Objection Handling Prompts for Competitive Deals

“Handle this objection” already contains its answer. Ask a model to handle one and it hands you a comeback. Ask for three and you get three. It does that on every objection, including the ones where the buyer is completely right.

So add a step before the drafting. An objection is a factual claim about two products. Your material either backs it up or it does not. Getting that verdict takes one pass and produces nothing a rep can say out loud, which is why it gets skipped.

It is also the step that decides everything after it. Three of the four possible verdicts send you somewhere other than an argument.

A polite “you’re right” is not a verdict

Models concede easily, and it costs them nothing. What they write is an opening move, not a judgement. “You are right that X, and” is a shape. Watch the clause after the comma: a real concession can stop there, and a polite one never does.

That is not a style point. A concession written as a shape gets filled with whatever the model can reach, and what it can reach about a competitor is general. The measured run further down is exactly that: three accurate but minor limits, in the right tone, missing the one sentence that settles the argument.

None of this is new. A good objection matrix has carried a true-or-false column for as long as anyone has kept one. What is new is that a model never adds that column by itself. A person filling in a matrix hesitates over a claim they are unsure about. A model asked for responses writes responses. You have to build the hesitation into the prompt.

Most competitive objections are two claims in one sentence. Take “they have a free tier and you do not”. That is a claim about their packaging and a claim about yours, and different documents settle each one.

Give the model only one side and it does something predictable. It sources the side it can see, then asserts the other in the same confident voice. Your rep ends up defending a sentence that is only half backed up.

Paste the objection first, on its own line. Lead with the competitor material instead and the model starts hunting for ammunition before it has read what it is answering. The responses then get built around the best thing in the pile rather than around what the buyer actually said.

Four prompts. The verdict prompt is second on the page and first in practice, and it deliberately produces nothing you can send. The other three each assume you already have a verdict: the full draft for objections worth answering, the split for partly-true ones, and the last for when the buyer is simply right.

The concession prompt bans any clause starting with “but”. That one rule does most of the work. Without it you get a reframe with a polite opening. With it, the model has to finish the concession. What it writes next is either a real boundary on where the trade-off stops mattering, or an admission that your material does not support one. Either is worth knowing before the call.

We picked an objection with a published answer. First we asked for the concede-and-reframe response with nothing in the window. Then we ran it again with the two relevant pages.

Nothing in the first response is false. Sequences and custom reporting really are behind the paid tiers. A buyer who checked would find each boundary where the answer said it was. It is a good answer to a question nobody asked.

The one line that ends the argument

The competitor states the limit themselves. That is what makes it the strongest answer available. A claim you make about a rival invites a check. A sentence quoted from the rival’s own page ends the conversation, because the only party who could argue wrote it.

Keep that beyond this example. The best answer to a competitive objection is very often something the competitor published about themselves.

The second miss is quieter and costs more. The answer points the buyer at the expensive tier and never mentions the cheap one. That kind of framing survives an internal review. It does not survive a search. The buyer finds the lower number in under a minute, and after that every figure your rep gives gets checked before it is believed.

It is three objections wearing one sentence. A model picks the most common reading and answers it without telling you it chose. That is how an objection library fills up with price arguments aimed at buyers who were never arguing about price.

The fix is not a better answer. It is making the model name the reading it picked, and quote the line in the transcript that put it there.

Sorting objections into types is older than any of this, and it belongs somewhere durable rather than being redone per deal. A sales objection framework is the version that survives a change of rep. What a prompt adds is the instruction: name the reading, quote the sentence it came from. A wrong reading then shows up before it becomes the house answer.

The objections themselves barely move. Buyers have asked about free tiers and seat minimums for years. What moves is the answer. Every response built on a competitor’s tier, limit or free plan sits on a page that changes without warning.

That matters more here than elsewhere. A stale battlecard is read silently. A stale objection response gets spoken, to somebody who probably has the competitor’s page open while you speak. That is the worst possible audience for an out-of-date fact.

A second kind of decay gets noticed even less. Objections your team stopped hearing stay in the library, because nothing removes them, and reps quietly learn that most of it is dead. Counting which entries actually get opened is what separates a library in use from one that is merely complete, and battlecard adoption is the number that does the separating.

Flares watches the competitor pricing and packaging pages these answers rest on. A changed tier or a withdrawn free plan reaches you dated, instead of reaching a rep mid-call. Deciding what the change means is still your job.

Which is the argument for storing the verdict alongside every response you keep. An answer with its verdict and its source line attached can be re-checked in a minute. An answer stored on its own can only be re-argued, and the person doing that will be a rep, in a deal, with a buyer waiting.