Assess
Review the current process and source data with your team. Agree the users, acceptance criteria, data boundaries and a feasible first scope.
- What you receive
- A scoped proposal with deliverables, assumptions and acceptance measures.
- How it is priced
- Fixed assessment fee.
Example deliverable: Scope brief
- Process and users
- Data boundaries
- Acceptance measures
Agree the first scope
Implement
Build the agreed connections, reporting or automation. Test representative cases, exceptions and recovery. Run alongside the current process where needed before cutover.
- What you receive
- A tested workflow, documentation and handover against the agreed scope.
- How it is priced
- Fixed project fee.
Example deliverable: Test & handover pack
- Working connections
- Exception and recovery tests
- Documentation and handover
Review the tested workflow
Operate
Set up scheduled runs, monitoring and incident handling. Review data quality and model performance, with clear responsibilities for support and change requests.
- What you receive
- Ongoing operation and support within a defined service agreement.
- How it is priced
- Recurring service fee.
Example deliverable: Service plan
- Run and incident monitoring
- Support responsibilities
- Change arrangements
Operate within the agreement
Expand
Use what the first implementation teaches us to scope the next useful addition. Reuse suitable components where contracts permit, while keeping client data separate.
- What you receive
- An agreed extension with its own scope and acceptance criteria.
- How it is priced
- Additional implementation and recurring fees.
Example deliverable: Extension brief
- Next useful addition
- Reusable components
- New acceptance criteria
Agree the next scope
Illustrative documents. The actual scope and terms are agreed in your proposal.
Agree the running costs before the build
Your proposal sets out usage allowances, third-party charges, hosting, support responsibilities, response arrangements and change limits. Prices follow the scope and expected operating effort; a subscription or support commitment is agreed explicitly.
Agree how success will be tested
Use representative data and realistic exceptions. For AI, keep final evaluation examples separate from training and tuning, and measure important errors, response time and cost per accepted result. Agree who reviews difficult cases and how the workflow recovers from failure.
Keep your data boundaries clear
Agree permissions, hosting and processing locations, retention and access before implementation. Reusable components do not give us permission to reuse your confidential data for another client or for model training.
Testing a system we deliver is part of delivery quality assurance. Independent evaluation of other vendors’ systems follows a separate service and conflict policy.
Read the independence ruleBring one process to the first conversation
Describe what happens today, where the data lives, who uses the result and what slows the work down. We’ll use that to identify a useful first scope.