Generative AI for sales: real use cases for small teams

Generative AI has gone from novelty to everyday tool in a couple of years, and sales is one of the places it fits most naturally. So much of selling is reading, writing and remembering, and those are exactly the jobs these models are good at.

This post explains what generative AI is in plain terms, walks through the sales tasks where it genuinely helps a small team, shows why connecting it to your CRM matters, and covers what to watch for before you let it near your customers.

What generative AI is

Generative AI is a type of artificial intelligence that produces new content (text, speech, images or code) in response to a prompt. The tools most salespeople use, chat assistants and AI features inside business software, run on large language models trained on huge amounts of text. You give the model an instruction and some context, and it predicts a useful response.

That’s different from older “AI” in sales tools, which mostly meant scoring or sorting: ranking leads, predicting which deals might close. Generative AI writes. It can turn a messy call transcript into a clean summary, a one-line request into a polite follow-up email or a question like “who haven’t I called back?” into an answer.

Usage is spreading fast, though surveys measure it differently. Microsoft’s 2024 Work Trend Index, which surveyed 31,000 knowledge workers in 31 markets in February and March 2024, found 75% of knowledge workers were using generative AI at work (Microsoft). Pew Research Center’s broader October 2024 survey of 5,273 employed U.S. adults found only 16% saying at least some of their work was done with AI (Pew Research Center). The gap comes from who was asked (desk-based knowledge workers versus all workers) and how the question was worded. Either way, many of your competitors are already experimenting.

Use cases for small sales teams

Drafting emails and texts

This is the use case almost everyone starts with, for good reason. Give the AI the context (“lead requested a fence quote, we visited Tuesday, quote sent Thursday, no reply”) and ask for a short, friendly follow-up. You get a draft in seconds and spend your time editing rather than staring at a blank screen.

Pew’s survey points the same way: among workers who had used AI chatbots for work, 47% had used them to draft written content and 52% to edit it (Pew Research Center). For templates that still sound human, see our sales follow-up templates.

Summarizing calls and records

A 25-minute discovery call becomes five bullet points: the need, the budget, the timeline, the objections and the agreed next step. A lead with weeks of messages becomes a short brief you can read in the truck before the appointment. Summaries also make handoffs easier when a deal moves between reps. Our post on AI call summaries covers what a good summary should include.

Research and meeting prep

Before a call with an agency prospect or a commercial property manager, AI can pull together what you already know from the CRM and suggest questions to ask. Research was the most common work use among chatbot users in Pew’s survey, at 57% (Pew Research Center). In Salesforce’s 2026 State of Sales survey of 4,050 sales professionals, sellers said they expected AI agents to cut prospect research time by 34% and email drafting time by 36% once fully implemented (Salesforce). Those are expectations reported in vendor research, not measured results, but they show where sellers see the time going.

First-draft quotes and proposals

If your quotes are built from a price book, AI can turn a conversation (“two zones of sod, irrigation repair, haul-away”) into draft line items for you to check. You still own the numbers. The time saved is in the assembly, not the judgment.

Answering questions about your pipeline

“Which deals are stalled?” “How much is outstanding on invoices?” “Which lead sources converted best this quarter?” When the AI can query your CRM, you get answers without building reports. For owners who don’t live in dashboards, this is often the feature that gets used daily.

Admin and data entry

Logging a call, creating a follow-up task, updating a deal stage, adding a note: AI can propose these from a conversation so the CRM stays current without someone typing it all in. This is where a confirmation step matters most, since these actions change your data.

Connecting AI to your CRM data

A general-purpose chatbot only knows what you paste into it. That creates two problems: it’s slow (you copy and paste context every time) and it’s risky (customer details end up in a tool your business doesn’t control).

AI that runs inside your CRM starts from your real data: the lead’s source, every message, the call notes, the quote, the deal stage. That makes its drafts more specific and its answers checkable, because it can link back to the records it used.

Three things to look for when AI is connected to CRM data:

• Permissions carry over. The AI should only see what the person asking can see. A rep shouldn’t be able to ask the AI about deals they’re not allowed to open.
• Answers link to records. If the AI says a deal has stalled, you should be one click from that deal.
• Changes need confirmation. Reading is low risk; writing isn’t. The AI should propose changes and wait for a person to approve them.

In Exa, for example, the AI assistant inherits each user’s role and permissions, and it only creates tasks, notes, messages or appointments after the user confirms an action card. More on the reasoning in why your AI assistant should ask before it acts.

What to watch for

Confident mistakes. NIST’s Generative AI Profile, published in July 2024, lists 12 risks specific to generative AI, including “confabulation,” the tendency to produce confident but false output often called hallucinating (NIST). In sales, that can mean a wrong price, a promised date nobody agreed to or details from the wrong customer. Review anything that reaches a customer.

Shadow AI and privacy. The same Microsoft study found 78% of AI users were bringing their own AI tools to work. If your team pastes customer data into personal chatbot accounts, you may be breaking promises in your own privacy policy. Our AI and customer data privacy checklist covers the questions to ask.

Compliance doesn’t change. A message drafted by AI is still your message. Texting consent, opt-outs and quiet hours apply exactly as they would if you’d typed it yourself.

Everyone sounds the same. If every rep sends the AI’s first draft, your messages start to sound like everyone else’s. Edit for your voice and add a specific detail only you would know.

Speed over quality. Among Pew’s chatbot users, 40% said the tools were extremely or very helpful for doing things more quickly, but only 29% said the same about improving the quality of their work. Use AI to go faster; keep quality checks human.

How to get started

You don’t need an AI strategy document. You need a two-week experiment:

1. Pick one task. Follow-up drafts or call summaries are good first choices because the output is easy to check.
2. Write down the baseline. How long does the task take now, and how often does it get skipped?
3. Set a review rule. For example: nothing AI-drafted goes to a customer without a person reading it.
4. Use it daily for two weeks. Keep a note of drafts you had to rewrite heavily and why.
5. Decide. Keep it, adjust your prompts or drop it. Then pick the next task.

If you need ideas for what to type, our list of AI prompts for sales is a good starting point. And if your CRM data is messy, spend an afternoon merging duplicates and closing stale deals first; AI answers are only as good as the records behind them.

Final thoughts

Generative AI won’t close deals for you, but it can give back the hours lost to typing, searching and summarizing. Pick one or two daily tasks, such as follow-up drafts and call summaries, use AI that’s connected to your CRM with permissions and confirmations, and review what goes out. If you want a feel for how it works inside a CRM, see how Exa’s AI assistant handles questions, drafts and actions.