← The Clair Blog

    ChatGPT for sales professionals: what to trust it with, and what to never

    Reply rates are falling because every inbox can now smell the sequence. Where AI actually helps a seller, and the three mistakes that cost revenue directly.

    A note before the split. This is general guidance, your company's policies on tools and data govern, and the tools themselves change constantly. The economics described below, which is what happens when everyone automates the same touch, will outlast every feature.

    Sales got to the AI future first, and the view from there is instructive. Outreach that once took an afternoon now takes a click, while sequences write themselves and reply rates fall in parallel. The same technology that lets you send a hundred personalised-looking emails lets everyone else do it too, and buyers adapted in months.

    The modern inbox can smell a sequence from the subject line. “I noticed you recently” has become a delete trigger, and the fake-personal touch now performs worse than no touch at all. Since the tools are not going back in the box, the useful question is where AI genuinely helps a seller now that everyone has it.

    The short answer. The durable value sits before and after the touch, not in it. Trust ChatGPT with account research before the call, call preparation, follow-ups drafted from your own notes, proposal structure, and the pipeline admin nobody became a seller to do. The touch itself, when a buyer decides whether you are a person worth replying to, is exactly where the machine voice costs money. Two hard lines sit beneath all of it: nothing it claims about your product or pricing goes out unchecked, and no pipeline or customer data enters a consumer version.

    What to hand over

    The job Why it wins
    Account research synthesis The annual report, recent news and public signals compressed into a one-page brief before the call, turning an hour of gathering into ten minutes plus your read
    Call preparation Likely objections, questions worth asking and the two things this buyer's world cares about right now, rehearsed against something that talks back
    Follow-ups from your notes Your rough call notes turned into the recap, actions and next-step email in the buyer's language, taking your edit rather than your evening
    Proposal structure The skeleton, sections and executive-summary first draft, with every substantive claim and number supplied and checked by you
    Pipeline admin The tidying of notes and next steps that keeps the system honest without eating selling time

    This is the useful division of labour in one table. The tool compresses material, gives the blank page a structure and clears the clerical residue. You still decide what matters, what is true and what the buyer should hear from an actual person.

    The three mistakes that cost revenue directly

    Fake personalisation at scale. The temptation is obvious: AI can reference something about each of two hundred prospects. The result is obvious to buyers too, because machine-personalised outreach is now a recognisable genre and being caught faking attention is worse than not paying it.

    Real personalisation means one true, specific observation that only someone who looked would make. It does not scale, which is now precisely its value. Use AI for the research that finds the observation, then write the observation yourself. The same principle runs through the edit that stops an AI email reading like one.

    Invented product claims. Ask the tool to draft a proposal or answer an objection and it will fluently generate capabilities, comparisons and figures about your own product, some of which will be wrong. A false claim in a proposal is a legal and commercial problem wearing your signature.

    Everything it asserts about what you sell, what it costs and what it beats gets checked against your actual materials before a buyer sees it. There are no exceptions for claims that happen to sound right, which are, rather inconveniently, the dangerous ones.

    The data paste. Pipeline exports, customer lists and deal details are commercially sensitive and often contractually protected, while consumer AI versions may retain or use what goes into them. That material lives in your CRM and whatever sanctioned tools your company runs. It does not travel to a consumer chatbot.

    The situation-shaped version still gets you the thinking help: a hypothetical buyer of a certain size, stalled at a certain stage, with the identifying details stripped out. The hypothetical method is the practical way to do that without handing over the account.

    The line that holds

    The sellers pulling ahead with AI have quietly reorganised around one distinction: the machine does the preparation, while the human does the moment. Research, rehearsal, recaps and structure are preparation, and handing them over returns hours to actual selling. The touch, the observation, the call and the message that makes a buyer feel seen are the moment.

    In a market flooded with machine attention, genuine attention became the premium product. It is now the scarcest thing in the profession precisely because everyone else automated it, and the tooling that freed your afternoon is what makes supplying it possible. For owners carrying sales alongside every other job, the business-owner version of this split draws the wider boundary.

    Not sure where to start?

    Preparation from the tool. The moment from you.

    The free finder tells you which AI to open for the task in front of you, in about 60 seconds, and the full “AI, sorted.” reference lands in your inbox.

    Try the free finder →

    One question in, one answer out.


    Clair helps non-technical professionals know when to trust their AI, when to check it, and when to skip it.

    Common questions

    How should salespeople use ChatGPT?

    For the work around the touch rather than the touch itself: synthesising account research into pre-call briefs, rehearsing objections, turning call notes into follow-ups and recaps, structuring proposals, and handling pipeline admin. The buyer-facing moment stays human, since machine-voiced outreach now underperforms.

    Does AI-written sales outreach still work?

    Increasingly poorly, because buyers have learned the genre. Mass personalised-looking sequences read as exactly what they are, and fake attention performs worse than none. What works is AI-powered research feeding one true, specific, human-written observation, which does not scale and is valuable for that reason.

    Can ChatGPT write my sales proposals?

    It can produce the structure, skeleton and summary first draft, which saves real hours. It should not own the substance, because it will fluently invent capabilities, comparisons and figures about your product, and a wrong claim in a proposal is a commercial and legal problem. Check every assertion against your actual materials.

    Is it safe to put customer or pipeline data into ChatGPT?

    Not on consumer versions, which may retain or use inputs, while deal data is commercially sensitive and often contractually protected. Keep it in your CRM and sanctioned tools, and bring AI the anonymous shape of the situation instead, such as a buyer of a certain size stalled at a certain stage.

    Will AI replace salespeople?

    It is replacing the mechanical layer, including sequences, admin and boilerplate, while increasing the value of judgement about accounts, genuine attention and the human moment where trust forms. Sellers who hand the machine the preparation and keep the moment are gaining ground on those who automated the moment itself.

    The AI Starter Kit

    Set up ChatGPT and Claude properly. Once.

    The workflows, rules and habits that hand the boring recurring work to your AI and give you your evenings back.

    Get the Kit →