AI Software Development
Custom AI software development
Production applications with authentication, data boundaries, audit trails, and an evaluation harness
- Duration
- 6 to 12 weeks
- Ladder stage
- Product
- NATIVE stages
- Integrate, Validate
The situation
When this engagement is the right one
Anyone can generate an application now. The scarce work is deciding what to build, wiring it into a system of record, proving it against a pass line, and making it maintainable by someone who did not write it.
You are probably seeing
- A pilot works and operations will not accept it.
- The application has no audit trail and a regulator has started asking.
- The team that built it has moved on and nobody can safely change it.
What we do
The work, in the order it happens
- 01Confirm the redesigned workflow and its decision rights before writing code.
- 02Build the application against your identity provider and data boundaries.
- 03Instrument the audit trail at decision level, not at request level.
- 04Stand up the eval harness where a model is in the loop.
- 05Run the pilot against the charter's pass line, then hand over.
What you get, and keep
- A production application in your environment, with source code and documentation.
- Authentication, authorisation, and data-boundary configuration, documented.
- An audit trail and retention configuration.
- An eval harness in CI with the initial evaluation set.
- An operational runbook and a named operational owner agreed with you.
Prerequisites
- A mapped workflow with Approve, Override, and Escalate named.
- A signed pilot charter with a pass line and a stop condition.
Not included
- We do not deploy into production without an escalation path and a kill switch.
- Ongoing managed operations are a separate agreement, not an assumption.
Duration and price shape
6 to 12 weeks
From a published figure. Three variables move it: number of integrated systems, regulatory review requirement, and whether an eval harness is required.
Where this sits in the method
Commercially this is a Product engagement on the Proof, Product, Platform ladder.
FAQ
Questions we get asked
- Who owns the code?
- You do. Code, prompts, evaluation sets, calibration data, and documentation, in exportable form, from day one.
- Do you use AI to write the code?
- Yes, where it is appropriate, under review standards we publish. Generated code is reviewed before it ships, always.
- What happens after handover?
- Your named operational owner runs it. We offer a defined support window, priced separately, never open-ended.
Related engagements
Work that usually sits either side of this
Next step
Ready to scope custom ai application development?
Bring the pilots you already have running. The first call is a scoping conversation, not a pitch.