How do I turn a paid pilot into a contract? Write the contract before the pilot starts

1. Blog
2. Strategy & Positioning

By Saroj Jha · November 2, 2026 · 16 min read

A paid pilot turns into a contract when the contract is settled before the pilot starts. Before day one, agree four things in writing with the person who will sign: what the pilot costs and how that fee is credited, no more than three success criteria with a baseline and a target, the contract that follows if the criteria are met (scope, price and start date), and a decision date that falls before the buyer’s budget is set. Then run the pilot to that date and read the result against what was agreed, not against what anyone hoped. Whether you sell an AI tool, a cleared medical device or an EdTech product, a pilot without those four agreements is free use with an invoice attached.

Here is how I would set up each part of a pilot so that a good result ends in a signature.

Running case: Example Co. Fictional

Example Co. is a fictional seed-stage AI company that drafts prior-authorization letters for specialty clinics, for a staff member to check and send. Its target is a health system with 10 specialty clinics, where staff spend about 40 minutes on each letter. A full contract would be 10 clinics at $9,600 a clinic a year, or $96,000 a year. The system’s revenue cycle director has agreed to a pilot in two clinics. Example Co., the health system and every number about them are fictional.

Why do paid pilots end without a contract?

Buyers ask for pilots because they want to see value before committing, and many of them pay for one. In Forrester’s survey of nearly 18,000 business buyers, more than 60% reported purchasing some form of trial, “from limited pilots to paid sandbox environments”, yet just over a third said they planned to convert to a fully paid version with the same provider^[1]. Forrester’s post does not say whether that third is a share of all buyers or of those who ran a trial, so read it as a sign that trials and contracts are different events, not as a conversion rate.

AI buyers in particular are still mostly in pilot mode. In McKinsey’s 2026 survey of 1,719 participants in 97 countries, 44% reported that AI is scaling across their enterprise, up from 38% a year earlier; larger organizations’ deployment continued to outpace smaller ones’, and larger companies were moving to scale AI agents much faster^[2]. A startup selling AI is therefore often one of several pilots inside the same buyer, competing for the few that get a budget.

In my view, pilots end without a contract for three reasons, and none of them is that the product failed. Nobody agreed in writing what success meant, the person who controls the budget was never part of the pilot, or the result arrived after the budget was set. Each section below closes one of those gaps.

Do this

List your last five pilots and mark, for each, whether success criteria were signed, whether the budget holder took part, and whether the decision date fell before the budget was set. The column with the most blanks is the one to fix first.

What should a pilot cost?

Charge for the pilot, and price it so that someone with a budget has to approve the spend. A free pilot is easy to start and easy to forget. A paid one puts a purchase order, a budget line and a named approver on the buyer’s side before you begin, which is most of what a contract needs later.

My rule of thumb, not a benchmark: price the pilot to cover what it costs you to deliver, and credit the full fee against the first year if the buyer signs within 30 days of the decision date. The credit makes the fee a down payment, and the window gives the buyer a reason to decide on time.

Two rules change what a device company can charge. An investigational device cannot be sold as a product: FDA’s regulation bars charging subjects or investigators more than is needed to recover the costs of making, researching, developing and handling it^[3].

Once a device is on the market, a free or deeply discounted hospital evaluation can raise questions under the federal Anti-Kickback Statute. It makes it a felony to knowingly and willfully offer remuneration to induce orders of items payable under a federal health care program, with a fine of up to $100,000, up to 10 years in prison, or both; properly disclosed discounts are an exception^[4]. Have counsel check any free units or discounts before you offer them. This is orientation, not legal advice.

Example Co., step 1 Fictional

Example Co. estimates the two-clinic pilot will take 120 hours of its team’s time to set up, support and report, $12,000 at an internal rate of $100 an hour (120 × $100). It prices the pilot at $15,000, which covers that cost (15,000 ÷ 12,000 = 125%) and is 15.6% of the full annual contract (15,000 ÷ 96,000). The fee is credited in full if the system signs within 30 days of the decision date, so year one would cost the system $81,000 (96,000 − 15,000).

Do this

Write down what your next pilot costs you to deliver, price it at no less than that, and put the credit and its deadline in the pilot agreement.

What should the success criteria say?

Success criteria are the clause that decides whether the pilot converts, so write them like a contract clause. Each criterion needs four things: a number and a stated way of counting it, a baseline measured before the pilot starts, an owner on the buyer’s side who will share the data, and a signature. If any of the four is missing, the result can be disputed after the fact.

Keep to three criteria at most, and choose ones that can move within the pilot. In my view, one measure of the work itself (time, errors, throughput), one of quality or safety that must not get worse, and one of adoption is enough. A satisfaction score alone will not survive a budget meeting, and annual retention cannot be read in twelve weeks.

Sector shapes the measures, not the rule. For an AI product, the work measure is often minutes saved and the quality measure is how often a person accepts the output unchanged; the Marketing AI Products Playbook sets out how to log acceptance. For a medical device, the MedTech Marketing Playbook explains why an evaluation should report a process and a cost measure at the hospital’s own volume. For a school district, use the measures the district already trusts; my post on selling EdTech to school districts covers those in its pilot step.

Example Co., step 2 Fictional

Example Co. and the clinic managers agree three criteria. First, staff minutes per letter in the two pilot clinics fall from a 40-minute baseline to 20 or fewer, measured in weeks 9 to 12. Second, the share of letters accepted by payers on first submission stays at or above the 88% baseline. Third, at least 80% of drafts are sent with only minor edits. The revenue cycle analyst owns the data, and all three are signed before kickoff.

Do this

Run each criterion through the four checks (number, baseline, owner, signature) and cut any criterion beyond the third.

Who has to sponsor the pilot?

The sponsor is the person who can sign the contract that follows, not the person who likes the product most. Most pilots have a champion, a user or manager who wants the product to win, and that person matters. But a champion who cannot commit next year’s budget can only forward your result, and forwarding is where pilots stall.

Buying groups are large and getting larger. In the same Forrester post, purchases that included generative AI features had buying groups twice the size of those without, 14 members against seven, and procurement professionals served as decision-makers more than half the time (53%)^[1]. In my view, that means naming two people on the pilot agreement: the champion who runs the pilot day to day, and the economic buyer who agrees in advance that meeting the criteria leads to the contract. The B2B Buying Committee post covers how to reach the others.

Ask for the sponsor before you agree the pilot’s price. A buyer who will not name the person who signs is telling you the pilot is exploratory, and in my view that calls for a shorter, cheaper test, not a full pilot and hope.

Example Co., step 3 Fictional

The champion is the clinic operations manager, who runs the pilot. The sponsor is the revenue cycle director, who owns the budget line and signs the pilot agreement. Before kickoff, Example Co. also asks the system’s procurement lead what a $96,000 contract needs: a security review, a business associate agreement and the director’s signature.

Do this

Write two names on your next pilot agreement, the champion and the economic buyer, and do not schedule kickoff until the economic buyer has signed it.

How do I agree the conversion path before the pilot starts?

Put the contract that follows into the pilot agreement. That means the scope of the rollout, the price, the start date, the length of the term, and the sentence that says what happens when the criteria are met. If the buyer has already reviewed those terms, the decision at the end of the pilot is yes or no, not the opening of a new negotiation.

Write three outcomes down, not one. In my view, the agreement should say what happens if the criteria are met (the contract starts on the stated date at the stated price), if they are partly met (one extension, with a new end date and the same criteria), and if they are missed (the pilot ends and both sides say why). A buyer who will not sign the first outcome in advance is unlikely to sign it later.

Example Co., step 4 Fictional

The pilot agreement states that if all three criteria are met, the system signs a one-year contract for all 10 clinics at $9,600 a clinic, starting on 1 July 2027, with the $15,000 pilot fee credited. If one criterion is missed, the system still reserves the contract’s budget at the 15 April lock, and the pilot is extended once by four weeks with the same criteria, moving the decision to 29 April. If two or more are missed, the pilot ends.

Do this

Add a one-page “if met, if partly met, if missed” section to your pilot agreement, with the price and start date of the contract filled in.

How long should a pilot run?

Long enough for the measure to move and settle, and short enough to finish before the buyer sets next year’s budget. Work it backwards from the budget date, not forwards from the day you are ready. Count back from the date the buyer locks its budget to the decision, the readout and the end of the pilot, and the kickoff date falls out.

The budget date belongs to the buyer, so ask for it. In my view, the commonest reason a successful pilot slips a year is a readout that lands after that date. For districts, my EdTech marketing calendar sets out when they plan and buy; for hospitals, AAJ’s report on how US hospitals buy medical devices explains why no public source can tell you how long a committee takes, so you have to ask each one.

Time-boxing also means not extending by default. A pilot that is extended twice has become free use with a reporting burden, and the buyer has learned that the deadline was soft. Allow one extension, only for a criterion that was partly met, and only with a new date in writing.

Example Co., step 5 Fictional

The system locks its budget for the year starting 1 July on 15 April 2027. Working back, with two weeks from decision to budget lock, two weeks from readout to decision and one week from pilot end to readout, the decision is 1 April, the readout 18 March and the pilot ends 11 March. A 12-week pilot therefore has to start by 17 December 2026, with a mid-point review on 28 January. The holiday weeks fall early in the pilot, which is one reason the minutes are measured in weeks 9 to 12.

Do this

Ask your buyer for the date its budget is locked, and plan the pilot backwards from it before you agree a start date.

How do I read the result?

Read it against the criteria you signed, and against what would have happened anyway. A before-and-after number in the pilot group mixes your effect with everything else that changed in those weeks. Where the buyer allows it, keep a similar team or site on its current process over the same weeks, and report the change in the pilot group minus the change in the comparison group.

Then write the readout yourself, on one page the champion can forward unchanged: the criteria, the baseline, the result, the decision the agreement now calls for, and what the contract costs. In my view, a met result should be presented as a decision already made, because it was made when the agreement was signed.

Be honest about what the result shows. Time freed is worth money only if the buyer uses it, so show the full value and a conservative case. A missed criterion should be reported plainly; it is better evidence for the extension, and for your next buyer, than a rounded-up result.

Example Co., step 6 Fictional

In weeks 9 to 12, the pilot clinics average 18 minutes a letter against their 40-minute baseline, a change of −22 minutes. A third clinic that stayed on its current process went from 38 to 37 minutes, a change of −1. The effect is (18 − 40) − (37 − 38) = −21 minutes a letter. First-pass payer acceptance is 89% (baseline 88%) and 84% of drafts were sent with minor edits only, so all three criteria are met. At 60 letters a week over 50 weeks, 21 minutes saved is 63,000 minutes, or 1,050 staff hours a clinic a year (21 × 60 × 50 ÷ 60). At a loaded cost of $45 an hour that is $47,250 a clinic, about 4.9 times the $9,600 price; counting only a third of it, $15,750, it is still about 1.6 times the price. On 1 April the director signs the contract the agreement described.

Do this

Agree with your champion before the pilot ends which comparison group the readout will use, and draft the one-page readout template now.

What changes by stage?

• Pre-seed — Run a few short, paid pilots by hand, and use them to learn which criteria buyers will sign and who signs the contract. Do not discount to zero to get a logo.
• Seed — Standardize the pilot: one agreement template with the four agreements, a fixed length, and a readout template. Track how many pilots convert and why the others did not.
• Series A–B — Make the pilot the first phase of the contract rather than a separate deal, with the rollout terms pre-signed, and give one person the job of reading every pilot’s result against its criteria.

Free tool

Work the pilot back from the budget date

The free Pilot-to-Contract Planner works a pilot backwards from the buyer’s budget date, checks each success criterion for a number, a baseline, an owner and a signature, and flags a missing economic buyer or conversion terms before you start.

The two illustrative engagements on AAJ show the same pattern in different markets: ProofDesk runs a pilot on one real questionnaire with a written bar and a decision date, and Example Co.’s first hospital, a different fictional company from the one in this article, agrees a 60-day evaluation with a named owner and a decision date. For a school district, the pilot step in How to sell EdTech to school districts adds the district’s own measures and budget calendar.

If you want me to look at your own funnel, positioning and channels, my Growth Audit ends with three prioritized fixes and a 30-day action plan.