Turn an AI idea into a decision your team can make.

Use an assessment to choose a first pilot.

This stage is useful when several AI ideas compete for attention, a prototype has unclear business value, or a team needs to understand what must be ready before implementation. Bring a description of the current task, its users and the problem you want to improve. Existing prototypes and process documents help ground the discussion.

Look at the task from four directions.

The assessment organises questions around the business outcome, usable information, technical environment and decision owners. These questions reveal whether a small pilot can provide useful evidence.

• Business: who uses the result, what the current process costs in effort, and which outcome would justify further work.
• Data: what material exists, who owns it, how it is updated and which users may access it.
• Systems: existing applications and accounts, required interfaces, access constraints and expected usage.
• Organisation: the business owner, technical contact, reviewers and the people who approve data use or changes.

Separate a promising task from its dependencies.

Compare candidate use cases by expected value, available information and implementation effort. Identify questions that need customer or specialist review, including data handling and jurisdiction requirements. Select a limited scope that can answer a meaningful question. Record external dependencies and unresolved choices so an attractive idea is not mistaken for an implementation-ready project.

Leave with a roadmap and a basis for estimates.

Agreed outputs can include a use-case shortlist, a readiness and dependency record, a proposed pilot scope and an outline of resource and delivery options. Any cost or effort estimate should state its assumptions and missing inputs. The roadmap identifies suggested owners and the decisions needed before the next stage, rather than presenting provisional dates as delivery commitments.

Review the recommendation before committing to implementation.

The customer checks that the proposed pilot addresses the real task, uses information that can be made available and has an owner for evaluation. Agree what evidence the pilot should produce and what would cause the plan to change. The assessment can lead to architecture work, further preparation or a narrower pilot. Its completion is the reviewable recommendation and decision record, with the next scope confirmed separately.