The output of an AI discovery programme is a report, and a report is easy to make look impressive and hard to make useful. A good one has two purposes: it is the basis on which the people who took part confirm that the use cases and their priorities are right, and it carries enough detail on each use case for the business to take it into its own investment process without rework. Here is what one contains, in the order we write it.
Executive summary
What was done, in numbers: workshops, participants, ideas captured, use cases consolidated. The three findings that shape everything else. The recommended first quarter. Two pages at most, and readable by a director who reads nothing else.
The method
Enough that the reader can judge the output: the divergent and convergent approach, the ten scoring elements and how they map onto the business’s own assessment template, how ideas were consolidated into use cases, the Opportunity Pyramid, the scoring notation, and what the report decides versus what it leaves to the business. A report that cannot explain its own method is an opinion.
The adoption pathways
Research-led, product-led, system-embedded, data-foundation: what each means and why a use case follows the pathway its nature demands. This section is where “switch it on” stops being the default assumption.
Common themes across the business
The capabilities that recurred in every function, each written up as a consolidated use case in the business’s own format: the problem in the participants’ words, what it consolidates (every idea, by reference), the scorecard, the pathway, the pyramid layer, candidate vendors, the recommendation. The data finding usually sits here, because it is the cross-cutting fact: data, not technology, is the constraint.
Findings by function
For each function: the cohort profile, the clustered ideas with their scores, the function-specific use cases, and the observations and priorities. This is where participants check that their ideas were captured correctly and rolled into the right use case.
Governance, access and responsible AI
The current state of access to AI tools, the legitimate concern behind any restriction, why a blanket block is the wrong answer, the governed-access model, the responsible-AI principles, the regulatory map, and how it all aligns with the governance the business already has.
Recommendations, priorities and roadmap
The principles for proceeding. The quick wins. The full portfolio grouped by readiness, ready now, nearly ready, strategic, research, with the data foundation underneath. The immediate recommendations for the next quarter. And how the consultancy can help with the next phase, as adviser, builder or taking a pilot on, stated plainly so the business can decide.
Appendices
The full idea register, every idea by reference with its summary and the use case it rolls into, so any decision in the roadmap can be traced back to the room it came from. The vendor landscape. The regulatory detail. Who wrote it.
What to ask of a sample
Before commissioning discovery, ask the consultancy for a redacted report. Then ask: can I trace a roadmap item back to an idea someone raised? Is each use case in a format my investment process could accept? Does it say which use cases should be bought rather than built? Does it have a first quarter I could start? And does it say, anywhere, that the data is not ready for something the business wants, because if every idea is a go, nobody has looked.
The report described here is the output of the Converge Methodology; the programme that produces it is under AI discovery workshops and opportunity prioritisation.