Competitive Displacement Prompts for Incumbent Accounts

Write down what actually decides whether an account can leave its current vendor. When the term ends. How they cancel, and by when. What they have built on top of the product, and who built it. Who signed the contract, and whether that person still works there. What a move would cost the people who would have to run it.

Four of those five are known only inside the account. The fifth sits in a legal document neither side of the deal has read.

A model has none of them. It will produce all of them anyway, because each one has an obvious ending. Annual term. Auto-renewal. Thirty days’ notice. Four to six weeks to migrate.

A complete-looking plan is the warning

The output has no gaps in it, and that is the problem. A plan with holes gets questions asked of it. A plan with a paragraph under every heading gets forwarded. By the time it reaches a pipeline review, the thirty-day notice window has been repeated twice and nobody remembers it came from nowhere.

The second one is the surprise. A vendor publishes more about how to leave them than most sellers ever look at. Contract terms and the data processing agreement are public, dated, and specific in exactly the places a plan tends to guess.

Mark every line in the first block with whether the buyer was reading or remembering. Renewal dates given from memory are wrong often enough to deserve their own column, and a plan built backwards from a date that is out by a quarter is a plan for the wrong quarter.

Four prompts, in a strict order. The plan comes first. The marking comes second, in its own message. The question list gets built from the marking. The last one exists to stop the plan promising to remove a switching cost nobody has ever removed.

That last prompt asks for a phrase rather than a number. Where your own record holds nothing about a cost, the output should say so instead of estimating. A short list of switching costs you can speak to from experience beats a complete one where half the entries came from an average, because the buyer paying the cost will ask who else has paid it.

The account context was a single sentence invented for the experiment, and the run condition says so, because nobody publishes somebody else’s contract. The vendor is real and its terms are public, so every contract claim in the output could be checked against the document it was describing.

An incumbent’s notice period is in their own terms

Thirty days’ notice is the most repeated fact in software buying. For this vendor it is not a fact at all. Their published terms set no advance notice period in the general agreement. Cancelling means switching off auto-renewal in the account before the current term ends. The renewal length is not fixed either: it runs for the shorter of the previous term or one year, so a two-year deal renews for one.

Both change the conversation rather than decorating it. A rep who says the notice window has closed when it has not talks a buyer out of a deal that was there. A rep who assumes a two-year renewal when it will be one has written off eighteen months that were never at risk.

An audit clause is not an audit

The prompt that produced this plan already asked for what had been guessed. It complied, with one sentence at the bottom saying the details above were typical for the company type. That is a disclaimer. It names no line, changes no line, and sits under thirty lines that read like findings.

Asked as its own message afterwards, the identical instruction produced a clean split and a ratio: one supplied fact, thirteen guessed. Worth carrying to every other prompt you write: there is a difference between an instruction that changes the shape of the output and one that asks for commentary after it. Tagging each claim as it is written works, which is why a battlecard prompt marks unsourced lines inline. Adding “say what you inferred” to the end of a request to generate something does not.

The guessed list looks like a weakness in the output. It is the most useful thing the exercise produces. Thirteen assumptions is thirteen things that decide the deal, already written down, already sorted by how much the plan rests on each. One more prompt turns a document nobody can act on into an agenda.

Which questions a buyer will actually answer

Order them by how fast the account can answer, not by how badly you want to know. The export question and the who-built-it question get answered in a sentence by somebody already on the call. The notice mechanism needs a person who is not in the room and may need a document, so asking it early reads as a request to help plan their exit before they have decided they want one.

The one to handle carefully is the signature question. It is the highest-value line on the list and it can land as a challenge to your champion’s authority, which is why the prompt above asks the model to flag intrusive questions and offer a softer version. Knowing who signed changes who the plan is written for, and the brief for the next call is where that has to show up, because a rep reads the brief and files the plan.

A displacement plan has an expiry date that most competitive work does not: the renewal it was built around. Everything in it depends on a window. Once the window passes the plan is not stale so much as void, and the account goes back to being unavailable for another term.

Two things move in the meantime and both are public. The vendor changes packaging or price, which is the most common reason an account that was not listening starts listening. And the vendor changes its own terms, which is the document every timing claim in the plan rests on. The terms behind the run above carry a last-updated date, and a plan built against last year’s version is describing a contract nobody signed.

What stays out of reach is the half that matters most. No amount of watching a vendor tells you when their customer is unhappy. That arrives through a conversation, a hiring change or a review. Displacement is the one competitive job where the trigger is almost always on the buyer’s side rather than the competitor’s.

Flares watches the competitor half and dates it: pricing, packaging and the product changes that give an outreach a reason to exist this week rather than in general. Getting the contract facts out of the account stays a conversation.

Which is why the question list is the thing to keep, not the plan. A plan ages into something confidently wrong about a contract nobody described to it. A list of thirteen open questions ages into a list of thirteen open questions, and each one gets crossed off by somebody asking.