2-Week Production Pilot

Most AI engagements start with a discovery deck and a 6-week assessment. We start with code. Two weeks, fixed fee, fixed scope: we pick one workflow, build a working prototype against your real data, and at the end you have something you can run and a recommendation memo that tells you honestly whether to keep going. If the AI isn't a fit for the workflow, we say so — and the prototype proves it.

What you walk away with

• A working AI prototype running against your real (or realistically anonymized) data — you get the code, the prompts, and a way to run it
• An eval against 10–30 cases pulled from your historical data, with the results documented honestly — including the cases where the AI got it wrong
• A recommendation memo: whether this workflow is a fit for AI today, where the risks are, what the next-step engagement would look like, and what the cost and timeline of going to production would be
• A live demo session with your team at the end of week two — Q&A, walkthrough, transparent discussion of the limitations we found
• Architecture notes covering model choice, integration shape, latency, and cost so you can have an informed conversation with any vendor going forward
• An exit option: if the recommendation is that AI isn't right for this workflow yet, you walk away with a paid-for, honest answer and no obligation to continue

How the engagement runs

1. 1

   Discovery · Days 1–3

   Kickoff call, data handoff, workflow walkthrough with your subject-matter expert. We define what the prototype will do, what it explicitly will not do, and what the eval set looks like.

2. 2

   Build · Days 4–7

   Prototype build against your data. Daily commits to a private repo, daily Slack update so you see progress in real time.

3. 3

   Test · Days 8–12

   Run the prototype against the eval set, document where it succeeds and where it fails, iterate on the failure cases. By the end of this window we know whether the use case is a fit.

4. 4

   Demo and memo · Days 13–14

   Live demo session with your team. Hand off the code, the eval results, and the recommendation memo covering whether to go to production and what that engagement would look like.

5. 5

   Next step (optional) · After day 14

   If the recommendation is to continue, we scope a production engagement — AI Product Launch, Workflow Automation, AI Chatbots & Voice Agents, or AI Integration depending on the shape of the work. The pilot fee is credited toward the next engagement.

By industry

Selected work

Recent two-week pilots.

Frequently asked

Is the pilot free?

No. It is a paid, fixed-fee engagement. Free pilots get treated like free pilots — deprioritized, never finished, never honest about failure modes. Paying for the pilot means we both treat the two weeks as real work, and you get a working prototype and an honest recommendation rather than a sales pitch.

What counts as a working prototype?

Runnable code that takes a real input from your workflow and produces a real output. Not a Figma mockup, not a slide deck, not a strategy memo. You can run it after we leave, and the eval set tells you how often it gets the answer right.

What happens if the prototype shows AI isn't a fit?

We tell you, in the recommendation memo and in the demo. The data the prototype produces in those two weeks is usually the most useful answer — it shows exactly where the AI breaks down and why. You walk away with a paid-for, honest answer instead of a vendor relationship you have to unwind.

What happens next if the pilot works?

We scope a production engagement — usually AI Product Launch, Workflow Automation, Voice Agents, or AI Integration depending on the shape of the work. The pilot fee is credited toward that next engagement. Some teams take the prototype code and the recommendation and run with it in-house instead — that's fine too.

What does it cost?

Fixed fee for the two weeks. Shared on the first discovery call. The fee is credited 100% against a follow-on engagement if you decide to continue.

How do we get our data to you safely?

Anonymized samples through a secure transfer method we set up on day one. We can also run the pilot against synthetic data modelled on your real data if anonymization is too heavy a lift for the timeline. Data is deleted at the end of the engagement unless you ask us to keep it.

Where does the prototype run?

In a private sandbox we set up for the engagement. If you want it deployed inside your cloud — AWS ca-central-1, Azure Canada Central, GCP Canada — we can do that, though it sometimes pushes the timeline. Inference providers chosen with your team based on cost, latency, and residency requirements.

Who does the work?

One or two engineers from our Toronto-based team, led by Vatsal. Same people you talk to in the kickoff call write the code. The full team is named on our team page.