AI in sales: how reps can use it well

How is AI used in sales day to day?

Most reps who use AI well use it for preparation and writing around the conversation, not for the conversation itself. The tool helps them arrive informed and follow up clearly; the rep still does the listening, the judging and the committing.

This guide is about individual sellers and sales managers using AI assistants in their own work. Buying a sales engagement platform or an AI feature inside a CRM is a separate decision with its own vendor review.

How should a rep research an account with AI?

Give the tool the sources and ask it to summarise them, rather than asking it what it knows about the company. A general assistant without web access may describe a company from outdated training data, and one with web access may still mix up companies with similar names. Either way, the summary is a lead to check, not a fact to repeat.

Suppose a rep at a fictional software firm is preparing to call Northgate Logistics, a fictional regional haulier. Asked 'tell me about Northgate Logistics', the assistant confidently said the company had recently raised investment and was expanding into Ireland. Neither statement appeared on Northgate's own website or news page. Had the rep opened with 'congratulations on the funding round', the call would have started with a mistake about the customer's own business.

The better prompt pasted the text of Northgate's About and News pages and asked: 'Summarise this into five points relevant to a fleet operations buyer. For each point, quote the sentence it comes from. List anything a rep should ask rather than assume.' The output was shorter and duller, and every line could be checked in seconds.

What does a reviewed AI follow-up email look like?

A good follow-up confirms what the buyer said, what you agreed and what happens next. AI drafts tend to drift towards what sales emails usually say, which is not the same thing.

Continuing the fictional example, the rep's call notes said: the buyer manages 40 vehicles, is frustrated by manual route planning, wants to see a demo with their own depot locations, and needs their finance lead in the next meeting. The first AI draft thanked the buyer, then added 'our platform will cut your fuel costs significantly' and 'companies like yours have seen fast returns'. The call notes contained no cost discussion and the rep had no case study for a haulier of that size.

The rep cut both lines and asked the tool to 'use only the points in my notes and end with the agreed next step'. The final email was four sentences: a recap of the routing problem in the buyer's words, confirmation of the demo with their depot data, a request for the finance lead's availability, and a proposed date. It made no promise the rep could not keep, which is what makes a follow-up trustworthy.

How can AI help with objections and call practice?

AI is a patient practice partner for objection handling. Ask it to play a specific buyer, give it the context and tell it to push back realistically rather than agree with you. For example: 'You are the finance director of a 60-person firm. You think our tool is too expensive and you have a cheaper spreadsheet process that mostly works. Challenge me for five turns. Do not accept vague answers.'

After the role-play, ask it which of your answers were vague or unsupported. The point is not to memorise a script. It is to find out where you reach for a claim you cannot back up, and to prepare a truthful answer instead. If your answer depends on a figure, such as implementation time, confirm it with your own team before using it on a call.

What sales information should stay out of AI tools?

Assume customer personal data, contract terms, negotiated pricing and anything under a non-disclosure agreement stay out unless your company has approved a tool and plan for that use. Call recordings and transcripts often contain other people's personal information and details the customer shared in confidence.

Business plans of the main assistants often have different terms from personal accounts on retention and on whether content is used to improve models, and AI features inside a CRM follow whatever your administrator has configured. Check the specific plan's terms rather than assuming. When in doubt, summarise from your own notes rather than pasting the raw transcript.

What rules apply to AI-written sales emails?

The same rules apply whether a person or an AI tool wrote the email. In the United States, the Federal Trade Commission's CAN-SPAM compliance guide says the law makes no exception for business-to-business email, requires honest subject lines and headers, a valid postal address and a working opt-out honoured within 10 business days, and that hiring another company to send your email does not transfer your legal responsibility. Other countries have their own direct marketing rules, which can be stricter.

AI makes it easy to produce a large number of personalised-looking emails quickly. That increases the importance of accurate targeting and honest content rather than reducing it. An email that invents a connection ('I saw your talk last week') or a deadline is a bad sales practice before it is a legal one. This is general information, not legal advice.

Which AI tools for sales are worth starting with?

Start with the general assistant your company already approves, such as ChatGPT, Claude, Gemini or Copilot, for research summaries, drafts and practice. If your CRM or sales engagement platform includes AI features, those are convenient for notes and summaries because the data is already there, but confirm with your administrator what the feature can read and who can see its output.

Dedicated AI sales tools, such as automated prospecting or call analysis software, are a buying decision for the sales leader and whoever reviews vendors. Judge them on a real task with your own data rules, not on a demo.

Where should AI not lead in sales?

AI should not make commitments on your behalf. Prices, discounts, contract terms, delivery dates and forecast numbers belong to people who are accountable for them. A drafted proposal that quotes a timeline the delivery team never agreed is a problem that surfaces after the deal closes.

It is also the wrong tool for the moments that build trust: a difficult renewal call, an apology for a service failure or a negotiation. Prepare for those with AI if it helps, then do them yourself. And if a buyer asks a question you cannot answer, the honest 'let me check' beats a fluent generated guess.

How can a sales team build these habits?

Use one shared exercise: the same fictional call notes, a follow-up email to write with AI, and a short explanation of what each rep removed. Comparing those explanations shows who checks claims and who sends what the tool wrote.

Tosheo.ai's planned sales path is being built around that kind of task: a diagnostic, short lessons with sales examples, role-play scenarios scored against a visible rubric, and a manager view of skill progress that does not expose chat logs. It is pre-build, so the details are intended scope, not a live service.

Sources and scope

US Federal Trade Commission, CAN-SPAM Act: a compliance guide for business ↗