Skip to main content
Updated July 26, 2026
Most of a selling week disappears into follow-ups, CRM updates, and meeting prep that never touches a customer. These are the eight AI use cases that hand that work to an agent, ranked by the time they give back.

All eight run on whichever AI agent you already use - ClaudeClaude, ChatGPTChatGPT, GeminiGemini, or Microsoft CopilotMicrosoft Copilot - connected to the tools listed with each use case.

Best AI Sales Use Cases


Automate meeting follow-ups

Best first workflow - the fastest payoff for most sales teams
4 hr/wkest. time saved
How it’s done today

After every call, you rewrite the same conversation three ways: a follow-up email, a CRM note, and a task list. Between back-to-back meetings it gets rushed, delayed, or skipped.

How AI helps

When the transcript is ready, your agent pulls out what was agreed and leaves ready-to-review drafts in the tools you already use.

GongCall endstranscript ready
Your AI agent

Pulls decisions, commitments, owners, dates, and next steps from the conversation.

Follow-up email Gmail
CRM note Salesforce
Tasks Asana
How to set it up

Required

GongGranolaMeetings

Reads the finished transcript with speakers, attendees, and timing.

Recommended

SalesforceHubSpotCRM

Reads accounts, contacts, and open deals to match the call.

Writes the call note and tasks, prepared for your review.

GmailOutlookEmail

Writes draft follow-up emails only - sending stays with you.

Skip the CRM and email connections and the agent still prepares everything - you just paste it in. Optional: a task manager if action items live outside your CRM, or Slack for a ping when drafts land.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have a past follow-up email you were happy with, one CRM note in your usual format, and three recent call transcripts ready before you start.

Setup prompt

What good looks like

Across the three test calls the email should need light edits at most, the note should land on the right account, and no owner or date should appear that was not said on the call. If it misses, tell it exactly what went wrong - too long, wrong tone, invented a date - and rerun the same three calls.

Choose your trigger

Trigger it on each new completed transcript if your recorder supports that event; otherwise check every 15–30 minutes. Restrict it to calls with at least one external attendee, or your standups and interviews start producing customer follow-up drafts too. If your recorder cannot filter by attendee, filter by calendar title and exclude anything marked internal.

What runs without you

Review every draft for the first five calls. Once four of five are sendable with light edits, let the CRM note and tasks write automatically and keep reviewing the emails. The customer email stays a draft permanently - that is not a phase you graduate from. Once a week, compare drafts created to calls held; a stopped automation does not announce itself.

Pairs well with automate pipeline hygiene and generate call coaching briefs - both reuse the transcripts you just connected.

Draft personalized outreach

Best for first touches that should feel researched, not templated
3 hr/wkest. time saved
How it’s done today

For every prospect, you research the account, choose a useful angle, find approved proof, and rebuild the message from scratch.

How AI helps

When a lead qualifies or a buying signal appears, your agent researches the person and company, picks one real reason to reach out, and leaves a concise draft in your email tool.

ApolloLead qualifiesor buying signal appears
Your AI agent

Researches the person and company, picks one useful angle, and applies your approved positioning.

Email draft Gmail
CRM activity Salesforce
How to set it up

Required

SalesforceHubSpotCRM

Reads leads, contacts, and lifecycle stage to pick who qualifies.

Writes the outreach activity, logged on the contact.

Recommended

ApolloClayResearch

Reads role, company, and recent buying signals for the angle.

GmailOutlookEmail

Writes draft outreach emails only - sending stays with you.

No enrichment tool? The agent researches from the public web instead - slower and shallower, but enough to test whether the drafts earn replies. Optional: a docs workspace holding your approved positioning and proof points.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have two or three approved messages that show your tone, and a list of ten accounts you know well ready before you start.

Setup prompt

What good looks like

Every company fact should be verifiable, the reason for reaching out should be specific to that account, and the claim should be one you approved. If a draft could have been sent to anyone, tell it so and rerun the same three prospects.

Choose your trigger

Trigger it when a lead meets your qualification rule, or run a capped daily batch over a saved CRM segment. Cap it - an uncapped outreach workflow is how you end up with a hundred drafts and no time to review them. Exclude existing customers and anyone already in an active conversation.

What runs without you

Review every draft for the first two weeks. Once the research is consistently right, let it log the CRM activity automatically and keep reviewing the messages. Sending always stays with you - outreach is where a wrong fact costs you the account. Once a week, check that drafts are still appearing for qualified leads.

Pairs well with prioritize accounts automatically - same CRM and research connections, and it tells you who to write to.

Draft proposals automatically

Best for deals that stall while the paperwork gets written
2 hr/wkest. time saved
How it’s done today

You move discovery context into a template by hand, hunt for relevant proof, and chase missing commercial, scope, and legal inputs deal by deal.

How AI helps

When a proposal is requested, your agent combines the discovery notes, the CRM record, and your approved content into a first draft - and visibly marks every gap instead of guessing.

SalesforceProposal requestedor deal reaches stage
Your AI agent

Maps discovery and approved content into your template and marks every missing fact or approval.

Proposal draft Google Docs
Approval tasks Asana
How to set it up

Required

SalesforceHubSpotCRM

Reads the deal, products, amounts, and discovery context.

Writes the proposal link and status back on the deal.

Recommended

Google DocsMicrosoft WordDocs

Reads your template, approved product language, and proof points.

Writes the proposal first draft.

No connections? Upload the transcript, template, and approved content by hand - same draft quality, manual trigger. Optional: your meeting assistant for discovery transcripts, or file storage for prior proposals.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have your current proposal template, your approved product language and proof points, and two past proposals you were happy with ready before you start.

Setup prompt

What good looks like

The recreated drafts should match the structure and claims of what you actually sent, with every missing price, term, or approval visibly marked rather than guessed. If it invented anything, say which line and rerun the same two proposals.

Choose your trigger

Trigger it from a short proposal-request form or from the opportunity stage change that means “we are proposing.” A form is usually better - it captures the scope questions the CRM record does not. Exclude renewals and anything using a standard order form.

What runs without you

Review every draft for the first month - proposals are low-volume and high-stakes, so there is no rush to hand this over. Once the structure is reliable, let it assign the approval tasks and update the CRM link automatically. The draft never goes to the customer without a human read. Check monthly that requests are still producing drafts.

Pairs well with prepare forecast reviews - same CRM connection, and proposals in flight are exactly what the forecast conversation is about.

Generate pre-meeting briefs

Best for days of back-to-back external calls
1 hr/wkest. time saved
How it’s done today

You dig through the CRM, email threads, prior call notes, and company pages in the minutes before a customer conversation - or walk in cold.

How AI helps

Before every external meeting, your agent matches the attendees to the account, gathers what happened since last time, and delivers a one-page brief you can scan in two minutes.

Google CalendarMeeting soon45 minutes away
Your AI agent

Matches the attendees, gathers recent account context, and drafts the questions worth asking.

Meeting brief Slack
How to set it up

Required

Google CalendarOutlookCalendar

Reads upcoming external meetings with attendees and timing.

Recommended

SalesforceHubSpotCRM

Reads the account, contacts, open deals, and recent activity.

Read access is all this needs - it writes nothing anywhere. Optional: email for recent threads, or your meeting assistant for what was said on the last call.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have three past meetings you remember well, to test against ready before you start.

Setup prompt

What good looks like

Each brief should be scannable in two minutes, name the right people and deal stage, and surface something you would have had to dig for. If it padded with history you already knew, say so and rerun the same three meetings.

Choose your trigger

Run it 30–60 minutes before each external meeting, or as one morning digest covering the day - offer both and let each rep pick. Skip internal, personal, and cancelled events, and skip meetings with no external attendee.

What runs without you

This one writes nothing anywhere, so there is little to hand over - the whole workflow is read-only by design. What changes with time is trust: after a week you will stop opening the CRM to double-check it. Once a week, confirm briefs are still arriving for the meetings that have them.

Pairs well with automate meeting follow-ups - you already have the calendar and CRM; adding a meeting assistant covers both ends of the call.

Automate pipeline hygiene

Best when the CRM is always three updates behind reality
1 hr/wkest. time saved
How it’s done today

Before every review, someone - usually you - scans the whole open pipeline for missing fields, stale next steps, contradictory dates, and deals with no recent evidence.

How AI helps

On a schedule, your agent checks every open deal against your freshness and completeness rules, and returns a cleanup queue with the issue and a proposed correction for each.

SalesforceNightly scanall open deals
Your AI agent

Checks fields, activity, next steps, close dates, and stage age against your rules.

Cleanup queue Slack
Proposed updates Salesforce
How to set it up

Required

SalesforceHubSpotCRM

Reads every open deal with fields, activity, next steps, and notes.

Writes proposed corrections and owner questions, prepared for review.

Start with read access and a separate cleanup queue; grant write access once the checks prove reliable. Optional: your meeting assistant, to catch deals whose recent calls contradict the CRM.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have one team’s live pipeline, and a few deals you know are clean and a few you know are stale ready before you start.

Setup prompt

What good looks like

Known-clean deals should stay unflagged and known problems should appear with evidence you can act on. If it flags healthy deals, name the rule that misfired and rerun the same pipeline - noisy rules are the reason these workflows get ignored.

Choose your trigger

Run it nightly after your CRM activity has synced, plus once a couple of hours before the weekly pipeline review. Scope it to open deals owned by the team you are testing with before widening it - a first run across the whole company produces a queue nobody reads.

What runs without you

Start with read access and a separate cleanup queue. Once the rules stop misfiring, let it write the safe corrections - missing next-step dates, stale activity flags - and keep stage, amount, and close-date changes as questions to the owner. Those three always need a human, because they move the forecast. Check weekly that the queue is still being generated.

Pairs well with prepare forecast reviews - a clean pipeline is what makes the forecast brief trustworthy.

Prepare forecast reviews

Best for managers who rebuild the same deck every Monday
1 hr/wkest. time saved
How it’s done today

You compare pipeline snapshots and read deal notes by hand to figure out which deals moved, which are at risk, and what to ask before the forecast call.

How AI helps

Before the call, your agent compares the current pipeline with the previous snapshot, flags the deals that materially changed, and prepares the question worth asking on each.

SalesforceBefore the callcurrent vs. previous
Your AI agent

Finds material changes, reads the recent deal context, and prepares one question per flagged deal.

Forecast brief Google Docs
How to set it up

Required

SalesforceHubSpotCRM

Reads the pipeline with stages, amounts, close dates, forecast categories, and notes.

Two exports work just as well - this week’s pipeline and last week’s are all the comparison needs. Optional: a spreadsheet holding prior snapshots or manager adjustments.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have this week’s pipeline and last week’s, and a past forecast call you remember ready before you start.

Setup prompt

What good looks like

The brief should catch the deals you actually discussed last week and skip the routine movement. If it surfaced noise, raise the material-change threshold and recreate the same week again.

Choose your trigger

Run it one or two hours before the weekly forecast call. In the last two weeks of the quarter, add a daily run that reports only what changed since the previous brief. Scope it to the team whose forecast you are reviewing, not the whole company.

What runs without you

This one produces a brief for a human conversation, so there is nothing to hand over completely - the output is the input to a discussion. What you can automate over time is the delivery: once the thresholds are right, let it post to the forecast owner privately without you triggering it. Judgments about commit and risk stay with you. Check weekly that the brief arrived before the call.

Pairs well with automate pipeline hygiene - run the cleanup the night before and the forecast brief has better data to work with.

Prioritize accounts automatically

Best for reps with more accounts than hours
0.5 hr/wkest. time saved
How it’s done today

You scan your book across CRM records, recent activity, usage signals, and renewal dates to decide what deserves attention - and mostly run on gut feel.

How AI helps

On a schedule, your agent ranks the accounts you own against a small set of buying, engagement, and timing signals - and shows the evidence and one next action for each.

SalesforceMorning refreshaccount data updated
Your AI agent

Applies your buying, engagement, and timing signals across every account you own.

Ranked action queue Slack
How to set it up

Required

SalesforceHubSpotCRM

Reads accounts, owners, open deals, renewal dates, and recent activity.

A current CRM export works if you would rather not connect it yet. Optional: enrichment sources for buying signals, or Slack to deliver each person’s queue privately.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have a live territory with a mix of urgent, healthy, and incomplete accounts ready before you start.

Setup prompt

What good looks like

The top ten should overlap heavily with the accounts you would have picked, and every recommendation should point to a real date or activity. Where you disagree, say why and retune the signals before you move on.

Choose your trigger

Refresh it each morning for active territories, or every Monday for larger books. Replace the previous queue rather than appending, and timestamp it - a stale priority list is worse than none. Scope it to accounts you actually own.

What runs without you

This one only recommends, so it never writes to your CRM unless you decide it should. Once the ranking matches your judgment, let it deliver the queue automatically each morning. The decision about where to spend the day stays with you - the queue is an argument, not an instruction. Check weekly that the queue refreshed.

Pairs well with draft personalized outreach - the queue tells you who to contact, and the outreach workflow drafts the message.

Generate call coaching briefs

Best for coaching at scale without replaying every call
0.5 hr/wkest. time saved
How it’s done today

You listen to a small sample of calls, take notes by hand, and struggle to keep coaching coverage consistent across the team.

How AI helps

Your agent reviews each selected call against your coaching rubric, finds the timestamped moments that matter, and prepares a private brief with one practice exercise.

GongCall selectedtranscript ready
Your AI agent

Scores the call against your rubric and picks the timestamped moments worth coaching.

Coaching brief Slack
How to set it up

Required

GongGranolaMeetings

Reads call transcripts with speakers and timestamps.

One connection is enough to start - paste individual transcripts if you want to trial the rubric first. Optional: Slack or Teams to deliver each brief privately to you.

Paste this into Claude, ChatGPT, or your automation tool. It interviews you, builds the workflow, and tests it before anything goes live - have three calls a manager has already reviewed - one strong, one weak, one mixed - and your coaching rubric if you have one written down ready before you start.

Setup prompt

What good looks like

The briefs should pick the same moments you would have picked and quote them accurately. If the advice is generic or the timestamps are wrong, say so and recalibrate on the same three calls before using it on anyone else.

Choose your trigger

Sample about three eligible calls per rep per week and run it before the next one-to-one. Restrict it to external customer calls above a minimum length - internal syncs and two-minute callbacks produce useless briefs and waste the sample.

What runs without you

Every brief lands with you first. Once they are consistently fair, let it run on the weekly sample without anyone triggering it - but a coaching brief never goes straight to a rep without you reading it first. Coaching is a conversation, not a report card. Check weekly that the sample is still being drawn.

Pairs well with automate meeting follow-ups - same transcripts, and one call then produces both the customer follow-up and the coaching evidence.

How to choose

  • Start with meeting follow-ups if calls create repeated email, task, and CRM work - it pays off fastest for reps and managers alike.
  • If you carry a quota: pre-meeting briefs prep each conversation, personalized outreach and proposals cut drafting time, and account prioritization decides where the day goes.
  • If you manage the team: forecast reviews and call coaching briefs are the manager workflows - pipeline evidence for Monday and fair coaching without replaying every call.
  • If your CRM is the mess: run pipeline hygiene first - clean records make every other workflow on this list more trustworthy.

What didn’t make the list (yet)

Two sales use cases are marketed heavily and deliberately missing here: Autonomous outbound (AI SDRs) - agents that send cold email without review - have the best-documented failure record in sales AI: publicly reported campaigns with thousands of sends and zero replies, and burned sender domains. What works is draft personalized outreach: AI researches and drafts, you send. AI pricing and discounting - letting an agent decide what to quote - has too little evidence and too much downside for a blanket recommendation. If a use case earns its way onto this list, we will add it with the same setup steps.

Frequently Asked Questions

For these use cases you don’t need a sales-specific AI tool - a general AI agent (Claude, ChatGPT, Gemini, Microsoft Copilot) connected to your CRM, meeting recorder, and email covers all eight. Specialized sales tools matter one level deeper, as the recording, enrichment, and delivery layers - they are the integrations, not the brain.
Yes - every use case here is agent-agnostic. The setup prompts in each card work in Claude (Cowork mode), ChatGPT (Work), Gemini, and Microsoft Copilot; what varies is which connectors your workspace has approved.
Use your company’s paid workspace plan, not a personal account - the major AI vendors’ business tiers do not train on your data by default. Check with IT before connecting call recordings, and keep customer-facing sends human-reviewed.
The practitioner consensus is compression, not replacement: AI removes the admin around selling, and the humans still do the selling. That is why every use case on this page drafts and prepares - none of them talk to your customers.
Deliberately - see what didn’t make the list. Autonomous cold outreach has the best-documented failure record in sales AI; the pattern that works is AI drafts, you send.