AI in Sales: Where It Helps, Which Tools Fit and How to Start

AI in sales is the use of language models and machine learning for the work around a sale: researching an account, preparing a meeting, writing up a call, drafting the follow-up and keeping the CRM current. Most of the time it saves today comes from that reading and writing. Scoring leads and projecting revenue count as AI too, but those models learn from past deals, so they pay off for teams with a long, clean deal history and much less for a five-person consultancy.

The selling itself stays with people: earning trust, judging scope and price, deciding when to push. A good start is one repetitive task, done as a draft that a person reads before anything goes out.

Customermates, an AI-native CRM, publishes this guide and appears in its own labelled section. Facts about other vendors come from their own pages, checked on 8 October 2026.

Where AI helps in a sales process

The table follows a deal from first research to the pipeline review. The last column is the part a person should keep.

The first seven rows work with the data a small team already has. Scoring and forecasting do not: a consultancy that closes 40 deals a year gives a model too little to learn from. At that size, a weighted pipeline, where each deal's value is multiplied by the probability you set for its stage, is easier to check and to explain to a partner. A wholesaler with thousands of orders a year is different: it has enough data for models that suggest reorders or add-on sales.

Generative AI, scoring models and plain automation

Four kinds of technology get sold as "AI in sales". They fail in different ways, so it helps to know which one a product uses before you trust it.

Many sales tasks need no model at all. If the rule fits in one sentence, workflow automation is cheaper and behaves the same way every time. Agents are a topic of their own, covered in the guide to AI sales agents.

Which AI tools fit which job

This is not a ranking. Each row names a kind of tool, examples described from the vendors' own pages, and the situation it fits.

A team that writes twenty follow-ups a week gains more from an assistant that sees the CRM and the threads than from a call recorder. A team that spends its day on calls gains more from the recorder. If you mainly want a better writing partner, start with the guide to ChatGPT for sales.

How to start: one workflow, four weeks

1. Pick the task you repeat most

   Write down how often it happens and how long it takes. Example: follow-ups on open proposals, ten a week at about ten minutes each.

2. Put the context in one place

   An AI writes a usable follow-up only if it sees the thread, the deal and what was promised. Paste that into a chat, connect your AI client to the CRM, or use a CRM that holds the conversations.

3. Run it read-only for a week

   Ask for summaries and lists only. Compare them with what you know about each deal and note every invented detail.

4. Move to drafts, not sends

   Let the AI save drafts that you edit and send yourself. Count how many go out with only small changes.

5. Write the rules down

   One page: tone of voice, what the AI may never promise (prices, dates, discounts) and which actions need a person. Walk the team through it and keep the page, because it also documents the AI literacy measures described further down.

6. Expand when the edits get small

   After two to four weeks, compare the time per accepted draft with your baseline. Add the next task only when this one runs with little correction.

Example: meeting briefs and follow-up drafts in Customermates

This section shows one setup in Customermates, the CRM that publishes this guide. Mate, the assistant inside it, runs in Customermates Cloud and uses hosted AI credits.

Take a three-person consultancy on the Pro plan:

• Records: contacts, organizations and deals. The deal stage is a Single choice field with stage probabilities: First call 10%, Scoping 30%, Proposal sent 60%, Won 100%, Lost 0%.
• Inbox: each consultant connects 1 account, usually the work mailbox in Gmail or Outlook. The unified inbox shows those threads next to the contacts and deals, and on Pro Customermates can read that account's calendar without writing to it.
• Knowledge Base: one page with the services, how the firm writes to clients and what Mate may never promise. Mate reads the Knowledge Base in chats and in routine runs, so the rules from step five live in one place instead of in every prompt.

Each consultant then adds a routine that runs On a schedule, Every day at 07:30:

Look at my calendar events for today. For each meeting with someone
outside our company, find the contact, the organization and any open
deal. Write a short brief per meeting: the last three messages in
either direction, the deal stage and value, open tasks, and one
question I should ask. Say what you could not find.
Do not send or change anything.

The briefs appear in the run's Run details, where the consultant can reply to ask a follow-up question. Later in the day, in a chat with Mate:

Find my deals in the stage "Proposal sent" where our last message is
older than seven days. For each, save a two-sentence follow-up as a
draft in the latest thread, in the language of that conversation.
Do not send anything.

Both instructions say "Do not send" for a reason. When asked to send an email or chat message, Mate sends it without a second confirmation; it stops for approval before deletions and a few admin actions. A routine runs as its owner, with the owner's channels and permissions, and refuses anything that would need approval, because nobody is there to give it.

What it costs: Pro for three people is €87 per month, billed monthly. That covers the inbox, 1 connected account and up to 5 routines per person, plus monthly hosted AI credits per active user; see pricing. Starter, at €36 per month, includes Mate and 1 routine per person but no inbox and no calendar, so the brief would only see CRM records. Pro includes the 7-day free trial.

With your own AI client instead: Claude, ChatGPT and other clients can do the same work through the Customermates MCP server, on every plan and on a self-hosted instance. The model then runs on your own AI subscription. Messaging and calendar tools need Pro or higher in the cloud. The server asks clients to confirm before sending or deleting, but whether you see that confirmation depends on the client.

Risks to settle before you start

• Invented details. Generative models write plausible facts. Ask for a source on every claim in a research brief, and never let a first message to a new contact go out unread.
• Sending without review. Look up which actions your tool performs without asking, and say in every instruction whether it may send.
• Customer data in prompts. GDPR applies as soon as personal data reaches an AI provider. Use an account covered by a data processing agreement and send only what the task needs.
• AI literacy. The EU AI Act requires every company that uses AI at work to take measures for its staff's AI literacy. The FAQ below has the details.
• Recording calls. A call recorder needs the other side's consent. In Germany, recording the non-public spoken word without authorisation is a criminal offence under § 201 StGB.

Frequently asked questions

Vendor and legal details checked on 8 October 2026 against Salesforce AI for sales, HubSpot AI, Gong conversation intelligence, LinkedIn Sales Navigator AI, Claude connectors, the European Commission's AI literacy FAQ and § 201 StGB.