Key points.
Define the task, permitted inputs and required output standard before selecting technology.
Make review authority, exception handling and approved tools explicit.
Evaluate ordinary cases, errors and operating costs before expanding a deployment.
An AI use case should describe a task and the standard required of its output. Selecting a model or an interface comes later. The operating arrangements determine who can supply information, who reviews the result and what happens when the system cannot meet that standard.
Establish the conditions for use
The Bank of England and FCA's 2024 survey examines AI use alongside governance and third-party dependencies in UK financial services. Its relevance extends beyond an adoption figure: deployment introduces responsibilities that need to be understood. [1]
For an individual process, the assessment should identify approved inputs, access controls, acceptance criteria and the consequences of error. A requirement for human review needs to specify what is checked and who has the authority to accept or reject an output.
Make the policy usable
Thomson Reuters' 2026 report finds that some professionals use tools their organisation has not approved. A written policy is therefore only one part of an adoption arrangement. People also need suitable tools, training and a clear route for questions or exceptions. [2]
Integration work must account for the systems receiving the output. A process can create additional risk if an unverified result is copied into a customer record, sent externally or used as an input to another automated decision.
Tiph's assessment
Start with a bounded pilot and a documented baseline. Test ordinary cases and exceptions. Record errors, review effort and the cost of maintaining the process. Define the conditions under which the system should be paused or referred to a person.
A deployment decision should follow that evaluation. Continuing to monitor the process is part of operating it, rather than a separate exercise after launch.
Published evidence. Scope and limitations.
- [1]Bank of England & FCA / 2024Artificial intelligence in UK financial services
The joint survey reports that 75% of responding firms were already using AI.
ScopeThe findings concern responding UK financial-services firms and reflect the position at the time of the 2024 survey.
- [2]Thomson Reuters / 2026Future of Professionals Report
In the survey, 34% of professionals report using AI tools their organisation has not approved.
ScopeSelf-reported behaviour within the report's professional-services sample. It is not a measure of all employees or organisations.
Analysis of digital strategy, service design and applied AI.
Further analysis
