AI Advisor · Dependable AI
Ezzard Batiste
How much authority AI gets. Where controls sit. What the evidence supports.
I bring 25+ years of building software for financial markets, and a combined technical and business perspective, to the evaluation and adoption of AI systems and agents. My current focus is dependable AI: authority boundaries, independently enforced controls, and evidence-based evaluation.
Capability is not dependability
As AI systems move from generating answers to taking actions, what they can do becomes inseparable from what they should be allowed to do.
Dependability means understanding the authority a system can actually exercise (its effective permissions through tools and connected credentials, not only what it has been told), setting boundaries around consequential actions, and testing whether those boundaries hold under realistic conditions.
Model instructions are part of that architecture. But for constraints that matter, a model's own instruction-following shouldn't be the only safeguard. Independently enforced controls, and evidence that they work, are what justify giving a system more responsibility.
The questions I work on
- How much authority should an AI system have?
- Which constraints matter too much to leave to the model alone?
- When should a person have to intervene?
- How do we test whether those controls actually work?
- What does the resulting evidence justify?
Where I focus: evaluating AI systems and agents · authority boundaries and human approval points · controls beyond model instruction-following · failure analysis and regression testing · AI strategy and executive education
Background
For 25+ years, I've built software for financial markets, including market data and trading infrastructure. In those systems, failures have real consequences, and one principle held: no data is better than bad data. I now apply that same systems discipline to AI.
- Systems experience: low-latency C++/Linux, market data, trading and digital-asset systems, with roles at Intercontinental Exchange, Interactive Data, BGC Partners and Galaxy Digital
- Education: Computer Science, Harvard University; MBA, Yale School of Management
- Hands-on AI work: working with current AI systems and multi-step workflows, examining where they fail, how approved state gets lost or overridden, and what stronger controls and evaluation might look like
The guide
Dependable AI: A Practical Field Guide (Edition 1)
A 30-page practical guide to risk, controls and evidence for AI systems that can act, not merely answer. It covers authority boundaries, effective authority, controls outside the model, evaluation and operations, with a worked example and a lens from electronic trading. Free, no sign-up.
Writing
I write on LinkedIn about dependable AI, especially the relationship between capability, authority, control, evaluation and evidence.
Get in touch
If you're evaluating an AI system, considering the use of agents, or working through questions of authority, control and oversight, I'd welcome a conversation. I'm available for AI advisory, technical evaluation and education, and interested in exchanging perspectives with others working on these challenges.