A useful brief starts with the decision.

A useful brief explains what the organization needs to understand, decide, or do. It does not need to begin with a tool, a model, or a technical architecture.

For a strategy question, describe the choice, the options being considered, and the constraints that make it consequential. For research, identify what is known, what remains uncertain, and how an answer would change the decision.

For an AI project, describe the present process before proposing an automated one. Who takes part, which information do they use, and what result moves to the next person?

Possible uses such as finding information, drafting, classifying material, or supporting an internal workflow raise different questions. Treat these as starting points for defining a need, rather than as a reason to adopt AI without an organizational purpose.

Explain what a useful result would look like and how someone could judge it. Identify where human review matters and who would own the system after it becomes part of daily operations.

Also describe information access, existing systems, and integration dependencies. A system that works in isolation still needs a place in the organization. Its quality, cost, and ongoing responsibility belong in the same conversation as its initial design.

Describe work already considered, important constraints, and the intended timing. Explain who needs to use the result and what they need to accomplish with it.

Halsted, Haines & Co. connects business strategy and technology through consulting, advisory, research, and AI development. Use the contact page to introduce the question; an unresolved need can be a useful starting point.

Prepare a short working brief using these fields. It should explain the situation well enough for someone outside your team to understand the need.

1. Business decision or process. What must someone understand, decide, or do? Identify the person responsible and the people affected.
2. Current workflow. List the main steps, the information entering each step, and the result passed onward. Note where people spend time correcting or finding information.
3. Desired outcome. Describe an observable improvement, with the current baseline if known. Identify quality requirements alongside speed.
4. Available information. Name the sources, their owners, access restrictions, and any known gaps. Explain whether information changes frequently.
5. Proposed system role. State what AI might assist with and which decisions or actions must remain under human control.
6. Constraints and ownership. Include timing, existing systems, integration dependencies, operating cost concerns, and who would own the result after introduction.

For example, a team might want to reduce the time spent finding an internal policy. “Use AI for knowledge management” is broad. “Help staff locate the current policy and its source, while routing unanswered questions to the policy owner” describes a task that can be examined.

A possible success measure would compare time to find a correct answer against the current process. Quality checks would also ask whether the answer cites the right source, respects information access, and declines to guess when the information is absent. Choose the measures and acceptable results for the actual business context.

Collect representative questions or inputs from the real workflow, with the expected outcome or review criteria for each. Include ordinary examples, ambiguous cases, missing information, and cases the system should hand to a person. Use information the organization is authorized to use.

Keep some examples separate from the material used to adjust the system so evaluation tests more than familiarity with the examples. A useful review considers errors and omissions as well as successful demonstrations.

Define who receives an escalated case, what information they need, and how the original user knows the next step. Also decide who will review quality as source information or the surrounding process changes. These responsibilities belong in the project brief alongside the initial idea.