The Two-Week Pilot a Cautious Owner Will Say Yes To
A stranger does not want your platform. They want to see one problem get smaller on their own numbers. Here is how to build that proof in two weeks and charge for it.
A cautious owner is not hard to sell to because they are unfriendly. They are cautious because they have paid for things that did not work. When a stranger arrives with an AI offer, the owner is not asking how clever it is. They are asking whether the specific problem costing them time or money right now will get smaller, in a way they can see.
That question is the whole job of a pilot. A pilot is not a sales pitch shrunk down to fit a smaller budget. It is a small, contained proof that your work actually works. You run it on their real data, in their real workflow, before either of you commits to anything bigger.
This piece is about how to build one. The earlier pieces covered the problem and the deliverable, the boundaries before the first call, and the choice between a build and a retainer. The last one covered the scope page that protects a project once it starts. This one covers the smallest version of an engagement that a stranger can say yes to without calling a meeting.
Why two weeks
The two-week window is not arbitrary. It is short enough that the prospect does not have to trust you first. They only have to be curious enough to try, which is a much lower bar than trust.
It is also long enough to do something real. In two weeks you can build one small thing, run it against actual inputs, and show a before and after. A shorter test would only be a demo, and a demo runs on clean sample data that proves nothing about the owner’s messy week.
Compare that with how the rest of the market prices this work. The pricing guides this series draws on describe full AI consulting engagements running four to six weeks. They tend to split that time into thirds: two weeks of discovery and planning, two of implementation and two of training at the back end. That is a sensible structure for a company with a committee and a budget line.
The pilot compresses that whole arc into something a stranger can agree to alone. The owner does not need approval from anyone, because the money is small and the time is short. You are not competing with the big engagement. You are offering the thing a small owner can actually say yes to this week.
What the owner is actually deciding
Look at the decision from the owner’s side of the table. They are not deciding whether AI is worth it. They are deciding whether this person, with this offer, is worth a small amount of money and a few hours of their attention.
Three worries sit underneath that choice. The owner wonders whether they will lose money they cannot get back. They wonder whether this will take over their week, and whether they can stop if it is not working. A big engagement answers none of them, because it asks the owner to commit first and learn later.
A pilot answers all three at once. The money is small enough to forget about. The time asked of the owner is a few short conversations, and the end date is already written down. Nothing in it requires the owner to believe your claims, which is why a cautious person can say yes to it.
The one sentence test
Most pilots fail because they are too big, not because the work is bad. The agencies that overcomplicate this step tend to make the same mistake. They treat a small business prospect as if they need an enterprise rebuild.
One write-up on agency pricing put it plainly: a small business does not need a $25,000 AI infrastructure build. They need one workflow that removes five hours of manual work per week. That sentence is a good filter for scoping any pilot you design.
Try it on your own idea. If you cannot describe the result in one sentence, the pilot is too big. “We will modernize their operations” fails. “Late invoices will get a polite reminder on day three, day ten and day twenty-one, so the owner stops chasing them by hand” passes.
The invoice reminder system is the invented example from the earlier pieces, so its shape should be familiar. The sentence names one workflow, one person who benefits and one thing that stops happening. A stranger can read it in ten seconds and know whether it matches their problem.
Writing the sentence also does something for you. It tells you what you will leave out. Every feature that does not serve that one sentence goes on a list for later. That list is what you bring to the owner when the pilot is over.
What happens in the two weeks
The days of a pilot are simple, and most of the work is not building. It is getting the right inputs and agreeing on what counts as better. Those two jobs fill most of the first week.
The first few days go to getting real material. You need the actual invoice list, or the actual inbox, or the actual spreadsheet, not a sample. You also need the owner’s honest description of how the job gets done today, including the awkward parts. This is the stage where a prospect finds out whether you listen, and where you find out whether the problem matches the one you assumed.
Send a short note on the first day that says what you need and by when. In the invented invoice example it might read: “To start, I need the list of open invoices from the last three months by Wednesday. I also need twenty minutes on Thursday to see how you chase them today.” A dated request keeps the two weeks from quietly turning into four. Most pilots slow down while they wait for a file.
Then comes the build, and it should be smaller than it feels like it should be. One narrow workflow, wired to their real inputs, doing the one thing the sentence promised. Resist the urge to add the second feature while the first is unfinished. A pilot that does one thing and works is worth more than a pilot that does four things and almost works.
The second week is for running it. Let it handle real inputs for a few days while you watch what it does. Models make mistakes, so this is also when you find the cases the build misses. Fix them, or write them down as not covered.
The last day is for showing the result. You sit down with the owner and put the before and after side by side, using their own numbers. Everything before that day exists to make that conversation short.
The before number
A pilot proves nothing without a number from before it started. This is the part that is easy to skip because it feels like paperwork. It is also the part that makes the whole thing believable.
Take the measurement in the first days, before you build anything. For the invoice example, the number might be the hours the owner spends chasing late payments in a normal week. It might also be how many days an invoice usually sits past due. Ask the owner to say the number out loud, then check it against something written down. People guess their own time badly, and the written number is the one that holds up at the end.
An invented version shows the shape. Say the owner reports about five hours a week spent on chasing. The invoice list shows the typical late invoice is paid about nine days after it comes due. After two weeks of reminders, the weekly chasing time is closer to one hour, and the late invoices are paid a few days sooner. Those figures are made up for illustration, and your pilot will produce its own.
Get Eddie Larsen’s stories in your inbox
Join Medium for free to get updates from this writer.
The point is that the owner is looking at their own numbers, not yours. A result they measured with you is harder to dismiss than a result you describe. It is also what makes the next conversation easy, because you are no longer asking them to trust a promise.
One caution about the number. Pick a measure the pilot can actually move in two weeks. Annual revenue will not change in that time, but hours spent and days late can. If no number will visibly shift in fourteen days, the workflow is probably too big for a pilot.
Price it low, but never free
The natural instinct with a stranger is to offer the pilot for nothing, to remove every reason to say no. It is a mistake, and the reason is easy to miss. Free work invites free attention, the kind that gets pushed aside the moment something else comes up.
An owner who has paid nothing has little reason to hand over the invoice list or answer your questions on day three. The pilot stalls, and then the free work looks like it did not work. A modest fee changes that, because the owner now has a stake in it getting done.
A modest paid pilot, even a few hundred dollars, signals that both sides are actually in it. In the invented invoice example, call it $400. That is a number a small owner can approve without a meeting, and it is also enough that they will return your messages.
The fee also gives you room to leave. If the fit is wrong, you can say so and stop, since nobody owes anyone a long relationship over a two-week test. The price makes the pilot a clean transaction in both directions, which is a good thing for a first engagement between strangers.
An earlier piece in this series covered the small paid first step before the build, a flat-fee review that ends with a one-page plan. The pilot is a different thing, and it comes after that kind of conversation or replaces it. The review produces a plan on paper. The pilot produces a working piece of the real thing, measured on real inputs.
Whatever number you choose, say it plainly and early. Do not apologize for it, and do not present it as a discount off something bigger. A pilot has its own price because it is its own product.
Keep it from growing
A cheap, friendly pilot is where scope creep hides best. The owner is excited, the price was small, and the requests start arriving. Someone asks if it can also handle quotes. Someone else asks if it can text the customer.
The scope page from the previous piece belongs here as much as in a larger project. For a pilot it can be half a page. It holds the one sentence from earlier, the two weeks, the number you are measuring and what is not included. Send it before the work starts, and walk through it in five minutes. Put the not-included list on it too, so a request for text messages in week two meets a line the owner already read.
When a large request shows up during the two weeks, the answer is simple. It is a good idea, and it goes on the list for after the pilot. That is not a refusal, because the pilot is exactly the place where a large request becomes its own small, priced project.
Set an end date as well as a start date. A pilot with no finish line becomes a free trial that never ends. Write down the day you will sit down with the owner to look at the numbers, and treat that meeting as fixed.
The result is a decision
The pilot should end with a short meeting and a clear choice. The work either did what the sentence promised or it did not, and both are useful outcomes. You put the before and after side by side, and then you ask what the owner wants to do next.
There are three honest answers. They can keep what you built as it is and move to a small monthly arrangement. They can extend it into a larger piece of work, scoped properly this time with what you both learned. Or they can stop, with something useful in hand and no hard feelings.
Stopping is a normal outcome, and saying so out loud is part of what makes the pilot easy to accept. An owner who does not feel trapped is far more willing to try. It also keeps you honest, because a pilot that always recommends a bigger engagement is a sales pitch wearing a different label.
If it worked, the next conversation is the easiest one you will have with that client. They have seen the result, on their own numbers, and they have a rough idea of what you charge. You are no longer a stranger with an offer.
The ceiling you are working toward
Here is what sits at the other end, so a small first number does not make you feel small. Larger pilot engagements in the AI consulting market run considerably higher, sometimes $15,000 to $50,000. That covers end to end implementation of one or two use cases. It includes tool configuration, workflow design, data integration, team training and performance monitoring.
That is not the number you are chasing with your first pilot. It is the ceiling you are working toward. You reach it once the small pilot has proven the work holds up and the client has seen it happen with their own numbers. Nobody pays that to a stranger on faith.
The path runs in the other direction. A small, measured win makes the next price easier to name. A client who has seen the result is easier to quote than a prospect who is guessing. The small pilot is not a discount on the big work. It is how the big work becomes possible.
This matters for how you hold your own pricing, too. If a pilot takes two weeks and a few hundred dollars, it will not pay for much of your time. That is fine if you treat it as the cost of the proof and can see the next step that makes the math work.
What this costs you
A small paid pilot costs you something, and it is better to know the price before you start. You will do more work than the fee covers. The pilot is a few hundred dollars of billing against many more hours of your attention, and that gap is the price of the proof.
Some owners will not pay, even a small amount, for something they cannot yet see. You will lose some of those. The ones who say yes will usually also answer your messages and hand over the right file. That makes them better first clients.
A pilot can also show that the work does not help. The workflow may turn out to be a small piece of a bigger problem, or the data may be too messy to be useful. That is a real result, and it is better learned in two weeks and a few hundred dollars than over a six-week build.
Handle the owner’s data with care, even in a short test. You are looking at real invoices, real customers and real inboxes. Agree up front on who sees what and what happens to it when the two weeks end.
You now have a way to offer a stranger something small, real and measurable. The next piece takes up the question that comes right after: what you promise when the pilot works. It also covers the parts of the outcome you clearly do not promise, and the small guarantee a solo builder can make without flinching.
Thanks for reading, now go build something fun.
Eddie
P.S. A pilot starts with someone who is willing to talk to you. The free chapter is How the First Client Actually Happens: how to build the list of thirty to sixty people who have already seen you finish something, and the first message to send them. Twenty minutes to read, one evening to build the list. The rest of the series is at e2larsen.gumroad.com.
