How we build
Clear responsibility. Work you can examine.
How UKBH coordinates AI tools, reviews development and sets security requirements before release.
AI development, human direction
Several AIs. One coordinated development process.
We use AI to help research, build and examine our work. Orchestration means giving different tools specific jobs, passing on the relevant context and bringing their findings together under human direction. You have one point of contact with UKBH throughout.
01
Plan around your business
We define the outcome, the information involved and what success looks like before development starts.
02
Give each AI a clear role
We coordinate tools such as ChatGPT, Codex, Claude, Gemini and Perplexity across research, planning, coding and review. The combination depends on the project; each task has a defined scope and a clear handover.
03
Challenge and verify the work
Development and review are separate tasks. A second AI can challenge an implementation, but agreement between models is not proof: we check the code, test the behaviour and review the evidence.
04
Keep people accountable
UKBH directs the work, resolves conflicting advice and remains responsible for delivery. You review the agreed customer journeys and approve the release.
Security before release
Every website and app needs a security review.
Our release requirement is a documented security review for every website and application, with checks matched to its features, data and risks. Security starts during planning and remains part of development, testing and ongoing care.
Sign-in and access
Where accounts exist, check authentication, password recovery, sessions and staff permissions. Test that one customer cannot access another customer’s records, files or reports.
Data and privacy
Review what is collected, where it goes, who can see it and how it is retained or deleted. Use synthetic or anonymised test data wherever possible and limit information shared with AI services.
Code, forms and integrations
Review input validation, file uploads, exposed credentials, dependency risks and external connections. Keep privileged operations on the server and give integrations only the access they need.
Payments and automated actions
Where relevant, check payment confirmation, repeated requests, failed transactions and the permissions behind emails, notifications and other automated actions.
Release and recovery
Record findings, fix and retest issues that block release, and review deployment settings, backups and recovery arrangements appropriate to the service.
Care after launch
Agree responsibility for updates, monitoring and responding to vulnerabilities. Significant changes need a fresh review of the affected areas.
Evidence before sign-off
Before launch, we document the checks completed, resolve findings that block release and explain any remaining limitations. Unresolved critical or high-risk security findings must block release.
For sensitive or complex projects, we identify the intended market, what data the app will handle and any need for specialist advice or independent testing. AI review and automated scans support the process alongside expert judgement.
Have questions about your project?
Tell us what your system needs to do and what information it will handle. We will explain the approach and checks it needs.


