AI Workflow Prototype Sprint
Build and test one AI workflow before committing to a larger product. The result is evidence for a decision, not a production launch.
What it is
Test the workflow before the larger build
Turn one AI workflow idea into a working thin slice that can be tried, measured, and used to decide what to build next.
What you receive
A working prototype and a clear next decision
- Thin end-to-end workflow with representative inputs
- Agreed evaluation cases and observed limitations
- Technical approach and next-build recommendation
- Source, setup notes, and handoff
Production hardening, rollout, and ongoing support are separate engagements.
What I need from you
People, evidence, and safe access
- One named user, workflow, and decision to inform
- Representative inputs and known failure cases
- A product or workflow owner who can review progress
- Named, least-privilege, customer-managed access where needed
- Customer approval of AI providers and permitted data
Access is named, least-privilege, and customer-managed; shared personal credentials are never used.
The boundary
What it does not include
This is a focused learning and validation engagement. It does not turn the workflow into a supported production system.
The proposal names the user, workflow boundary, prototype hypothesis, representative inputs, and evidence needed for the next decision.
How it works
From question to evidence
Each step should reduce uncertainty about the workflow and the next investment.
- Name the decision
We agree on the user, workflow, hypothesis, and evidence needed to decide what comes next.
- Build the thin slice
I connect the smallest useful path across interface, model, data, and integration points.
- Test representative cases
We review agreed examples, failure cases, observed behavior, and practical limitations.
- Choose the next step
You receive the source, setup notes, evidence readout, and a recommendation to build, narrow, redesign, or stop.
Price and boundary
Price one testable workflow
The starting price covers one agreed workflow and the standard output above. Extra integrations, data work, or prototype variants are priced in the written proposal.
Normal studio tooling is included unless agreed otherwise.
Customer-specific licenses, API usage, cloud costs, and travel are stated in the proposal.
Review points are stated in the proposal; the public offer makes no unlimited revision promise.
Starting prices are set independently in each currency, not converted live. Applicable taxes are confirmed in the proposal.
Good fit
A useful prototype when
- The user and workflow are named.
- A working thin slice would answer a real product or investment question.
- Representative inputs and a review owner are available.
- The team can accept a prototype rather than a production launch.
Before work starts
Confirm the question, then begin
The website is a commercial summary. The written proposal and contract confirm the binding scope, timing, price, taxes, expenses, and terms.
- Send a short email
Describe the desired workflow, current blockage, constraints, and decision-makers.
- Check the fit
We use a short call to confirm the smallest useful scope.
- Review the documents
Paid work follows a written proposal and contract, then an invoice.
Discuss the prototype sprint
Tell me what the prototype needs to prove, which representative cases are available, and who will make the next decision.