Consulting
A working prototype in four weeks, with its limits written down
For the owner who wants to see an AI agent do a real job in the business before paying to build it properly. We build it on your own material in four weeks. You get a working system, a written statement of what it does and does not prove, and a recommendation.
Who it suits
When a demo is not enough to decide
It suits the owner, partner or manager who has to be convinced before the business spends more.
A vendor demo runs on the vendor's sample files and answers a different question than yours. A prototype built on your own documents, requests or records shows how an agent handles the mess in your actual work, and where it goes wrong.
The engagement has a fixed scope and ends on a date agreed before it starts. A prototype is a test, not a first version of the real thing, and we never present it as one.
How is this different from a vendor demo?
It runs on your material, so the failures you see are yours. It comes with its own boundary in writing, including the cases it was not tested on. And it ends in a recommendation on an agreed date, so nobody has to decide when to stop paying for it.
Four weeks
What happens each week
The four weeks are set in the scope we agree with you. Weeks three and four can be shorter when your material is already easy to reach, and we agree that before we start.
Show the detail
| Week | What happens | What it produces |
|---|---|---|
| 1 | Scope, boundary and material. What the prototype will attempt, what it will not, and which files or samples it runs on. We write the boundary now, not at the end. | Scope and boundary statement, agreed in writing |
| 2 | We build against the agreed material and show you the first working version at the end of the week. | A working system you have seen operate |
| 3 | We test it on real cases with known right answers, taken from your work rather than invented. | Results across the cases, including the failures |
| 4 | Readout and decision. What it showed, what it cost, what it would take to build it properly, and our recommendation. | A decision pack with a written recommendation |
What you receive
- A working system
- Running on your material, in a setup agreed in week one, that you and the people who do the work have both used.
- The boundary statement
- What the prototype shows, what it does not, which cases it was tested on and which it was not. It travels with any screenshot or recording of the prototype.
- Results across the cases
- Including the ones it got wrong, with the pattern behind them described.
- Cost to date and cost to run
- What the four weeks used, and what the same workload would cost each month if it ran for real.
- Decision pack and recommendation
- Build it properly, fix a named prerequisite first, or stop. A recommendation to stop arrives on the same schedule and at the same price.
Warning
The expiry date, and why there is one
Every prototype carries a date after which it must not be shown as current. A system built in four weeks reflects that month's files, model versions and assumptions, and a screenshot that outlives them misleads whoever sees it next. We write the date into the boundary statement and do not maintain a prototype past it. If it is worth keeping, it is worth building properly.
Boundaries
What a prototype cannot be used for
We name these now because the pressure to use a prototype for one of them arrives after it works.
Show the detail
- Running the business
- It has no testing schedule, no change control, no monitoring and nobody on call. It must not carry real client work.
- Proof that the controls are in place
- A prototype does not have the controls a live agent needs, so it cannot show them.
- A final decision to build on its own
- It shows what is possible. Whether this workflow should go live, and under which controls, is what the governance sprint answers.
- Client or personal data without a decision to allow it
- We settle in week one what data it may touch and where it runs. Where the real material cannot be used, the prototype runs on samples or made-up cases, and the boundary statement says so.
If you have several candidate workflows and no choice yet, or the value is in doubt, the sprint is the better first step. If you have no candidate at all, we will say so rather than sell you this.
How it is sold
Sold through Consulting
The prototype is part of our consulting work, and we scope it with you before we quote.
You start with a 30-minute call at no charge. If a prototype is the right step, we write the scope and quote it before any work begins. The price depends on how many files and systems it touches, whether it uses real material or samples, and where it has to run.
Want both? We can contract the sprint and the prototype together, and the prototype goes ahead only if the sprint recommends it. Want the real thing instead? Consulting is the fixed-price route. The Diagnostic is CAD $5,000 and Core Setup is CAD $9,000, plus GST/HST. Each skill passes the test you sign, or we fix it free.
We publish no performance figures for this service, because we cannot back them. Under a non-disclosure agreement (NDA) we walk a reviewer through the method and a sample set of deliverables.
Next step
Bring the workflow and the person who has to be convinced
A 30-minute conversation shows whether a prototype is the right step, or whether the sprint answers your question more cheaply.
- The workflow, as it runs today
- Who has to see it work, and what would convince them
- The material it would run on, and whether we can get access within four weeks
- Any client or personal data in that material
- The decision the prototype should help you make, and when you make it