How to Use Claude for Sales: A Role-by-Role Guide (SDR to VP)

Most “Claude for sales” advice is written as if every seller does the same job. They do not. An SDR needs volume with personalization, an AE needs deal strategy, a sales engineer needs technical translation, and a VP needs pattern recognition across the whole pipeline. The prompts that transform one role are noise for another.

So this guide is organized by role. For each one: what Claude is actually for in that seat, a prompt you can run today, the pro move that separates power users, and the mistake that wastes the tool. At the end, the two upgrades that change everything: connecting live data over MCP and encoding your playbook as skills.

TL;DR

Role	Use Claude for	Not for
SDR	Personalized outreach at scale, prospect research	Sending drafts untouched
BDR	Trigger events, account plans, persona targeting	Working from model memory
AE	Discovery prep, objection simulation, post-call analysis	Pre-call use only
Sales Engineer	Technical-to-business translation, ROI models	Generic feature dumps
Sales Manager	Call scoring, coaching plans, at-risk deal reviews	Admin tasks only
VP of Sales	Win/loss patterns, forecast interrogation, playbook design	Report summarization
RevOps / Sales Ops	Territory models, comp-plan stress tests, funnel diagnosis	Vague ‘fix our funnel’ asks

For SDRs: volume without the template smell

The SDR job is personalization at a scale no human sustains manually. Claude’s job is the 80%: research the prospect, find the angle, draft the variants. Your job is the 20% that proves a human did the work.

Write 8 cold email variants for a [title] at a [size] [industry]
company. My product solves [problem]. Under 80 words each,
each from a different angle: trigger event, peer proof, cost of
inaction, contrarian take, question-only, one-line, before/after,
and referral-style. No "I hope this finds you well." No em-dashes.

The pro move: Paste the prospect’s LinkedIn About section and their last three posts, then ask for an opening line that references what they actually care about. Thirty seconds per prospect, and it reads nothing like a mail merge.

The common mistake: Sending Claude’s drafts untouched. Every model has a detectable default voice; buyers have read it a thousand times. Edit for the one detail only you could know.

For BDRs: trigger events, not lists

Account-based outbound lives on timing. Claude’s job is finding the reason to reach out this week rather than any week.

I am targeting [company]. Find 3 trigger events from the last 90
days I can use as outreach angles: funding, leadership changes,
product launches, expansions, layoffs, earnings commentary.
For each: why it matters to a [your category] conversation and
which persona to contact first.

The pro move: Run the four-prompt discovery-prep chain: raw research, then filter through what you sell, then talk tracks and questions, then landmines to avoid. That last step, asking what NOT to bring up (sensitive layoffs, bad press, failed initiatives), is the one everyone skips and the one that saves calls.

The common mistake: Doing this from model memory. Claude’s training data is months old; a trigger event from last week only shows up if Claude can search or query live data. This is exactly what connecting a research tool over MCP fixes.

For AEs: the work between the calls

AEs get the most from Claude on either side of the meeting: prep going in, analysis coming out.

I have a discovery call tomorrow with a [title] at [company].
They are evaluating us against [competitor]. Give me 5 questions
that uncover real pain and differentiate us without naming the
competitor. For each, what a good answer and a bad answer
sound like for my deal.

The pro move: After every call, paste your notes or transcript and ask: “What did this buyer reveal about their decision criteria, timeline, and internal politics? What did I assume that they did not actually say? What should I ask next?” The gap between what you heard and what was said is where deals slip.

The common mistake: Only using Claude before calls. The post-call analysis is the bigger unlock, and almost nobody does it.

For Sales Engineers: translate, quantify, disarm

The SE’s craft is turning technical capability into business outcome for two audiences at once: the engineer who evaluates and the executive who signs.

My prospect's engineering team cares about [requirement].
Write a one-pager showing how our [feature] solves it, with 3
quantified benefits and the technical caveats an honest vendor
would disclose. Then a 3-sentence executive summary of the
same thing for their CFO.

The pro move: Ask Claude to build an interactive ROI calculator as an Artifact: deal-specific inputs (team size, current costs, time savings), live outputs, shareable with your champion. Your champion re-runs it internally with their own numbers, which means they are selling for you.

The common mistake: Generic feature dumps. Technical buyers detect boilerplate instantly. Give Claude the prospect’s actual environment and constraints, or do not bother.

For Sales Managers: coach from patterns, not anecdotes

A manager’s scarce resource is attention. Claude’s job is telling you where to point it.

Here is my team's pipeline export. Which deals are at risk based
on days in stage, last activity, and deal size relative to rep
average? For each at-risk deal, give me 3 coaching questions to
ask the rep, not statements to tell them.

The pro move: Upload call transcripts and have Claude score each rep on discovery depth, objection handling, and next-step setting, then build the coaching plan from the gaps. One hour of setup replaces the ride-along schedule you never keep.

The common mistake: Using Claude for meeting agendas and admin. The leverage is pattern recognition across your team’s calls and deals, the thing no manager has time to do manually.

For VPs of Sales: interrogate the funnel

At VP level Claude is a strategic thinking partner with infinite patience for your data.

Here are our last 4 quarters of pipeline data. What patterns do
you see in win rates by segment, deal size, and cycle length?
Where are we leaking revenue? Rank your 3 highest-confidence
findings and tell me what data would falsify each one.

The pro move: Upload win/loss data and ask for the top three reasons you lose and the process change that would fix each. Then the harder question: “Challenge my current playbook with this data. What am I rewarding that I should not be?”

The common mistake: Treating Claude as a reporting tool that summarizes. Ask it to challenge your assumptions; agreement is not analysis.

For RevOps and Sales Ops: the modeling partner

Ops teams get Claude’s most underrated capability: structured modeling with explained tradeoffs.

Here is our territory data. Model 3 territory splits optimized for:
equal opportunity, geographic efficiency, and account potential.
Show the tradeoffs of each, which reps win and lose under each
model, and which split you would pick and why.

The pro move: Paste your comp plan and ask: “Where are the perverse incentives? What behaviors does this plan accidentally reward or punish?” A second opinion before rollout is cheaper than a quarter of sandbagging. Same trick for lead scoring: upload marketing-qualified leads and closed-won data together and ask where the definitions disagree.

The common mistake: Asking Claude to build dashboards or fix “the funnel.” It works with what you give it: export, clean, upload, then ask a specific question about a specific stage and metric.

The two upgrades that change everything

1. Connect live data (MCP)

Every workflow above gets sharper when Claude can query instead of recall. Over MCP, Claude reads your CRM directly (HubSpot’s connector, Salesforce through Agentforce) and pulls verified contact data: Salesgear’s server exposes search across 800M+ contacts, enrichment, and per-prospect deep research. The BDR trigger-event prompt stops being a memory exercise and becomes a live query. We ranked the 7 servers worth connecting and published 5 full CRM workflows with prompts and guardrails.

2. Encode the playbook (skills)

Prompts in this guide are per-conversation. A skill file makes them permanent: your ICP definition, voice rules, stage conventions, written once and applied every time, by every rep, identically. We published 5 downloadable skill files for sales teams. And if you want these workflows running on a schedule instead of on demand, that is Claude Cowork.

Where Claude stops and infrastructure begins

Claude drafts, researches, analyzes, and models. It does not manage deliverability, warm up inboxes, sequence across email and calls and LinkedIn tasks, detect replies, or verify an address at send time. That is execution infrastructure, and it lives in a sales engagement platform. The teams doing this well use Claude for judgment and the platform for execution, with the same data underneath both.

Frequently asked questions