Most freelancers lose money on discovery calls because they ask what they want to hear, not what they need in order to price the work. A discovery call is not a sales call. It is the only conversation in which the client tells you what the work costs them if it goes wrong, and that answer, not the hours you guessed, is your price.
The nine questions below do that job. Each one forces a specific fact into the open, and the prompts underneath turn your call notes into a written scope you can send for approval.
Why most discovery calls produce the wrong price
The typical freelance call goes: what do you need, how long will it take, when can you start. Three questions, none of them about difficulty. The client answers with a mood, and you convert the mood into hours using whatever rate you had last time.
That produces the standard failure. You quote four hundred dollars for something that takes twelve hours. The client is delighted, signs, and comes back in month three with a request that assumes you had budget for a rebuild.
Scoping fails when the client describes an outcome instead of a deliverable, when the constraints are missing, or when nobody writes the scope down and the two of you remember it differently six weeks later. The questions below fix the first two. A written scope fixes the third.
The nine questions that price a job
Ask them in this order. The first three establish what success means in their words, the next four establish the constraints that decide your hours, the last two establish the commercial shape of the engagement.
- What happens if this is done well? Ask for the business outcome, not the feature list.
- What is happening now that made this urgent? Urgency usually names a deadline with a cost attached.
- Who is the audience for the output? Size of audience changes review and compliance work.
- What already exists that I must work with? Existing site, brand assets, copy, data, accounts, approvals.
- Who approves the work, and how many people? Number of approvers sets the number of revision cycles.
- What does your current process cost you? Establishes whether your fee is small relative to the pain.
- What is explicitly out of scope for this phase? The single most useful question on the list.
- What is the deadline, and what makes it fixed? A soft deadline and a hard one price differently.
- What is the approval process and the turnaround on feedback? Slow feedback eats your margin quietly.
Question seven is the one people skip because it feels rude early in a relationship. It is the one that protects the relationship later, because everything excluded in advance is a conversation you never have to have badly.
Ask client discovery questions ai prompts should never skip
Use AI before and after the call, never instead of it. Before, it turns your service description into a question set tailored to that kind of client. After, it structures the answers.
I am a [your discipline, for example freelance web developer]. My client is [client type, for example an owner of a two-location dental practice]. The project they discussed was [one-line description]. Write [ten] discovery questions for a first call. The questions must uncover: what success means in their terms, existing assets and constraints, who approves the work and how many people, what is out of scope for this phase, the deadline and whether it is fixed, and the approval turnaround. Order them so the first three feel like small talk and the last three feel like a natural continuation. Keep each question to one sentence, phrase them politely, and avoid anything that sounds like an audit.
Fill in these details
- Discipline: the model returns different questions for a designer, a copywriter, and a data consultant. Generic discipline gives generic questions.
- Client type: narrow it. "Small business owner" produces questions a fifteen-year-old could ask.
- Project description: one line. The model needs to know which of your service areas this is, or it will hedge across all of them.
- Question count: ten gives you room to drop two that do not apply.
If you are starting the engagement rather than repairing one, ask the model instead for the missing information you would need before you could quote at all. The answer list is usually the brief in disguise.
A project brief prompt ai turns into a scope you can send
After the call, paste your notes and let the model structure them. Keep the raw notes too.
Turn these discovery call notes into a written scope for [client name]'s [project name]. Notes: [paste raw notes]. Produce: (1) an objective in two sentences, (2) a deliverables list where each item is specific enough that two people would agree whether it is done, (3) an explicit out-of-scope list, (4) open questions the client still needs to answer, and (5) assumptions I made so I can correct them. Do not invent a price, a timeline, a number of revisions, or a technology. Do not add deliverables I did not mention. If a note is too vague to turn into a deliverable, list it under open questions instead of resolving it yourself.
The negative constraints do the real work. A model asked to produce a scope will happily invent a two-week timeline because a document that says "schedule TBD" feels unfinished.
Worked example: a website for a two-location practice
The client asks for "a new website, we need it soon". The nine questions produce this.
| Question | What the client said | What it decided |
|---|---|---|
| Success | "We lose calls because people cannot book online." | Deliverable is booking, not pages |
| Urgency | "Our competitor launched in July." | Real deadline, driven by a rival |
| Assets | "Photos from the last photographer, probably still usable." | Asset check needed before design |
| Approvers | "Me and my practice manager." | Two approvers, expect two rounds |
| Out of scope | "We are not moving off the current booking system yet." | No integration in this phase |
| Feedback | "I am bad at replying to email, honestly." | Every review round costs two weeks |
The last row is the whole quote. Once you know that each review round takes a fortnight, a project with three review stages is not a two-week job. It is a two-week job with two extra months of holding time attached, and it must be priced as such.
Send the structured scope back before sending a number. Clients routinely correct scope without feeling pressured, and a corrected scope is the cheapest price increase available.
What to do when the client says the budget is fixed
Roughly one call in five ends with a number the client has already decided. Do not accept it as the price. Find out what the number buys.
The client has a fixed budget of [amount] for [project]. Write a scope that fits it. Give me three options: a minimum viable version that delivers the primary outcome only, a recommended version that adds [the most requested secondary feature], and a complete version that meets everything they asked for. For each, list what is included, what is excluded, and what you would need to change about your approach to fit the budget. Do not quietly shrink my standard scope and call it the same deliverable.
Then read the options honestly. If the minimum version requires you to work below a rate that makes the project worth taking, say so and decline. A scope that fits is not always worth having.
Check the scope before you send the quote
1. Test each deliverable against the word done
Read each item and ask whether a client could dispute whether it is complete. "Modern website" fails. "Five pages built and uploaded, mobile responsive, booking form connected and tested" passes. Rewrite anything that could be argued about.
2. Count the review rounds and price them separately
Approvals are work. Multiply your hourly rate by the expected hours per round and by the number of people who will comment. Most projects contain two to three rounds, and almost no quote prices them explicitly.
3. Confirm what you are excluding
Read the out-of-scope list aloud to the client in the follow-up email. An exclusion that the client has heard is fair. An exclusion buried in a document they skimmed becomes a dispute later.
Frequently asked questions
How many discovery questions should a call actually have?
Enough to remove every assumption that would change your price. That is usually between eight and twelve. More than that and you lose the call to fatigue, fewer and the fee carries the guesswork. The nine above are a floor that covers most service work.
Can AI run the discovery call for me?
It cannot, and it should not. A model can generate questions, structure notes, and pressure-test your scope, but it cannot read the pause where a client says "it is not really a deadline, do not worry about it". Use it on both sides of the call.
What if the client will not answer the scoping questions?
That is information, not an obstacle. Someone who will not say what is out of scope, who approves, or how fast they reply will also not accept a change order. Quote the smaller version, put a limit on revisions in writing, and let the behaviour set your assumptions.
Run the next call with the nine questions written down
Put the list in front of you, take the notes verbatim rather than in summary, and send the written scope back within a day. The habit costs twenty minutes and it is the difference between a project that earns what you planned and one that quietly did not. For the operational side of running a one-person business around this work, see our collection of small-business prompts and operating notes.