AI agents for sales teams

The average sales rep spends more time updating their CRM and writing follow-up emails than actually talking to prospects. That ratio is backwards, and agent skills can flip it.

I watched a sales team track their time for a week. Out of a 40-hour workweek, the top performer spent 11 hours on actual selling conversations. The rest went to data entry, research, email drafts, internal reporting, and hunting through LinkedIn. The least productive rep? Six hours of selling. The admin work was eating everything.

Agent skills won’t close deals for you. They’re terrible at reading a room, building trust, or knowing when to push and when to back off. But they’re absurdly good at the other 29 hours of your week. Let’s go through the workflows where they make the biggest difference.

Lead enrichment before every call

You’ve got a call with someone at Acme Corp in 20 minutes. What do you actually know about them? Their name, their title, maybe their company’s website. If you’re disciplined, you spent 10 minutes on LinkedIn. If you’re busy, you’re going in blind and hoping to figure things out from the conversation.

An agent can build you a one-page brief in under a minute. Here’s the prompt I’ve seen work best:

“Research this prospect for an upcoming sales call. Name: Sarah Chen, VP of Engineering at DataFlow Inc. Find: company size and recent funding, their tech stack (check job postings and their engineering blog), any recent news or press releases from the last 90 days, Sarah’s recent LinkedIn activity or conference talks, and any mutual connections I might have context on. Summarize everything in a one-page brief with the most important conversation starters at the top.”

The output gives you something to actually talk about. “I saw you just raised a Series B, congratulations. Are you scaling the engineering team?” is a completely different opening than “So, tell me about your company.”

For teams doing high-volume outbound, you can run this in batch. Give the agent a list of 20 prospects, get 20 briefs back, and sort them by which ones have the most interesting conversation hooks. That sorting step alone saves you from wasting calls on leads where you have nothing to work with.

CRM hygiene nobody wants to do

Every CRM turns into a graveyard of half-filled records, duplicate contacts, and deals that have been “closing next month” for six months. Nobody wants to clean it up because it’s mind-numbing work. But dirty CRM data means bad forecasts, missed follow-ups, and embarrassing moments when two reps call the same person.

An agent can audit your CRM data and generate a cleanup report. You’ll need to export your data (most CRMs let you export to CSV), then run something like this:

“Here’s a CSV export of our CRM deals. Analyze it and create a cleanup report. Find: contacts with missing phone numbers or email addresses, duplicate records (same person at same company with slightly different names), deals marked as ‘negotiation’ or ‘proposal’ with no activity in the last 21 days, contacts with job titles that look outdated (like ‘intern’ for someone we’ve been talking to for 3 years), and any deals with a close date in the past that are still marked as open. For each issue, give me the record ID and a suggested action.”

This works because it’s pattern matching across structured data, which is exactly the kind of thing agents handle well. You’ll still need to make judgment calls on what to do with the results. Maybe that stale deal actually has an informal verbal commitment you haven’t logged. But at least now you know it exists.

Run this monthly. Set a calendar reminder. Your pipeline accuracy will improve noticeably after the second cleanup.

Proposal and quote drafting

First drafts of proposals are 80% boilerplate. The company overview, the methodology section, the “why us” page. They barely change between deals. The 20% that matters, the specific scope, pricing, and timeline, is what you should be spending your time on.

Here’s how to get a solid first draft:

“Draft a proposal for DataFlow Inc based on these inputs. Our product: CloudSync, an enterprise data integration platform. Prospect requirements: they need to connect 12 data sources including Snowflake, PostgreSQL, and three REST APIs. They want real-time sync with under 5-minute latency. Their budget range is $80K-$120K annually. Our pricing: base platform is $60K/year, each additional connector beyond 5 is $4K/year, real-time sync add-on is $20K/year. Write an executive summary, proposed solution section, pricing breakdown, implementation timeline (assume 6-8 weeks), and next steps. Keep it under 4 pages. Tone should be confident but not pushy.”

The agent will produce something you can actually send after 20 minutes of editing. Without it, you’re looking at an hour of writing from scratch, or worse, copy-pasting from the last proposal and forgetting to change the company name. We’ve all done it. It’s not a good look.

For teams that do a lot of proposals, keep a document with your standard positioning, differentiators, and case study summaries. Paste it in as context and the output quality jumps significantly.

Meeting prep that goes beyond the basics

Pre-call research and meeting prep overlap, but they’re different. Research is about the prospect. Meeting prep is about the conversation strategy.

“I’m meeting with the CTO and VP of Sales at DataFlow Inc tomorrow. Here’s what I know so far: they’re evaluating us against two competitors (Fivetran and Airbyte), their main concern is implementation time because they have a board deadline in Q3, and the CTO is technical and will want architecture details while the VP of Sales cares about reporting capabilities. Prepare talking points for each person, suggest three questions I should ask to uncover their real decision criteria, and give me a one-paragraph response for the likely objection: ‘your competitor is 30% cheaper.’”

This kind of prep used to be something only the best reps did consistently. With an agent, every rep on your team can walk into meetings with this level of preparation. That’s a real competitive advantage because most salespeople don’t prepare like this.

Follow-up emails that don’t sound like templates

Post-meeting follow-ups are one of those things that should be easy but somehow take 30 minutes. You want to reference specific things from the conversation, outline next steps, and sound like a human being. Templates feel generic. Writing from scratch takes too long.

Feed the agent your meeting notes (even rough ones) and let it draft:

“Write a follow-up email after my meeting with Sarah Chen at DataFlow Inc. Meeting notes: she’s most interested in the real-time sync capability, concerned about their team’s bandwidth for implementation, wants a reference customer in fintech, and asked about SOC 2 compliance. Next steps are: I’m sending a case study from our fintech client, scheduling a technical deep-dive with her engineering lead, and providing our SOC 2 report. Keep the email under 200 words. Professional but warm tone.”

The result will sound like you actually paid attention in the meeting, because you did. The agent is just turning your notes into a polished email faster than you could type it.

For the “checking in” emails that feel awkward to write, try: “Write a brief check-in email to a prospect I haven’t heard from in 10 days. Last conversation was about their data integration needs. Add value by mentioning that we just published a case study about reducing data pipeline latency by 60%. Don’t be pushy.” These come out surprisingly natural compared to the “just touching base” emails most of us default to.

If you want to get better at these prompts in general, read tips for better results.

Pipeline reporting without the spreadsheet gymnastics

End-of-week pipeline reviews shouldn’t require an hour of pivot tables. Export your pipeline data and ask:

“Summarize my sales pipeline. Break it down by stage, show the total value and deal count at each stage. Flag any deals with no logged activity in the last 14 days. Identify the three highest-value opportunities closing this month and note any risks I should be aware of (like missing next steps or no champion identified). Format this as a report I could share with my manager.”

This replaces the Friday afternoon scramble to build a pipeline slide. You still need to add your judgment, your gut feeling about which deals are real and which are wishful thinking. But the data assembly happens in seconds instead of an hour.

Where agents fall short in sales

I want to be honest about the limits because I’ve seen teams over-rely on this stuff and make expensive mistakes.

Agents are bad at relationship judgment. They can’t tell you that a prospect’s enthusiasm is performative, or that the real decision-maker is the quiet person in the corner of the Zoom call. They can’t read the political dynamics inside a company or sense when a deal is about to go sideways.

They’re also bad at deciding which leads to prioritize when the data is ambiguous. A lead might look perfect on paper but have no budget. Another might seem small but is actually the entry point to a massive account. That’s your pattern recognition from years of selling, and no agent can replicate it.

Use agents for the work that’s clearly defined: research, drafting, data cleanup, reporting. Keep the judgment, the relationship-building, and the strategy in your own hands. The goal is to reclaim those 29 hours of admin so you can spend more of them on the 11 hours that actually generate revenue.

For more on how agents help with the writing side of sales, see writing and communication with AI. And if your marketing team is using agents too (they should be), check out agents for marketing teams so you can coordinate your messaging.