals, and — this is the part that matters — tell your reps what to do next, not just who visited. Claude is brilliant at reasoning over a list. MarketBetter makes sure the list is made of real, high-intent companies instead of cold guesses.

Claude tells your SDRs what to do. MarketBetter tells them who to do it for — with the intent data that makes every message land.

Pair the two and the five stages above stop being a manual afternoon and become a system: high-intent leads in, prioritized worklist out, every day.

Start generating better leads​

Claude is a force multiplier, not a lead database. Give it a real source, a sharp ICP rubric, and clear instructions, and it will do the qualification work of a small team — consistently, in minutes.

The one thing it can't manufacture is a good source of leads. That's worth solving first.

Want to see high-intent leads flow straight into a Claude-ready worklist? Book a demo →

Most SDRs we talk to use Claude the same way they use ChatGPT — open a tab, paste a question, copy the answer, repeat. That works, but it leaves most of Claude's value on the floor.

The SDRs who get a real multiplier out of Claude don't treat it as a faster Google. They treat it as a morning copilot — a fixed 90-minute block, the same five sub-tasks every day, with prompts they've sharpened over months. The output isn't "AI content." It's a stack of ready-to-send messages, a triaged signal queue, a clean account plan, and an inbox at zero.

This post is the exact routine. Five blocks. The prompts. What they're actually replacing.

If you want the broader strategic case, the complete Claude-for-SDRs pillar guide walks through why Claude wins for SDR work. This post is the operating manual.

The 4 tools this routine replaces​

Before the breakdown, what the 90 minutes is actually compressing:

Total replaced: roughly 3 hours of common SDR busywork compressed into a 90-minute block, before your first dial.

The catch: every minute saved comes from prompt structure, not from Claude being magic. The prompts below assume you've fed Claude your ICP, your product positioning, and three or four "good email" examples in a saved project. If you haven't, start here on prospect research and here on cold email personalization.

Block 1 — Signal triage (15 minutes)​

What you're doing: sorting overnight signals — visitor ID hits, intent topic spikes, job change alerts, replies — into three buckets: call-now, sequence-today, snooze.

Why Claude is good at this: it's pattern recognition over noisy fields, exactly the kind of thing a human gets bored doing by 9:05 AM.

Prompt:

You are my SDR signal triage copilot. I will paste a CSV/list of overnight signals.
For each row, classify into one of: CALL_NOW, SEQUENCE_TODAY, SNOOZE_7D.

Rules:
- CALL_NOW: returning visitor on /pricing or /book-demo, OR open champion at a target account who just changed roles, OR an intent spike on a top-3 use case at an account already in pipeline.
- SEQUENCE_TODAY: first-touch ICP fit signals — new visitor on a product page, fresh intent spike, persona-fit job change at a fit account.
- SNOOZE_7D: weak signals — single page view on /blog, off-ICP firmographics, signals at accounts already in late-stage with another seller.

Return a table with: account, contact, signal, classification, 1-line "why this bucket".
At the bottom, list the CALL_NOW accounts with the strongest "first 90 seconds" opener I should use, referencing the specific signal.

What replaces this otherwise: a 30-minute scroll through Sales Navigator + your visitor ID tool + your intent platform, trying to remember which accounts are already in flight. If you're doing this manually every morning, you're paying the same cost twice — the platform fee and the SDR's morning.

For the deeper case on why signal-first SDRs out-book signal-blind ones, see From Buying Signal to Booked Meeting in 24 Hours.

Block 2 — Targeted prospect research (20 minutes)​

What you're doing: taking your 3-5 CALL_NOW or top-priority accounts from Block 1 and turning each into a one-screen account brief.

Why batch: a single research session with context loaded once is faster than five context switches.

Prompt (one per account):

Research brief for [ACCOUNT NAME]. I'm an SDR selling [your product, 1 line].
Contact I'm reaching is [NAME, TITLE].

Pull from the company site, recent press, their LinkedIn page posts, and the contact's
LinkedIn activity in the last 90 days. Return:

1. Company snapshot — 2 lines max. What they actually do, not their tagline.
2. Recent "why now" — top 3 events in the last 90 days that justify outreach today.
3. Strategic priorities — what leadership is publicly talking about.
4. Personal hook for [NAME] — something they personally posted, said, or shipped that
   I can reference without being weird about it.
5. Three opener angles, ranked by likely reply rate, with the explicit pitch each implies.

If a section has nothing solid, say "no clean signal" — do not invent.

That last instruction matters. Hallucinations are 90% prompt-permission errors. If you give Claude an explicit out, it takes it. If you don't, it fills the gap. (More on that pattern in our writeup on Claude vs ChatGPT for sales teams.)

Block 3 — Account plan drafting (15 minutes)​

What you're doing: for the 2-3 hottest accounts, converting the research brief into a one-page plan you can drop in your CRM or hand to an AE.

Prompt:

Convert the research brief above into a one-page account plan in this structure:

- Account: name, segment, size, deal-trigger event
- Buying committee — likely roles, who I have, who I'm missing
- Use-case fit — which 1-2 of our use cases match their stated priorities
- Risks — what kills this deal at each stage (no demo, no champion, no budget cycle)
- 14-day plan — day-by-day, what I do, what I expect back, when to escalate to AE
- Discovery questions — 5 I would actually ask on a first call, ranked by signal value

Be specific. No platitudes. If a section is weak, write "needs more research" — do not pad.

This is the step that makes the AE conversation different. Most SDRs hand over a contact and a paragraph. The AEs who book repeat business get a 14-day plan with a discovery agenda. For more on what that handoff actually looks like, see the SDR-to-AE handoff playbook and the 15-minute pre-demo prep playbook.

Block 4 — Inbox at zero (20 minutes)​

What you're doing: processing replies, scheduling pings, and "soft no" responses. Every reply gets one of four actions: book, nurture, re-engage, archive.

Prompt:

I will paste replies from overnight. For each reply, return:

- Intent classification: HOT, WARM, COLD, NEGATIVE, OUT_OF_OFFICE, REFERRAL
- Suggested action: book meeting / send Loom / soft re-engage in 30d / mark closed-lost / forward to AE
- A 3-sentence draft reply in my voice (matching the examples I gave you in the project), ready to paste.
- A "do not send if" line — the 1-2 conditions that should make me NOT send the draft.

Group output by intent. Put HOT and REFERRAL at the top.

The "do not send if" line is the only reason this block stays at 20 minutes instead of 45. It tells you which drafts to skim past versus which to actually review. Without it, every Claude-drafted reply gets the same level of scrutiny — and you end up reading 30 drafts to decide on 8.

For replies where the prospect went quiet mid-cycle, the Champion Goes Quiet playbook has the specific re-engagement sequences worth pasting in as Claude context.

Block 5 — Personalized first-touch drafts (20 minutes)​

What you're doing: drafting first-touch sequences for the 8-12 accounts you'll add to your active list today. Three-touch sequence per account: cold email, LinkedIn note, follow-up email.

Prompt:

For each account in the list below, draft a 3-touch personalized outbound sequence:

Touch 1: Cold email, max 90 words. Open with the specific "why now" from the research
brief — not a flattery line. CTA is a soft ask (book 15 min OR a specific question).
Touch 2: LinkedIn note, max 280 chars. Reference touch 1 obliquely, not directly.
Touch 3: 4 days later, plain-text follow-up. New angle, not "just bumping this up."
Reference a different "why now" if one exists.

Voice: match the examples in the project. No "I hope this finds you well." No "saw you
guys are doing great things." Concrete or skip the line.

For each account, also output the 1-line subject line you'd actually open with.

Two non-obvious things matter here:

1. Voice examples in the project beat any prompt instruction. Telling Claude "write in my voice" without 3-5 saved examples produces marketing copy. With examples, it produces something you'd actually send. We dig into this in the Claude 200K context for sales workflows post.
2. The "new angle on touch 3" rule is what stops the sequence from feeling like a follow-up. Most AI-generated sequences fail at touch 3 because they reuse the touch-1 hook. Force a different angle and reply rates climb.

What this routine doesn't do​

It doesn't replace dials. It doesn't replace your call recording review. It doesn't replace 1:1 coaching from your manager. And it doesn't make a bad ICP fit into a good account.

It compresses the administrative and research overhead that's traditionally eaten 60% of an SDR's day, so you can spend more of the remaining day on the things only a human does well: phone, video, and judgment.

The teams getting outsized returns from Claude pair this routine with a platform that surfaces the right signals into the morning queue in the first place. That's the gap MarketBetter fills — every account in your CALL_NOW bucket came from visitor ID, intent, or champion-tracking signals the platform pushed to you, not a list you remembered to check. Claude turns those signals into ready-to-send work in 90 minutes. The rest of your day is selling.

Where to go next​

If you want to keep going deeper into Claude-for-SDR specifically:

• Claude for SDRs: The Complete Guide — the pillar that ties everything together
• Claude Code SDR Part 1: AI-Powered SDR — the 10-part hands-on series
• Claude Code SDR Part 2: Prospect Research
• Claude Code SDR Part 3: Personalized Cold Emails
• Claude Code SDR Part 9: Follow-up Sequences
• Claude vs ChatGPT for Sales Teams — why we keep going back to Claude
• Visitor ID to First Outreach in 30 Minutes — pairs with Block 1
• The Inbound Triage Tier System — adjacent triage framework for inbound

Want a signal queue that actually fills your CALL_NOW bucket every morning? That's what MarketBetter does for the SDRs running this routine. Book a demo and we'll show you what your morning queue would look like with real visitor ID, intent, and champion-tracking signals running into it.

If you're an SDR in 2026 and you're not using Claude for at least a third of your daily workflow, you're getting outworked by people who are.

This isn't speculative. It's the consistent pattern we see across the GTM teams using MarketBetter: the SDRs who pair Claude with their existing tools (Sales Navigator, CRM, sequencer, enrichment) are booking 2–3x more qualified meetings — not because they grind harder, but because Claude eats the parts of the job that used to eat their day.

This pillar is the single page that pulls it all together. It's a map. Each section links into a deeper, hands-on guide so you can go as shallow or as deep as you want.

Use this guide if you want to:

• Understand which sales tasks Claude is actually good at (and which still need a human)
• See concrete workflows for prospect research, Sales Navigator, email personalization, and CRM hygiene
• Compare Claude vs. ChatGPT vs. Codex for SDR work
• Get a daily routine you can copy and run starting tomorrow

Let's get into it.

What Claude actually is (and why SDRs care)​

Claude is Anthropic's family of large language models — the same kind of underlying technology behind ChatGPT, but built with a different design philosophy. For sales work, three things matter:

1. Long context. Claude can hold the equivalent of a 500-page document in working memory. You can drop in a whole company's 10-K, a quarter of call transcripts, or a CSV with 2,000 leads, and ask questions across all of it. Most sales workflows benefit from this more than from raw "intelligence."
2. Reasoning that holds together. When you ask Claude to compare 30 prospects against your ICP and prioritize them, it doesn't lose the thread halfway through. That matters when the output is a worklist you're about to grind through.
3. Claude Code. The CLI version of Claude can read files, run scripts, hit APIs, and do real work in a terminal — not just chat. That's what unlocks the workflows in this guide.

If you've never opened Claude Code, start with The AI-Powered SDR: How Claude Code + MarketBetter Changes Everything. It's the on-ramp.

For a deeper head-to-head on which model to use when, see Claude vs ChatGPT for Sales Teams and Codex vs Claude Code for Outbound Sequences.

The five things Claude is genuinely good at for SDRs​

Most SDR teams trying AI fail because they pick the wrong tasks. AI is not magic — it's a very specific kind of leverage. After watching dozens of GTM teams roll this out, five jobs consistently produce a return.

1. Prospect research at scale​

The before: an SDR opens a LinkedIn profile, copies the bio into a doc, hunts for the company's last funding round, reads the latest blog post, then attempts a "personalized" opener. Twenty minutes per prospect, fifteen prospects a day.

The after: Claude reads the LinkedIn profile, the company about page, the last three blog posts, and a Crunchbase entry, then drafts a one-paragraph "what to actually open with" briefing. Two minutes per prospect, sixty prospects a day, and the openers are sharper because Claude can hold all four sources in working memory at once.

Hands-on walkthrough: Claude Code SDR Part 2: Prospect Research and Automate Lead Research with Claude Code.

2. Personalized cold email at volume​

There's a chasm between "generic AI-written email" and "actually personalized email." The difference is the inputs. If you hand Claude a job title and a company name, you get generic slop. If you hand it the prospect's last LinkedIn post, a snippet from their company's earnings call, and your ICP framing, you get something a human couldn't tell from a hand-written email — at 30x the speed.

We've broken the workflow down step by step in Claude Code SDR Part 3: Personalized Cold Emails and AI Email Personalization at Scale. Templates that have produced real opens: AI Sales Email Templates with Claude Code.

3. Sales Navigator → enriched list pipeline​

Sales Navigator is a goldmine, but it's also a UI nightmare. Most SDRs end up exporting CSVs and gluing tools together. Claude Code can sit in the middle of that pipeline — taking a raw export, hitting enrichment APIs, scoring against ICP, and dropping a ready-to-sequence list into your CRM or MarketBetter campaign.

Full walkthrough: Automate LinkedIn Sales Navigator with Claude Code and Claude Code SDR Part 4: LinkedIn to Pipeline.

4. CRM cleanup and duplicate hunting​

This is the boring, underrated win. Every SDR org we look at has tens of thousands of dirty records — duplicate companies, inconsistent job titles, missing fields, accounts owned by reps who left two years ago. Claude is unreasonably good at this kind of pattern work because it can hold the whole CSV in context and make consistent, explainable decisions.

For a real example of what dirty data costs you and how to fix it: When CRM Has 3 Records for the Same Company and Claude Code SDR Part 7: CRM Cleanup.

5. Pipeline analysis and reporting​

The other underrated win. Once a week, drop your CRM export into Claude and ask: "What changed in pipeline this week? Which deals look at risk? Which reps are leaning on a single mega-deal?" In ten minutes you get a weekly business review most ops teams take two days to produce.

Deep dive: AI Pipeline Velocity Optimization with Claude Code and Claude Code SDR Part 6: Lead Scoring.

What Claude is NOT good at (don't waste time here)​

This is the part most "AI for sales" content skips. The list of things Claude shouldn't be doing in your workflow:

• Actually sending the email. Claude drafts; your sequencer sends. Mixing the two is how you end up with deliverability problems and brand damage.
• Live discovery calls. Claude is a research and prep tool, not a replacement for the conversation. The SDRs who try to use it on live calls sound exactly like what they are.
• Anything that needs a relationship. Referral asks, expansion conversations, exec sponsorship — these are still 100% human. Claude can help you prep, but a Claude-written DM to a CFO will read as Claude-written, and they will clock it instantly.
• Hard objections you don't understand yet. If you can't articulate why a prospect might say no, Claude can't either. It can help you brainstorm, but it can't shortcut the muscle of actually understanding your market.

We wrote a longer take on this: Why General AI Won't Replace the SDR Stack and Why Open-Source GTM Agents Won't Replace the SDR Platform.

Claude vs. ChatGPT vs. Codex: which one when?​

Short version of a long argument:

• ChatGPT — Best for one-off brainstorms and quick rewrites in a browser. The product layer is more mature for non-technical users.
• Claude (web) — Best when you need to drop in a long document (an RFP, a deck, a transcript) and ask deep questions. The long-context advantage is real.
• Claude Code — Best when the work is repeatable and touches files, APIs, or your terminal. This is where the 10x leverage lives.
• Codex / OpenAI CLI — Best when the work leans heavier on code generation than on reading/reasoning over content. Decent for sequencer integrations.

Full comparison matrices: Codex, Claude, ChatGPT for GTM Comparison, Claude vs ChatGPT for Sales Teams, Codex vs Claude Code for Outbound Sequences, and the practical OpenAI Codex CLI GTM Guide.

If your team is debating whether to build something custom or buy a platform, read Build vs Buy: The AI SDR Stack Decision before the next meeting.

The 10-part Claude Code SDR series, in order​

If you want the hands-on path, work through the series in order. Each part is ~10 minutes to read and another 15–30 to set up:

1. Part 1 — The AI-Powered SDR: How Claude Code + MarketBetter Changes Everything
2. Part 2 — Prospect Research with Claude Code
3. Part 3 — Personalized Cold Emails at Scale
4. Part 4 — LinkedIn to Pipeline
5. Part 5 — Competitive Intelligence
6. Part 6 — Lead Scoring with AI
7. Part 7 — CRM Cleanup
8. Part 8 — Meeting Prep
9. Part 9 — Follow-up Sequences
10. Part 10 — The Complete Playbook

Tangential but useful: AI Buyer Persona Research Automation with Claude Code, AI Objection Handler with Claude Code, Multi-language Cold Outreach with AI, and AI Sales Onboarding Automation.

A realistic Claude-powered SDR day​

Here's what a 9-to-5 actually looks like for an SDR who has internalized this workflow. Adjust to taste.

9:00 — Triage and target list (30 min)​

Open Claude Code. Hand it last night's MarketBetter signal feed plus your CRM export. Ask: "Which 25 prospects should I prioritize today, ranked by signal strength and ICP fit, with one sentence each on why?" Paste the output into your day list.

Underlying mechanics covered in: From Buying Signal to Booked Meeting in 24 Hours and Visitor ID to First Outreach in 30 Minutes.

9:30 — Research sprint (45 min)​

For the top 10 prospects, run a research macro. Claude reads LinkedIn, the company about page, last earnings call (if public), and last 3 blog posts. Produces a one-paragraph "what to open with" briefing per prospect. Total time: ~4 minutes per prospect, parallelized.

10:15 — Personalized outbound block (75 min)​

For each researched prospect, Claude drafts an email + LinkedIn DM + voicemail script using your templates and the research briefing. You read, edit (always edit), and queue in the sequencer. Expected output: 20–25 outbound touches that don't read as templated.

11:30 — Live calls (90 min)​

This is human time. Claude shouldn't be on the call. But before each call, give Claude 30 seconds: "Pull the meeting prep brief for [prospect name]." It hands you the angles, the questions you should ask, and the likely objections.

Covered in Claude Code SDR Part 8: Meeting Prep.

1:00 — Lunch (you, not Claude)​

2:00 — Follow-ups and replies (60 min)​

For replies that came in overnight, paste them into Claude and ask for a draft response in your voice. Same for follow-ups on cold opens. The model gets better at "your voice" the more you correct it — keep a one-page style doc and feed it in every time.

Workflow: Claude Code SDR Part 9: Follow-up Sequences.

3:00 — Round 2 outbound block (90 min)​

A second outbound sprint, weighted toward prospects from this morning's research that you didn't get to yet. Same flow as 10:15.

4:30 — Pipeline hygiene + end-of-day reporting (30 min)​

Claude runs the daily CRM cleanup macro — flags duplicates, missing fields, stale opportunities, and accounts assigned to nobody. You spend ten minutes resolving the top five issues. Then Claude drafts your end-of-day update for your manager from your activity log.

The longer template version of this day: Claude Code SDR Part 10: The Complete Playbook.

Common questions​

Do I need to know how to code to use Claude Code?

No. Claude Code is a command-line tool, not a programming language. You type instructions in English. The reason it's powerful for SDRs is that it can read your CSVs and hit web pages — not that you're writing software.

Will my SDR manager freak out about prospects being touched by AI?

If they're paying attention, the question they'll actually care about is the output, not the tool. SDRs using Claude well are not the ones sending mass-templated AI slop — they're the ones sending sharper, more researched messages than the rest of the team. That conversation tends to land on "show me your workflow," not "stop using it."

What about deliverability? Doesn't AI content get flagged?

Email providers don't flag content because "AI wrote it" — they flag patterns: same body across thousands of sends, links to suspicious domains, low engagement, sudden volume spikes. Claude-drafted but human-edited emails sent at SDR cadence don't trigger any of that. If you want to go deep, we wrote about it in the context of why most signal-based selling rollouts fail in 90 days.

How does Claude compare to a purpose-built AI SDR tool like 11x, Regie, or Nooks?

Different categories. Claude is a general-purpose model you wire into your existing tools. Purpose-built AI SDR platforms are end-to-end products that try to replace the SDR seat. We have a strong opinion on this — Why General AI Won't Replace the SDR Stack — and you can see the head-to-heads in our reviews like Landbase Review 2026.

Where does MarketBetter fit?

MarketBetter is the signal and orchestration layer underneath the workflows in this guide. Claude is the research and writing engine; MarketBetter is the system that surfaces which accounts are in-market right now, routes them, and tracks what happens. The 10-part series is named "Claude Code + MarketBetter" for a reason — they're complements, not competitors. See the AI SDR tech stack for the full picture, or how to build an AI SDR with MarketBetter.

Where to start tomorrow​

If you read nothing else from the links above, do these three things this week:

1. Read Part 1 and install Claude Code. Twenty minutes.
2. Pick one workflow from the five above — most teams start with prospect research because the time savings are immediate and obvious.
3. Run it on your real worklist for one week. Don't try to automate the whole stack at once.

The SDRs who win at this don't move fastest. They move first on the workflow they understand best and then expand from there.

If you want the signal layer that decides which prospects belong in your Claude pipeline in the first place — that's what we built MarketBetter for. Book a demo or keep reading the SDR automation pillar and the B2B intent data pillar for adjacent territory.

Ask any sales rep: "How do you decide who to call first?" You'll get answers like:

• "I work alphabetically through my list"
• "Whatever came in most recently"
• "Gut feeling based on company size"
• "Whoever my manager tells me to"

None of these are strategies. They're coping mechanisms for a broken system.

The best accounts—the ones with the highest likelihood to close and the highest deal value—are often buried in a spreadsheet, never contacted. Meanwhile, reps waste hours on accounts that were never going to buy.

This guide shows you how to build an AI-powered account scoring system with Claude Code that identifies your highest-potential accounts automatically. Stop guessing. Start knowing.

The Real Cost of Poor Prioritization​

Here's what happens when sales teams prioritize badly:

Time Waste:

• Average SDR spends 2+ hours daily deciding who to contact
• 67% of time is spent on accounts that will never convert
• Best accounts get the same attention as worst accounts

Revenue Loss:

• 35-50% of deals go to the vendor that responds first
• High-fit accounts that go uncontacted convert at competitor sites
• Reps hit quota on volume, miss it on value

Burnout:

• Calling dead accounts kills morale
• "Spray and pray" feels pointless (because it is)
• Top performers leave for companies with better systems

The data is clear: teams that score and prioritize accounts effectively see 30% higher conversion rates and 20% shorter sales cycles.

Why Traditional Lead Scoring Fails​

Most lead scoring systems are built on two flawed premises:

Flaw 1: Static Rules​

"Companies with 500+ employees get 10 points."

This ignores:

• Industry context (500 at a tech startup vs. 500 at a hospital = totally different)
• Current buying signals
• Relationship history
• Market timing

Flaw 2: Incomplete Data​

You score what you can measure, but the most predictive signals are often qualitative:

• "They mentioned they're evaluating competitors"
• "Their CTO attended our webinar AND read our pricing page"
• "They just raised a Series B and need to scale sales"

Claude Code can synthesize both structured and unstructured data to create scoring that actually predicts conversions.

The Architecture of AI Account Scoring​

Here's how an intelligent prioritization system works:

1. Data Aggregation​

Pull from every source: CRM, enrichment tools, website behavior, email engagement, social signals.

2. ICP Matching​

Score firmographic fit against your ideal customer profile.

3. Intent Detection​

Identify behavioral signals that indicate active buying.

4. Relationship Mapping​

Account for existing touchpoints and engagement history.

5. Timing Analysis​

Factor in buying cycles, budget periods, and urgency signals.

6. Composite Scoring​

Combine all factors into a single prioritization score.

Building the System with Claude Code​

Step 1: Define Your ICP Criteria​

First, codify what makes an account "ideal":

const ICP_CRITERIA = {
  firmographic: {
    employeeRange: { min: 50, max: 1000, weight: 0.2 },
    revenueRange: { min: 5000000, max: 100000000, weight: 0.15 },
    industries: {
      include: ['SaaS', 'Technology', 'Financial Services', 'Healthcare'],
      exclude: ['Government', 'Education'],
      weight: 0.15
    },
    geographies: {
      include: ['US', 'Canada', 'UK', 'Germany'],
      weight: 0.05
    }
  },
  
  technographic: {
    required: ['Salesforce', 'HubSpot'],
    positive: ['Outreach', 'SalesLoft', 'Gong'],
    negative: ['Competitor X', 'Legacy CRM'],
    weight: 0.15
  },
  
  departmentSignals: {
    hasSalesTeam: { minSize: 5, weight: 0.1 },
    hasMarketingTeam: { minSize: 2, weight: 0.05 },
    hasRevOps: { weight: 0.1 }
  }
};

Step 2: Aggregate Data Sources​

Pull everything you know about each account:

async function aggregateAccountData(companyId) {
  // CRM data
  const crmData = await crm.getCompany(companyId);
  const contacts = await crm.getContacts({ companyId });
  const deals = await crm.getDeals({ companyId });
  const activities = await crm.getActivities({ companyId });
  
  // Enrichment data
  const enrichment = await clearbit.enrich(crmData.domain);
  const techStack = await builtwith.getTechStack(crmData.domain);
  
  // Website behavior
  const webActivity = await analytics.getCompanyActivity(companyId, {
    days: 30
  });
  
  // Email engagement
  const emailEngagement = await emailPlatform.getEngagement(companyId);
  
  // Social signals
  const linkedInActivity = await linkedin.getCompanySignals(crmData.domain);
  
  // News and events
  const recentNews = await newsApi.getCompanyNews(crmData.name, { days: 90 });
  
  // Competitor mentions
  const competitorSignals = await detectCompetitorActivity(companyId);
  
  return {
    company: crmData,
    contacts,
    deals,
    activities,
    enrichment,
    techStack,
    webActivity,
    emailEngagement,
    linkedInActivity,
    recentNews,
    competitorSignals
  };
}

Step 3: Score with Claude Code​

Now use Claude to synthesize all signals into a comprehensive score:

async function scoreAccount(accountData) {
  // Calculate structured scores
  const icpScore = calculateICPScore(accountData, ICP_CRITERIA);
  const engagementScore = calculateEngagementScore(accountData);
  const intentScore = calculateIntentScore(accountData);
  
  // Use Claude for qualitative analysis
  const qualitativeAnalysis = await claude.messages.create({
    model: 'claude-3-5-sonnet-20241022',
    max_tokens: 1000,
    system: `You are a B2B sales strategist analyzing accounts for 
prioritization. You excel at identifying hidden buying signals and 
assessing account quality beyond basic metrics.

Provide:
1. OPPORTUNITY_SCORE (0-100): Likelihood to close
2. VALUE_SCORE (0-100): Potential deal size relative to effort
3. TIMING_SCORE (0-100): Urgency/readiness to buy
4. KEY_INSIGHTS: 2-3 critical observations
5. RECOMMENDED_APPROACH: Best first touch strategy`,
    messages: [{
      role: 'user',
      content: `Analyze this account for prioritization:

COMPANY: ${accountData.company.name}
INDUSTRY: ${accountData.enrichment.industry}
SIZE: ${accountData.enrichment.employeeCount} employees
REVENUE: $${accountData.enrichment.annualRevenue}
TECH STACK: ${accountData.techStack.join(', ')}

RECENT ACTIVITY:
- Website visits: ${accountData.webActivity.pageviews} (${accountData.webActivity.uniqueVisitors} unique)
- Pages viewed: ${accountData.webActivity.topPages.join(', ')}
- Email engagement: ${accountData.emailEngagement.openRate}% open, ${accountData.emailEngagement.clickRate}% click
- Last activity: ${accountData.webActivity.lastActivity}

CONTACTS:
${accountData.contacts.map(c => `- ${c.name} (${c.title}): ${c.engagementScore} engagement`).join('\n')}

RECENT NEWS:
${accountData.recentNews.map(n => `- ${n.headline}`).join('\n')}

COMPETITOR SIGNALS:
${accountData.competitorSignals.length > 0 ? accountData.competitorSignals.join('\n') : 'None detected'}

RELATIONSHIP HISTORY:
- Previous deals: ${accountData.deals.length}
- Total activities: ${accountData.activities.length}
- Last touch: ${accountData.activities[0]?.date || 'Never'}

Provide your analysis as JSON.`
    }],
    response_format: { type: 'json_object' }
  });
  
  const aiAnalysis = JSON.parse(qualitativeAnalysis.content[0].text);
  
  // Combine all scores
  return {
    companyId: accountData.company.id,
    companyName: accountData.company.name,
    scores: {
      icp: icpScore,
      engagement: engagementScore,
      intent: intentScore,
      opportunity: aiAnalysis.OPPORTUNITY_SCORE,
      value: aiAnalysis.VALUE_SCORE,
      timing: aiAnalysis.TIMING_SCORE
    },
    composite: calculateComposite({
      icp: icpScore,
      engagement: engagementScore,
      intent: intentScore,
      ...aiAnalysis
    }),
    insights: aiAnalysis.KEY_INSIGHTS,
    recommendedApproach: aiAnalysis.RECOMMENDED_APPROACH,
    tier: determineTier(/* composite score */)
  };
}

function calculateComposite(scores) {
  // Weighted combination
  return (
    scores.icp * 0.2 +
    scores.engagement * 0.15 +
    scores.intent * 0.25 +
    scores.OPPORTUNITY_SCORE * 0.2 +
    scores.VALUE_SCORE * 0.1 +
    scores.TIMING_SCORE * 0.1
  );
}

Step 4: Create the Daily Prioritized List​

Generate a ranked list for each rep every morning:

async function generateDailyPrioritization(repId) {
  // Get rep's assigned accounts
  const accounts = await crm.getAccountsByRep(repId);
  
  // Score all accounts (parallelize for speed)
  const scoredAccounts = await Promise.all(
    accounts.map(async account => {
      const data = await aggregateAccountData(account.id);
      return scoreAccount(data);
    })
  );
  
  // Sort by composite score
  const ranked = scoredAccounts.sort((a, b) => b.composite - a.composite);
  
  // Assign daily tiers
  const dailyList = {
    mustTouch: ranked.slice(0, 5).map(addContactReason),
    highPriority: ranked.slice(5, 15).map(addContactReason),
    standard: ranked.slice(15, 50).map(addContactReason),
    nurture: ranked.slice(50).map(addContactReason)
  };
  
  // Push to CRM and Slack
  await crm.updateDailyPriorities(repId, dailyList);
  await slack.sendDM(repId, formatPriorityList(dailyList));
  
  return dailyList;
}

function addContactReason(account) {
  return {
    ...account,
    whyNow: generateWhyNow(account),
    suggestedAction: getSuggestedAction(account),
    talkingPoints: getTalkingPoints(account)
  };
}

Real-World Example: Tech Company Prioritization​

Input: 500 accounts assigned to an SDR

AI Analysis Output (top 3):

[
  {
    "companyName": "CloudScale Inc",
    "composite": 94,
    "scores": {
      "icp": 92,
      "engagement": 88,
      "intent": 96,
      "timing": 98
    },
    "insights": [
      "CEO visited pricing page 3x this week",
      "Currently using Competitor X (known pain: data accuracy)",
      "Just closed Series B—scaling sales team is top priority"
    ],
    "recommendedApproach": "Reference Series B news, position as infrastructure for scaling sales team. CEO is actively evaluating—this is hot.",
    "whyNow": "Series B + active pricing page visits = buying now"
  },
  {
    "companyName": "DataFlow Systems",
    "composite": 87,
    "scores": {
      "icp": 95,
      "engagement": 75,
      "intent": 89,
      "timing": 82
    },
    "insights": [
      "VP Sales attended our webinar last week",
      "Hiring 5 SDRs according to LinkedIn",
      "No current solution in place"
    ],
    "recommendedApproach": "Reference webinar attendance, offer to help structure their new SDR team. Timing is good with their hiring push.",
    "whyNow": "Building SDR team from scratch = greenfield opportunity"
  },
  {
    "companyName": "NextGen Analytics",
    "composite": 84,
    "scores": {
      "icp": 88,
      "engagement": 91,
      "intent": 78,
      "timing": 75
    },
    "insights": [
      "3 different people from the company have downloaded content",
      "Tech stack includes Salesforce + Outreach",
      "Last contacted 6 months ago—went dark after demo"
    ],
    "recommendedApproach": "Re-engage with new angle. Multiple stakeholders engaged now vs. single contact before. Ask what's changed.",
    "whyNow": "Re-engagement opportunity with broader buying committee"
  }
]

Continuous Learning: The Feedback Loop​

The system improves by tracking outcomes:

async function logPrioritizationOutcome(accountId, outcome) {
  const originalScore = await getHistoricalScore(accountId);
  
  await analyticsDb.log({
    accountId,
    scoredAt: originalScore.timestamp,
    composite: originalScore.composite,
    outcome: outcome, // 'converted', 'stalled', 'lost', 'disqualified'
    daysToOutcome: daysBetween(originalScore.timestamp, new Date()),
    dealValue: outcome === 'converted' ? await getDealValue(accountId) : null
  });
  
  // Quarterly: Retrain weights based on what actually converted
  if (isQuarterEnd()) {
    await retrainScoringWeights();
  }
}

async function retrainScoringWeights() {
  const outcomes = await analyticsDb.getOutcomes({ months: 6 });
  
  // Analyze which factors actually predicted conversions
  const analysis = await claude.messages.create({
    model: 'claude-3-5-sonnet-20241022',
    messages: [{
      role: 'user',
      content: `Analyze these prioritization outcomes and recommend 
weight adjustments:

CONVERSIONS:
${outcomes.filter(o => o.outcome === 'converted').map(summarize).join('\n')}

LOSSES:
${outcomes.filter(o => o.outcome === 'lost').map(summarize).join('\n')}

Current weights: ${JSON.stringify(currentWeights)}

What factors were most predictive? Recommend new weights.`
    }]
  });
  
  // Update scoring algorithm
  await updateScoringWeights(analysis);
}

Integration with Daily Workflow​

Make prioritization seamless:

Morning Slack Notification​

// 7am daily
cron.schedule('0 7 * * *', async () => {
  const reps = await crm.getActiveReps();
  
  for (const rep of reps) {
    const priorities = await generateDailyPrioritization(rep.id);
    
    await slack.sendDM(rep.slackId, {
      blocks: [
        {
          type: 'header',
          text: `🎯 Your Priority Accounts for Today`
        },
        {
          type: 'section',
          text: `*Must Touch (5 accounts)*\n${priorities.mustTouch.map(a => 
            `• *${a.companyName}* (Score: ${a.composite}) — ${a.whyNow}`
          ).join('\n')}`
        },
        {
          type: 'actions',
          elements: [
            {
              type: 'button',
              text: 'View Full List',
              url: `https://crm.com/priorities/${rep.id}`
            }
          ]
        }
      ]
    });
  }
});

CRM Priority Field Updates​

async function syncToCRM(priorities) {
  for (const account of [...priorities.mustTouch, ...priorities.highPriority]) {
    await crm.updateCompany(account.companyId, {
      priority_tier: account.tier,
      ai_score: account.composite,
      last_scored: new Date(),
      recommended_action: account.suggestedAction,
      score_reasoning: account.insights.join(' | ')
    });
    
    // Create task if high priority
    if (account.tier === 'mustTouch') {
      await crm.createTask({
        companyId: account.companyId,
        subject: `Priority Touch: ${account.companyName}`,
        notes: account.whyNow,
        dueDate: new Date()
      });
    }
  }
}

Measuring Prioritization ROI​

Track these metrics:

The compound effect: If better prioritization increases conversions by 64% and deal size by 29%, and you're running 1,000 qualified accounts/quarter at a $30K baseline ACV, that's an additional $620K in ARR quarterly.

Advanced: Dynamic Reprioritization​

Don't just score once—reprioritize throughout the day:

// Real-time triggers
async function handleSignificantEvent(event) {
  const { accountId, eventType, data } = event;
  
  const significantEvents = [
    'pricing_page_visit',
    'competitor_search',
    'demo_request',
    'executive_engagement',
    'funding_announcement'
  ];
  
  if (significantEvents.includes(eventType)) {
    // Immediately rescore
    const newScore = await scoreAccount(await aggregateAccountData(accountId));
    
    // If jumped to Tier 1, alert immediately
    if (newScore.tier === 'mustTouch' && (await getPreviousTier(accountId)) !== 'mustTouch') {
      await sendUrgentAlert(accountId, newScore, event);
    }
  }
}

async function sendUrgentAlert(accountId, score, triggerEvent) {
  const rep = await crm.getAccountOwner(accountId);
  
  await slack.sendDM(rep.slackId, {
    text: `🚨 *HOT ACCOUNT ALERT*\n\n*${score.companyName}* just jumped to Tier 1!\n\nTrigger: ${triggerEvent.eventType}\n${score.whyNow}\n\nDrop what you're doing. This one's live.`
  });
}

Getting Started with MarketBetter​

Building AI account prioritization from scratch is powerful but complex. MarketBetter provides the complete solution:

• Daily SDR Playbook — Every rep gets their prioritized list each morning
• Real-time scoring — Accounts reprioritize based on live signals
• AI-powered reasoning — Not just a score, but why and what to do
• CRM integration — HubSpot, Salesforce out of the box
• Learning loop — Improves automatically based on your conversion data

Stop letting your best accounts go unworked. Stop wasting time on accounts that were never going to buy. Let AI tell you exactly where to focus.

Book a Demo →

Key Takeaways​

1. Poor prioritization costs deals — 67% of rep time goes to accounts that won't convert
2. Static lead scoring fails — Rules can't capture qualitative buying signals
3. Claude Code enables intelligent scoring — Synthesize structured + unstructured data
4. Make it actionable — Daily ranked lists with clear reasoning and suggested actions
5. Continuous learning — Track outcomes and retrain weights quarterly

Your CRM is full of gold. The problem is it's mixed in with thousands of accounts that look the same on the surface. AI-powered prioritization separates signal from noise—so your team spends 100% of their time on accounts that can actually close.

Your sales team just had a great discovery call. The prospect is ready for a proposal. Now comes the bottleneck: someone needs to spend 2-4 hours pulling together a customized deck with the right case studies, accurate pricing, and messaging that addresses this specific buyer's pain points.

What if that proposal could write itself?

With Claude Code and the right architecture, you can reduce proposal generation from hours to minutes—while actually increasing personalization. This guide shows you how to build an AI proposal generator that pulls context from your CRM, incorporates meeting notes, and produces polished documents ready for review.

Why Manual Proposals Kill Deal Velocity​

Proposals are a critical bottleneck in the sales cycle. Here's why:

Time Cost:

• Average proposal takes 2-4 hours to create
• Senior AEs spend 6-8 hours/week on proposals
• At $150K OTE, that's ~$18K/year per AE on document creation

Quality Variance:

• Junior reps produce weaker proposals than veterans
• Copy-paste errors creep in (wrong company names, outdated pricing)
• Generic messaging fails to address specific prospect concerns

Velocity Impact:

• Deals stall waiting for proposals
• Prospects go cold while documents are in progress
• Competitors who respond faster win the deal

The math is simple: faster proposals = higher close rates. Teams that respond to pricing requests within 1 hour are 7x more likely to close than those who wait 24+ hours.

The Anatomy of a Great Proposal​

Before automating, understand what makes proposals convert:

1. Personalization That Shows You Listened​

• References to specific pain points from discovery
• Industry-relevant examples and metrics
• Prospect's own language reflected back

2. Clear Value Narrative​

• Business impact, not feature lists
• ROI calculations specific to their situation
• Timeline to value that feels realistic

3. Social Proof That Resonates​

• Case studies from similar companies (size, industry)
• Relevant testimonials and metrics
• Recognizable logos when possible

4. Transparent Pricing​

• Clear breakdown of what's included
• Options that give them control
• Investment framed against expected return

5. Easy Next Steps​

• Single clear CTA
• Low-friction way to move forward
• Multiple contact options

Building the Proposal Generator with Claude Code​

Step 1: Design Your Proposal Schema​

First, define the structure Claude will generate:

interface Proposal {
  metadata: {
    prospectCompany: string;
    prospectContact: string;
    generatedDate: string;
    validUntil: string;
    version: string;
  };
  
  executiveSummary: {
    headline: string;
    painPointsSummary: string[];
    proposedSolution: string;
    expectedOutcomes: string[];
  };
  
  situationAnalysis: {
    currentState: string;
    challenges: Challenge[];
    businessImpact: string;
  };
  
  solution: {
    overview: string;
    capabilities: Capability[];
    implementation: ImplementationPlan;
  };
  
  socialProof: {
    caseStudies: CaseStudy[];
    testimonials: Testimonial[];
    relevantLogos: string[];
  };
  
  investment: {
    options: PricingOption[];
    comparison: string;
    roi: ROICalculation;
  };
  
  nextSteps: {
    cta: string;
    timeline: string[];
    contacts: Contact[];
  };
}

Step 2: Create the Context Gatherer​

Claude needs rich context to generate personalized proposals. Build a function that aggregates everything:

async function gatherProposalContext(dealId) {
  // Get CRM data
  const deal = await hubspot.getDeal(dealId, {
    associations: ['contacts', 'companies', 'meetings', 'notes']
  });
  
  // Get company info
  const company = deal.associations.companies[0];
  const companyData = {
    name: company.name,
    industry: company.industry,
    size: company.numberOfEmployees,
    revenue: company.annualRevenue,
    website: company.website,
    description: company.description
  };
  
  // Get meeting transcripts/notes
  const meetingNotes = deal.associations.meetings.map(m => ({
    date: m.meetingDate,
    notes: m.notes,
    attendees: m.attendees
  }));
  
  // Get relevant case studies from our database
  const caseStudies = await findRelevantCaseStudies({
    industry: company.industry,
    companySize: company.numberOfEmployees
  });
  
  // Get product/pricing info
  const productInfo = await getProductCatalog();
  const pricingTiers = await getPricingForDealSize(deal.amount);
  
  // Compile competitors mentioned
  const competitorMentions = extractCompetitorMentions(meetingNotes);
  
  return {
    deal,
    company: companyData,
    meetings: meetingNotes,
    caseStudies,
    products: productInfo,
    pricing: pricingTiers,
    competitors: competitorMentions
  };
}

Step 3: Build the Generation Prompt​

The prompt is where the magic happens. Here's a production-tested approach:

const PROPOSAL_SYSTEM_PROMPT = `
You are an expert B2B sales proposal writer. Your proposals have an 
exceptional win rate because you:

1. Lead with the prospect's specific pain points, using their exact language
2. Connect each capability to measurable business outcomes
3. Include relevant social proof (similar company size, industry)
4. Present pricing as an investment with clear ROI
5. Make next steps frictionless

STYLE GUIDELINES:
- Write in confident but not arrogant tone
- Use "you" and "your" heavily (prospect-focused)
- Avoid jargon unless the prospect used it first
- Keep sentences punchy—average 15 words
- Use numbers and specifics over generalities

FORMATTING:
- Output as JSON matching the Proposal interface
- Include 2-3 case studies maximum
- Provide 2-3 pricing options (good/better/best)
- Keep executive summary under 200 words
`;

async function generateProposal(context) {
  const response = await claude.messages.create({
    model: 'claude-3-5-sonnet-20241022',
    max_tokens: 8000,
    system: PROPOSAL_SYSTEM_PROMPT,
    messages: [{
      role: 'user',
      content: `Generate a proposal for the following opportunity:

PROSPECT COMPANY:
${JSON.stringify(context.company, null, 2)}

DEAL CONTEXT:
- Deal Size: $${context.deal.amount}
- Stage: ${context.deal.stage}
- Products of Interest: ${context.deal.products?.join(', ')}

MEETING NOTES (Discovery Insights):
${context.meetings.map(m => `
[${m.date}]
${m.notes}
`).join('\n---\n')}

AVAILABLE CASE STUDIES:
${JSON.stringify(context.caseStudies, null, 2)}

PRICING TIERS:
${JSON.stringify(context.pricing, null, 2)}

${context.competitors.length > 0 ? `
COMPETITORS MENTIONED:
${context.competitors.join(', ')}
(Address differentiators tactfully)
` : ''}

Generate a complete, personalized proposal.`
    }],
    response_format: { type: 'json_object' }
  });
  
  return JSON.parse(response.content[0].text);
}

Step 4: Transform to Final Format​

Claude outputs structured JSON. Now transform it to your preferred format:

async function renderProposal(proposalData, outputFormat = 'docx') {
  switch (outputFormat) {
    case 'docx':
      return await renderToWord(proposalData);
    case 'pdf':
      return await renderToPDF(proposalData);
    case 'slides':
      return await renderToSlides(proposalData);
    case 'notion':
      return await renderToNotion(proposalData);
    default:
      return proposalData; // Return raw JSON
  }
}

async function renderToWord(proposal) {
  const doc = new Document({
    sections: [{
      properties: {},
      children: [
        // Cover page
        new Paragraph({
          children: [
            new TextRun({
              text: `Proposal for ${proposal.metadata.prospectCompany}`,
              bold: true,
              size: 48
            })
          ],
          alignment: 'center'
        }),
        
        // Executive Summary
        new Paragraph({
          children: [
            new TextRun({ text: 'Executive Summary', bold: true, size: 32 })
          ]
        }),
        new Paragraph({
          children: [
            new TextRun({ text: proposal.executiveSummary.headline, bold: true })
          ]
        }),
        ...proposal.executiveSummary.painPointsSummary.map(point => 
          new Paragraph({ children: [new TextRun(`• ${point}`)] })
        ),
        
        // ... continue for all sections
      ]
    }]
  });
  
  return await Packer.toBuffer(doc);
}

Real-World Example: SaaS Proposal Generation​

Let's walk through a complete example:

Input Context:

• Company: Acme Corp (500 employees, Manufacturing, $80M revenue)
• Deal: $45,000/year platform subscription
• Discovery Notes: "SDR team of 12, currently using spreadsheets to track prospects. Manager complained they're 'flying blind' on pipeline. Asked about Salesforce integration."

Generated Proposal (excerpts):

{
  "executiveSummary": {
    "headline": "From Flying Blind to Crystal Clear: Transforming Acme's SDR Performance",
    "painPointsSummary": [
      "Your 12-person SDR team currently operates without centralized visibility—you described this as 'flying blind' on pipeline",
      "Spreadsheet-based tracking creates data silos and makes it impossible to spot trends or coach effectively",
      "Without integrated systems, reps waste hours on manual data entry instead of selling"
    ],
    "proposedSolution": "MarketBetter's AI-powered SDR platform consolidates your prospecting, outreach, and pipeline management into a single source of truth—with native Salesforce integration you specifically asked about.",
    "expectedOutcomes": [
      "50% reduction in time spent on manual data entry",
      "Real-time pipeline visibility for managers",
      "15-20% increase in meetings booked per rep"
    ]
  },
  
  "socialProof": {
    "caseStudies": [
      {
        "company": "Precision Manufacturing Co",
        "industry": "Manufacturing",
        "size": "450 employees",
        "challenge": "SDR team working off disconnected spreadsheets",
        "result": "67% increase in pipeline visibility, 23% more meetings in first quarter",
        "quote": "Finally, I can see what my team is actually doing without asking for status updates."
      }
    ]
  },
  
  "investment": {
    "options": [
      {
        "name": "Growth",
        "seats": 12,
        "annual": 36000,
        "features": ["Core platform", "Basic automation", "Standard integrations"],
        "recommendation": false
      },
      {
        "name": "Scale",
        "seats": 12,
        "annual": 45000,
        "features": ["Core platform", "Advanced automation", "Salesforce integration", "Priority support"],
        "recommendation": true,
        "whyRecommended": "Includes the Salesforce integration you specifically need"
      },
      {
        "name": "Enterprise",
        "seats": 12,
        "annual": 60000,
        "features": ["Everything in Scale", "Dedicated CSM", "Custom reporting", "API access"],
        "recommendation": false
      }
    ],
    "roi": {
      "currentCostOfInefficiency": "$15,000/month in lost productivity",
      "expectedSavings": "$8,000/month",
      "paybackPeriod": "6 months"
    }
  }
}

Handling Edge Cases​

Multiple Stakeholders​

When proposals need to address different personas:

async function generateMultiStakeholderProposal(context) {
  const stakeholders = context.meetings
    .flatMap(m => m.attendees)
    .filter(a => a.role !== 'our_team');
  
  // Identify personas
  const personas = await claude.analyze({
    messages: [{
      role: 'user',
      content: `Categorize these stakeholders:
                ${JSON.stringify(stakeholders)}
                
                Categories: Executive, Finance, Technical, End-User`
    }]
  });
  
  // Generate proposal with persona-specific sections
  return generateProposal({
    ...context,
    stakeholderPersonas: personas,
    additionalInstructions: `
      Include these targeted sections:
      - Executive Summary (for ${personas.executive?.name})
      - Technical Specifications (for ${personas.technical?.name})
      - ROI Analysis (for ${personas.finance?.name})
    `
  });
}

Competitive Situations​

When prospects mention competitors, address it tactfully:

if (context.competitors.includes('competitor-x')) {
  context.additionalInstructions += `
    The prospect mentioned evaluating Competitor X. Include:
    - A brief, factual comparison (no FUD)
    - Differentiation on the specific pain points they mentioned
    - A case study of a customer who switched from Competitor X
    
    Keep comparison professional—never bash the competitor.
  `;
}

Custom Pricing Requests​

When standard tiers don't fit:

if (context.deal.customPricingRequested) {
  const customPricing = await calculateCustomPricing({
    baseSeats: context.deal.seatCount,
    addOns: context.deal.requestedAddOns,
    term: context.deal.contractTerm,
    volume: context.company.numberOfEmployees
  });
  
  context.pricing = {
    custom: true,
    breakdown: customPricing,
    flexibility: 'Pricing reflects your specific requirements. Let\'s discuss if anything needs adjustment.'
  };
}

Integration with Your Workflow​

Trigger: CRM Stage Change​

// HubSpot workflow trigger
app.post('/webhooks/hubspot/deal-stage-change', async (req, res) => {
  const { dealId, newStage } = req.body;
  
  if (newStage === 'proposal_requested') {
    // Gather context
    const context = await gatherProposalContext(dealId);
    
    // Generate proposal
    const proposal = await generateProposal(context);
    
    // Render to PDF
    const pdf = await renderProposal(proposal, 'pdf');
    
    // Save to deal
    await hubspot.uploadFile(dealId, pdf, 'proposal.pdf');
    
    // Notify rep
    await slack.notify(context.deal.owner, {
      text: `📄 Proposal generated for ${context.company.name}`,
      actions: [
        { text: 'Review', url: `https://app.hubspot.com/deals/${dealId}` },
        { text: 'Send to Prospect', callback: 'send_proposal' }
      ]
    });
  }
  
  res.sendStatus(200);
});

Human-in-the-Loop Review​

Always allow reps to review before sending:

async function queueForReview(proposal, dealId) {
  // Create review task
  await hubspot.createTask({
    dealId,
    subject: 'Review Generated Proposal',
    priority: 'HIGH',
    notes: `AI-generated proposal is ready for review.
            
            Check:
            - [ ] Pain points accurately captured
            - [ ] Pricing correct
            - [ ] Case studies relevant
            - [ ] No copy/paste errors
            
            Make edits directly in the attached document.`,
    dueDate: addHours(new Date(), 4)
  });
}

Measuring Success​

Track these metrics to quantify proposal automation ROI:

The compounding effect is significant. If faster proposals increase close rates by just 6%, and your reps can produce 4x more proposals, the revenue impact is dramatic.

Getting Started with MarketBetter​

Building your own proposal generator is powerful, but it takes time. MarketBetter offers proposal automation as part of the complete AI SDR platform:

• One-click proposals from any deal in HubSpot or Salesforce
• Smart case study matching based on prospect industry and size
• Dynamic pricing that pulls from your CPQ configuration
• Brand-compliant templates that match your company guidelines
• Version tracking so you know what was sent when

Combined with AI lead research, automated follow-ups, and pipeline monitoring, it creates a system where proposals are generated in the flow of work—not a bottleneck that delays them.

Book a Demo →

Key Takeaways​

1. Manual proposals waste senior AE time — $18K/year per rep on document creation
2. Speed wins deals — Responding in 1 hour vs 24 hours increases close rate 7x
3. Claude Code enables intelligent generation — Pull CRM data + meeting context + case studies
4. Structure matters — Define schemas so output is consistent and renderable
5. Always human-in-the-loop — AI generates, humans approve and send

Your proposals are often the first professional deliverable a prospect sees. Make sure they're personalized, polished, and prompt. With AI, you can have all three.

Your SDRs spend just 35% of their time actually selling. The rest? Research, data entry, writing emails, prepping for calls. Both Claude and ChatGPT promise to automate this busywork—but they take different approaches.

After running both AIs on real sales workflows at MarketBetter (and building an AI SDR with OpenClaw), here's what we learned about when to use each.

GPT-5.3-Codex dropped three days ago. Claude Code has been the go-to for AI-powered development. If you're building sales automation, which one should you use?

The honest answer: Both. For different things.

This isn't a "which is better" post. It's a practical guide to when each tool excels—specifically for GTM use cases.

The Core Difference​

Here's the fundamental distinction:

OpenAI Codex is optimized for building software. It excels at:

• Writing code from scratch
• Multi-file refactoring
• Creating integrations
• Building applications

Claude Code is optimized for reasoning and judgment. It excels at: