The Prompt to Identify Competitors You Actually Lose To

Ask any AI assistant who competes with a company and you get a clean, ordered list of vendors. The ordering is the part worth examining, because it is not arbitrary and it is not what you need. It reflects how much has been written about each company, which is a real quantity that correlates with funding, marketing budget and press attention, and correlates only loosely with how often a name comes up in your deals.

Those two quantities come apart hardest at the ends. A well-funded vendor that publishes constantly will be named first even where it sells to a different buyer entirely. A competitor that beats you twice a quarter, sells quietly through partners and has never issued a press release is invisible to the same process. Neither error announces itself, because the list arrives in one tone.

The list you need is not published anywhere

This is the uncomfortable part, and it is also why the prompts below look the way they do. Your competitive set is a fact about your own deals. It lives in your closed-lost records, in the queries buyers use before they ever speak to you, and in what buyers say when somebody asks them. None of that has ever been published, so no amount of recall reaches it and no better-worded question will either.

Each of these answers a different question, and each catches competitors the others miss. Supplying one of the four is common and produces a set skewed in a knowable direction.

The first builds the set. The second exists because closed-lost reasons are almost always free text, and reading a hundred of them is exactly the work worth handing over. The third finds names nobody in sales has met. The fourth is the one that keeps the set alive after the week you built it.

The instruction doing most of the work appears in all four: cite which input put this name here. Without it, an unsourced name and a name backed by nine lost deals arrive looking identical, and the whole point of building the set from evidence is lost in the formatting.

We asked about a real company, then checked the answer against a source anybody can open: that company’s own published comparison pages. The result was not the one the complaint about these tools usually predicts.

The recall was fine. The category was the problem

Every product Pipedrive publishes a comparison page against appeared in the answer, which is a better result than the invention story suggests and worth saying plainly. What went wrong was structural. Ten names came back and all ten were CRM vendors, because a question about competitors resolves to a list of companies, and a spreadsheet is not a company.

That is a silent omission rather than an error. Nothing in the output is false, nothing is marked uncertain, and the thing missing from it is in many categories the single largest reason deals are lost. An answer can be entirely accurate and still describe the wrong contest.

The third tier in the prompt above exists because of exactly that result. It holds what a buyer does instead of buying anything from anyone, and it is the tier a model will not populate unless told to.

Three things usually sit in it. A spreadsheet, which is free, already works and belongs to nobody who has to justify it. An internal build, which is a competitor with a roadmap, a sponsor and a sunk cost. And doing nothing, which is what a closed-lost record means when the reason field reads no decision and the deal is filed as though nobody won it.

Adjacent competitors arrive earlier than direct ones

The other tier people collapse is the adjacent one. An adjacent competitor solves the same problem in a different shape, so it meets the buyer while they are still deciding what kind of thing to buy rather than which one. By the time a direct competitor is in the deal, that question has been settled, often by somebody you never competed with. Keeping the tiers apart is what stops a set from treating both as the same urgency.

The set you just built describes the deals you have already lost, which is the only honest basis for one and also its limitation. It is a rear-view measurement, and the companies that will cost you deals next year are the ones entering your market now.

The events that change a set are public and individually unremarkable. A funding round that turns a quiet competitor into one with a sales team. An acquisition that puts a product you never competed with inside a suite your buyers already own. A vendor from an adjacent category shipping the feature that made you different. Each is a headline for a day, and the cumulative effect is a set that stopped describing your market some months before anybody noticed.

The set is still yours to decide, and one thing about it is not automatable at all: something has to nominate a company before anything can watch it. That first judgement comes from the deals you lose, which is where this page started.

Flares watches funding, launches, pricing and positioning across the companies you name, so a new entrant reaches you as a signal rather than as a surprise in a closed-lost reason. The set is still yours to decide. What changes is how long you go without knowing it needs deciding again.

Which leaves the set exactly where it belongs, as a description of deals you have actually had rather than of a category you have read about. Rebuild it from that evidence each quarter, and let the watching tell you when a quarter was too long to wait.