How to Ask for a Commitment Without Pitching Your Idea

Strip away the language and a commitment has four parts. If any part is missing, you do not have a commitment, you have a mood.

The action should cost the other person almost nothing. This matters more than most people realize. In his discussion of AI and customer interviews, Rob Fitzpatrick points out that when you pay people for their time you lose the signal about whether they actually care, because now they have a reason to answer that is unrelated to the topic. The same logic applies to asks: the smaller and more concrete you make the next step, the more the acceptance itself tells you something. A 15-minute call is a cheap yes. "Let's definitely stay in touch" is a free non-answer.

Time should be a real slot, not a range. "Sometime next week" gives the other person an exit that costs no social capital: they can agree and never follow up. "Friday at 10" forces a decision now.

The stop is the part people skip. After you ask, say nothing. Count to three in your head. Most people fill silence with a mini pitch ("and I could also walk you through the roadmap, and there's a deck..."), which converts a clean ask into a sales conversation and gives the other person permission to retreat into politeness. Silence is the ask's enforcement mechanism.

The shape overall: name the ask, name the person, name the time, stop. One sentence. No justification beyond a clause about why it helps them.

Most closes fall into one of three categories: a next meeting, an introduction, or a decision. Each needs different language, because each asks the other person to spend a different kind of capital.

Script 1: The next meeting

Use this when the interview surfaced a problem worth digging into and you are not ready (or allowed) to pitch.

"This was genuinely useful. Would it help if I sent you a two-page summary of what I heard by Thursday, and we speak for 15 minutes on Friday at 10 to check whether I got it right?"

Why it works: the deliverable is about them, not your product. It gives you a second conversation with a legitimate reason to exist, and the summary doubles as written evidence of your listening. The time is exact.

If they hesitate, drop the deliverable and keep the slot: "Or skip the summary. Fifteen minutes Friday at 10 just to go a level deeper on the approval step you described?"

Script 2: The introduction

This is the highest-value close in most interviews and the one people fumble worst. At the end of an interview, the Desert and Pacific I-Corps Hub recommends asking whether the interviewee knows someone else you should talk to and would introduce you, noting that a warm introduction beats cold outreach. Their follow-up tip is the operational detail that makes it work: send a thank-you email that reminds them of the introduction and includes a short blurb they can paste into the intro email.

The close, spoken live:

"You mentioned Dana owns the budget for this. Would you be comfortable introducing me to her? I can send you three lines you can forward as-is, so it takes you thirty seconds."

Then send exactly that, the same day, no longer than four sentences: who you are, what you are trying to understand, one question, and an explicit out ("happy to be told no").

The named-person test applies here too. "I'll connect you with someone on the team" is not an introduction. "Dana Whitfield, she owns the budget" is.

Script 3: The decision

Use this when the conversation has run long enough that the other person expects to hear about what you are building, and you need a real answer rather than a compliment.

"Based on what you described, I think there are two ways this could go, and I would rather you tell me which. Would it be useful to show you a rough version on Tuesday at 2 for 20 minutes, with the explicit understanding that you will tell me if it misses? Or is the honest answer that this is not a priority this quarter?"

Notice what this does. It offers two commitments, one to look and one to decline, so the other person can say no without rejecting you personally. A decline with a reason is worth more than a vague yes, which is the same principle that makes bad news good news in a discovery program: the I-Corps write-up calls finding out your innovation does not fit the market a success, because it saves the months or years you would otherwise spend on a product with no customer.

You cannot audit a yes by tone of voice. You audit it by whether something exists afterward that did not exist before the close.

The cheapest test, useful halfway through the close, is a single question: "What would need to be true for that to happen?" A real yes produces constraints you can work with: "I'd need my manager to sign off, and she's back Monday." A polite yes produces fog: "Yeah, I think that could work, let's see how things go."

A second test is the reverse commitment. Ask them to do one small thing before the next meeting: send the document they mentioned, forward the thread, or check a date with a colleague. People who intend to engage do small things. People being polite do not.

There is a sharper version of this that costs money and therefore filters hard. Fitzpatrick's notes, collected in a public summary of The Mom Test, make the point that if you are far enough along, literally asking for money (a deposit or pre-order) is how you know they were telling the truth, because the order in hand is not an opinion. Most interview closes are far too early for that, but the principle scales down: the more the yes costs (a slot, a name, a document), the more it means.

One caution about reading signals too eagerly. Interviewing enthusiasts rather than mainstream buyers produces stronger signals than the market will, according to the Harvard guide, so a real yes from an enthusiast is real enthusiasm and not necessarily a real market. Weight your yeses by whether the person resembles the buyer you actually need.

Suppose you run 12 interviews in a week. Here is what the raw counts look like and what they mean.

• 12 interviews held.
• 7 end with "sounds good, let's stay in touch." No date. Count as zero.
• 3 end with a booked 15-minute call. Count as three commitments.
• 1 ends with a named introduction to a budget owner, plus a forwarded blurb. Count as one commitment, worth more than a call because it buys access you cannot buy.
• 1 ends with an explicit no and a reason ("we already solved this with an internal tool"). Count as a result, not a failure. The I-Corps guidance treats this as a win because it redirects you early.

So your honest pipeline from 12 conversations is four commitments and one documented no. That is a far more useful number than "12 people were interested," and it is the number to track week over week. If the count of dated, named commitments does not grow while interview count does, your questions are producing warmth rather than need, and the close is telling you so before you build anything.

A useful decision rule from this: if fewer than one in four interviews produces a dated or named commitment, stop scheduling new interviews and fix the close first. You are generating conversations faster than you are converting them into information.

A no is data, and the only thing that ruins it is arguing. Here is the sequence.

1. Thank them and accept it plainly. "That's useful to hear, thank you." No second attempt in the same breath.
2. Ask one diagnostic question. "Can I ask what would make this a yes later?" or "What would have to change on your side?" This is the same question as the real-yes test, used after the fact.
3. Record the reason verbatim. Not "not interested" but "no budget until the new fiscal year" or "we just signed a two-year contract with a vendor." Verbatim reasons cluster, and the cluster is the finding.
4. Ask for the introduction anyway, if it fits. A no on buying is not a no on connecting you with someone who has the problem worse.
5. Set a re-contact condition, not a date. "Can I check back when your renewal comes up?" beats "I'll follow up in a month," because the first has a trigger attached.

The checklist version:

• Accepted the no without a rebuttal
• Asked what would make it a yes
• Wrote down the reason in their words
• Asked for one introduction
• Logged a re-contact trigger
• Sent a two-line thank-you the same day

What you must not do is treat the no as a negotiation. Pushing after a no teaches the other person that honesty costs them time, and the next time you ask for an interview they will decline.

The close is a skill with a rehearsal path. Fitzpatrick describes running mock conversations in workshops where one person plays the customer, one interviews, and a third observes and marks where the conversation drifted into a pitch, noting that people find it hard to stay off their own idea even when the idea was made up two minutes earlier. The close belongs in that rehearsal.

A simple loop that takes 20 minutes:

1. Write your three closes on one page (meeting, intro, decision).
2. Have a colleague play a busy buyer. Give them one constraint:
   "You are mildly interested but will not commit to anything."
3. Deliver the close. Your colleague answers only with a polite yes
   or a fog answer. No real commitments allowed.
4. Run the test question: "What would need to be true?"
5. Observer scores three things: did you name a time? a person?
   did you stop talking after the ask?
6. Swap roles and repeat.

AI can help here, but only in the rehearsal seat. The Great Lakes I-Corps Hub publishes prompts for this, including asking a model to act as a specific customer persona while you practice, then to give feedback on your follow-up questions and on what you should have asked but did not. That is a useful use of a model. Letting one run the interview is not: Fitzpatrick's view is that a machine cannot run the interview because the interviewee disengages once they realize no caring human is on the other side.

The same idea shows up in the jobs-to-be-done tradition, where interviews are treated as a craft you build reps in rather than a script you read. A practitioner field report on putting the JTBD interview into practice describes deliberately starting with easier interviews before taking on harder ones, which maps well onto the close: practice the next-meeting close on warm conversations before you attempt the decision close on a skeptical buyer.

If you want to run practice sessions across several models and compare the feedback, a single key for text models is enough, and Quanta Inference exposes text, image, audio and embedding models through one OpenAI-compatible API.