By David Bartas · Updated
What this work helps you decide.
AI strategy consulting helps a team decide which business problem is worth testing with AI, what information the system can use and how to judge the result. I work with founders and product leaders to turn an idea into a bounded pilot with explicit assumptions, evaluation criteria and a decision about what happens next.
Where would AI improve a real workflow—and what would make the pilot worth continuing?
What to bring.
- The workflow, its users and the problem they face
- A sample of the information the system would use
- The current process and a useful comparison baseline
- Budget, timing, access restrictions and the people who own the decision
Typical deliverables.
- A prioritised shortlist of AI use cases and the reasoning behind it
- A data and delivery readiness review with unresolved questions
- A pilot brief covering scope, evaluation, ownership and next steps
The scope depends on your question and available context. A consultation frames the decision; a complete review, workshop or delivery engagement is scoped and quoted separately.
How I approach the work.
01
Identify the decision
Describe the user’s task and the business outcome before choosing a model. Separate a promising demonstration from a workflow the team can operate.
02
Review readiness
Check whether the required information is available, usable and appropriate for the proposed workflow. Record gaps in permissions, system access and team ownership.
03
Define an evaluation
Choose representative tasks, a comparison baseline and criteria for useful and unacceptable results. Include missing information, incorrect answers and situations that need human review.
04
Plan the pilot
Set a limited scope, dependencies, review points and a decision at the end: continue, revise or stop. Make the reasons for that decision reviewable.
Related project evidence.
These cases describe project scope and personal contributions. Each links to a public product, profile or source record.
InvestingStack
Sep 2026 – present
A curated ecosystem directory, source-linked research and reusable datasets.
Read the case study →TradingAgents × Investboard
Sep 2026
An open-source connector that supplies portfolio context and records completed analysis.
Read the case study →Investboard
2025 – present
Bringing portfolio data, a personal mandate and decision history into one product.
Read the case study →David Bartas is the founder of Investboard and creator of InvestingStack. His own products and ecosystem work are identified in the case studies.
Common questions.
What should an AI pilot brief include?
A pilot brief should name the user and workflow, the problem to solve, the comparison baseline, the information the system can access, evaluation criteria, human review points, ownership and the decision to make at the end. A useful brief also records what is outside the pilot’s scope.
Do we need to choose a model before the consultation?
No. Start with the workflow, constraints and useful outcomes. Model and integration options can be compared once the task and evaluation are clear. If you already have a prototype, bring a few representative examples and its known limitations.