AI Software Development
AI PoC to production
A twelve-gate path from a working demonstration to a system operations will accept
- Duration
- 6 to 12 weeks
- Ladder stage
- Product
- NATIVE stages
- Integrate, Validate, Expand
The situation
When this engagement is the right one
The gap between a working demonstration and a system operations will accept is not model quality. It is twelve unglamorous gates, each of which someone must sign.
You are probably seeing
- The PoC has been finished for two quarters and is not live.
- Security review has become an unbounded conversation.
- Operations will not take support responsibility and nobody has asked them what they need.
What we do
The work, in the order it happens
- 01Run the gate assessment: what is present, what is missing, what is disputed.
- 02Close the gates in dependency order, with a named signer for each.
- 03Rebuild what cannot be hardened, rather than layering controls on a demo.
- 04Complete operational acceptance with the team who will hold the pager.
What you get, and keep
- A twelve-gate status register with named signers.
- A hardened production deployment.
- An operational runbook accepted by the operations team.
- A post-deployment monitoring specification.
Prerequisites
- An existing PoC and access to whoever built it.
- Operations engaged before, not after, hardening starts.
Not included
- We do not certify a PoC as production-ready without closing the gates, whatever the deadline pressure.
Duration and price shape
6 to 12 weeks
From a published figure, scoped after a gate assessment.
Where this sits in the method
- IIntegrateRewriting workflows and decision rights
- VValidateProve it before you scale it
- EExpandWhen the pilot becomes reality
Commercially this is a Product engagement on the Proof, Product, Platform ladder.
FAQ
Questions we get asked
- Can you just harden what we have?
- Sometimes. Where the demo made structural assumptions that do not hold, we will tell you rebuilding is cheaper and show the arithmetic.
- What are the twelve gates?
- Workflow, decision rights, identity, data boundary, audit trail, evaluation, failure modes, cost ceiling, monitoring, runbook, support model, and handover.
Related engagements
Work that usually sits either side of this
AI Software Development
Custom AI Application Development
Production applications with authentication, data boundaries, audit trails, and an evaluation harness
AI Governance and Risk
Model Risk and Assurance
Evaluation harnesses, quality bars, drift monitoring, and audit-ready documentation for regulated workflows
Next step
Ready to scope poc to production?
Bring the pilots you already have running. The first call is a scoping conversation, not a pitch.