Your buyers brought AI to work
Structure and markup refreshed for current answer engines; the original analysis is preserved.
This month OpenAI gave companies a private, sanctioned way to use ChatGPT at work. The feature list is not the point. The point for a B2B vendor is quieter: the AI your buyers were already sneaking onto their phones is becoming an approved part of how they do their jobs — including how they get oriented in a category and decide which providers are worth a closer look. Including how they research you.
the short answer
What was blocking AI at work was never interest — most big companies’ teams were already using ChatGPT — it was data risk. A private enterprise version removes that, turning hidden use into sanctioned use. For a B2B vendor that means buyers will use AI earlier in the process, to get oriented in a category and seed a shortlist, often before reading a single vendor page. So you now have to be legible to a new reader — a model that surfaces clearly described, corroborated, unambiguous providers — because it is becoming the gatekeeper to the buyer’s attention. It is early and uneven: prepare, don’t panic.
key takeaways
- This month OpenAI launched a private, sanctioned ChatGPT for companies — the thing that was blocking AI at work was never interest, it was data risk, and an enterprise version removes it.
- The consequence for a B2B vendor is not the feature list. It is that the AI your buyers were already using on the sly is becoming a tool their employer approves — so they use it more, and earlier.
- Earlier is the operative word: buyers will use an approved AI to get oriented in a category, and that orientation step is where a shortlist quietly begins — including the shortlist you are or are not on.
- That adds a reader you did not have to satisfy before: a model doing the first pass, surfacing providers that are clearly described, corroborated by third parties, and unambiguous as entities.
- It is early and uneven, not universal — which is the moment to prepare, not panic. The buyer still decides; the AI is becoming the gatekeeper to the buyer’s attention, and you have to be known to it.
where the AI now sits
The buyer still evaluates and still decides. What changed is the stage before: an approved AI now often runs the first pass — naming providers, framing the category — and a brand it does not surface at "research and shortlist" rarely makes it to "evaluate". The new work is getting named in that first pass.
Why a privacy feature is a distribution event
It is easy to file an enterprise launch under procurement news and move on, but for anyone who depends on being discovered, this is a distribution event in disguise. The mechanism is simple: behaviour that is against the rules stays small, hidden, and late; behaviour that is sanctioned grows, surfaces, and moves earlier. By making AI use safe and approved, the launch does not create your buyers’ appetite for asking a machine to explain their options — that already existed — it removes the friction and the fear that kept the appetite in check. The predictable result is more of that asking, done openly, by more senior people, about more important decisions. And the most important decisions a vendor cares about — which category to trust, which providers to consider — are exactly the kind a buyer will now feel free to explore with an approved assistant.
Worth stating plainly: this is a beginning, not an arrival. Plenty of companies remain cautious or are still restricting these tools, and a buyer leaning on AI to build a shortlist is, this month, doing something early rather than something standard. We are not claiming the buying committee has been handed to a chatbot. We are claiming that a barrier just dropped, that what is behind it is a real change in where and how buyers get oriented, and that the moment a barrier drops is the moment to start preparing — well before the change finishes arriving and the easy positions are taken.
The shift, in three parts
The barrier was permission, not interest; sanctioned use moves the AI earlier in the buying process; so you now have to satisfy a reader before the human ever sees you. Open each part for where it changes the plan.
01 The barrier was never interest — it was permission
Read the announcement for what it removes, not what it adds. By OpenAI’s own account, teams at the large majority of big companies were already using ChatGPT within nine months of its launch — the appetite was never in question. What held it back was that the use was unsanctioned: sending company information to a consumer tool that might train on it was a real data and compliance risk, and plenty of firms responded by restricting or quietly banning it. An enterprise version with proper security and an explicit promise not to train on company data is aimed squarely at that barrier. It converts a furtive, against-the-rules habit into an approved, on-the-record tool. That conversion is the story for a vendor: not that AI arrived at work — it was already there, in pockets and on personal accounts, doing small jobs in the margins of the workday — but that it just stopped being something employees had to hide, which means it is about to be used far more, and far more openly, than it was last month.
02 Sanctioned use moves the AI earlier in the buying process
When a tool is approved rather than smuggled, people reach for it sooner and for bigger things. The same buyer who used a hidden chatbot to polish an email will use an approved one to do something more consequential: get their bearings in a category they do not know well. Faced with an unfamiliar problem, the natural first move becomes asking the AI to explain the landscape — what the main approaches are, who the notable providers seem to be, what the trade-offs and pitfalls are. That is not the moment a deal is won, but it is the moment a shortlist is seeded, and it now often happens before a human has read a single vendor page. The buying process did not get a new step so much as a new opening act, performed by a machine, in which some providers get named and framed favourably and others never come up at all. Where your category’s buyers start with that opening act, the act decides who is even in the running for the human rounds that follow. None of the later stages get a chance to favour you if the opening act has already left you out.
03 So you now have to satisfy a reader before the human
The practical consequence is a new reader standing between you and the buyer, and it does not read like a person. It will not be charmed by your hero copy or your case-study design; it surfaces providers it has come to understand as real, credible entities in the category, drawing on what it has learned and what it can corroborate. So the work extends beyond persuading the human who lands on your page to being legible and trustworthy to the system that decides whether a human lands there at all. In concrete terms that means being clearly described in plain language, corroborated by independent third parties rather than only by your own marketing, and unambiguous as an entity so the model is confident it knows who you are. The buyer is still the one who chooses; the AI is becoming the gatekeeper to the buyer’s attention. This is the same shift the AC Group has helped clients navigate for 27 years each time a new layer inserted itself between a company and its customers — learn to satisfy the new reader without losing the human behind it.
Two vendors, one curious buyer
A buyer at a mid-sized company is handed a new problem in a category they do not know. A year ago they would have opened a search engine and started clicking; this month, with an approved AI on their desktop, they open that instead and ask it to explain the space and name the credible options. Two vendors solve this buyer’s problem about equally well. The first has spent its effort on a polished site and paid search, confident that a buyer who looks will be impressed. The second has spent comparable effort being clearly described in plain terms, earning write-ups and mentions from independent sources, and keeping its entity unambiguous everywhere it appears.
When the buyer asks, the model names the second vendor among the options to consider and does not mention the first — not out of malice, but because it understands the second as a real, corroborated entity in the category and has no clear picture of the first. The buyer takes that shortlist into the human rounds, and the polished site never gets the visit it was built for, because the buyer never learned it existed. Both vendors were equally good and equally available; only one was legible to the reader that now goes first. That gap — invisible when buyers started with a search box, decisive once they start with an answer — is the whole reason this month’s quiet enterprise launch belongs on a vendor’s radar.
What to do with this
Do not overreact, and do not wait. Overreacting looks like abandoning everything that works with human buyers to chase a machine; waiting looks like assuming this is a curiosity until your pipeline quietly tells you otherwise. The measured move is to add the new reader to the list of audiences you serve, alongside the searcher and the buying committee, and to ask a simple question of your category: if a buyer asked an AI to explain it and name the options today, would your name come up, and would it come up described the way you would want?
Then close the gap where the honest answer is "no". Make sure you are described in plain, unambiguous language a model can understand, not just in brand copy written to impress. Earn corroboration from credible third parties, because a model trusts a category’s consensus more than a vendor’s self-description. Keep your entity clean and consistent so the model is sure who you are. None of this means deserting the human buyer; it means making sure the human still gets the chance to choose you, by being known to the reader that now decides whether they ever see you. That is the discipline the AC Group has applied through every shift in how customers find companies for ' + years + ' years — meet the new gatekeeper on its terms, without forgetting who it is a gatekeeper to.
What this is not
It is worth drawing the boundaries clearly, because the easy version of this argument overshoots and loses credibility. This is not a claim that buying committees have been replaced by chatbots, or that human judgement, references, demos, and procurement are going away — they are not, and the consequential decisions still run through people. It is not a claim that every buyer in every category is doing this now; in plenty of industries, and at plenty of companies that remain cautious or are still restricting these tools, the old path is fully intact. And it is not a prediction with a date attached, because the pace will vary enormously by sector, regulation, and how technical the buyer is. Overstating any of that would be its own mistake.
What it is, is narrower and sturdier: a barrier that kept AI use small and hidden has been lowered, the behaviour behind it — buyers asking a machine to make sense of their options — is real and was already latent, and lowering the barrier reliably makes such behaviour grow and surface. You can disagree about how fast and how far without disagreeing about the direction. The case for acting now does not rest on the change being complete; it rests on the change being directional and on early positions being cheaper to take than late ones. We would rather be accurate and modest about this than loud and wrong, because a vendor who is told the sky is falling will either panic or, worse, stop listening.
How to tell if it is already happening to you
You do not have to guess whether your buyers have started here; the signs show up in your own pipeline if you look. Watch for buyers who arrive already oriented — using the category’s vocabulary correctly, referencing the main approaches, sometimes naming competitors you were not expecting — before they have spoken to anyone on your team or, as far as you can tell, visited your site. Listen for the offhand "I asked ChatGPT and it suggested…" in discovery calls, which is becoming as ordinary as "I googled it" once was. Notice when deals appear later in their own process than they used to, as if a round of narrowing happened somewhere you could not see.
Then run the direct test, which costs nothing: ask an AI the questions your buyers would ask to get oriented in your category — what the main options are, who the credible providers seem to be — and read the answer as a buyer would. If you are named, well described, and in good company, you are in reasonable shape for now. If you are absent, described wrongly, or framed as a minor player, that is the gap to close, and it is far cheaper to close while the behaviour is still emerging than after it has become the default. Do this for the handful of questions that matter most, repeat it on a calm cadence, and you will know where you stand long before your win rate forces you to find out.
AI in the buying process: quick answers
Why does ChatGPT Enterprise matter for a B2B vendor?
Not because of what it does inside a company, but because of what it opens up. The barrier to using AI at work was never enthusiasm — by OpenAI’s own figure, teams at most large companies were already using ChatGPT. The barrier was that it was unsanctioned: data and privacy risk meant many firms quietly banned it or left it in a grey zone. An enterprise version with real security and a promise not to train on company data removes that barrier and turns informal, hidden use into approved, everyday use. For a vendor, the relevant shift is not the feature list; it is that the AI your buyers were already consulting on the sly is becoming a tool their employer endorses. When research happens on an approved tool rather than a furtive one, it happens more, more openly, and earlier in the process — including the research where they decide which vendors are worth a closer look.
Are buyers really using AI to choose vendors yet?
Some are, many are not, and it is early — which is exactly the right moment to notice rather than the moment to panic. As of this month the honest picture is mixed: plenty of companies are still cautious or restricting AI, and a buyer who asks a chatbot "who are the leading options for X" is doing something new, not something universal. But the direction is not subtle. The same people who now draft emails and summarize documents with an approved AI will, sooner rather than later, use it to get oriented in an unfamiliar category — to ask what the main approaches are, who the notable providers seem to be, what to watch out for. That orientation step is where a shortlist quietly begins. You do not need every buyer to be doing this for it to matter; you need enough of your buyers to be doing it that being absent from those answers starts to cost you real consideration.
What does this change about how I should be found?
It adds a reader you did not used to have to satisfy. For years the job was to be found by a buyer searching and by the humans on a buying committee reading. Now there is an intermediary that may do the first pass: a model the buyer asks to explain the category and name the options. That model does not browse your site the way a person does; it draws on what it has learned and what it can retrieve, and it surfaces the providers that are clearly described, corroborated by third parties, and unambiguous as entities. So the work shifts from persuading a human on your page to being legible and credible to the system that decides whether your page is one a human ever sees. The buyer is still the decision-maker; the AI is increasingly the gatekeeper to the buyer’s attention, and you have to be known to the gatekeeper to reach the decision-maker.
Isn’t this just SEO with a new name?
It overlaps with good SEO but it is not the same job, because the reader is different. Search rewards a page that matches a query and earns clicks; an AI asked to name the options in a category rewards a brand it has come to understand as a real, corroborated entity in that space. You can rank well and still be invisible in the answer if the model has no clear sense of who you are, and you can be named in the answer even for a query where your page does not rank, if your reputation is well established off your own site. The overlap is real — clear content and a clean technical foundation help both — but the new task is reputational and entity-level as much as it is on-page. Treating it as "SEO with a new name" will leave you optimizing for the human searcher while missing the machine that now answers before the human searches.
A note on sources and timing
This is written in late August 2023, just after OpenAI launched its enterprise tier of ChatGPT — a private, secured version that, by the company’s account, does not train on customer data, aimed at companies where teams were already using the consumer tool informally. We have described only what was public as of this writing, and we have been careful not to overstate adoption: this is an early signal, not a finished shift, and many buyers are not yet researching vendors this way. The durable point does not depend on the pace: when AI use at work moves from forbidden to approved, buyers will reach for it earlier and more openly, including to get oriented in categories and seed shortlists — which adds a reader you have to be known to before the human ever looks. That is the kind of shift the AC Group has helped clients meet for 27 years, each time a new layer slid in between a business and the people it hopes to reach.