What must remain true from an investment mandate to a reconciled outcome?
A prompt is not a mandate
Latest writing
A prompt is not a mandate
Designing the mandate-to-outcome control plane for agent-initiated investing.
Read the essayAI engineering · Product systems · London
Sohaib Benamor
I write about how intelligent decisions become reliable system outcomes.
AI engineer with a product and backend focus, writing about agents, financial infrastructure, optimisation, and the software that makes them dependable.
- Question
- What must remain true from an investment mandate to a reconciled outcome?
- Claim
- A prompt is not a mandate.
- Artifact
- 18 minutes · 3 figures · 18 sources
- Status
- Published · 12 Jul 2026
QQuestion
SSource
HHypothesis
DDecision
EEvidence
Working questions
The problems I keep returning to.
- 01
How should an agent receive authority without inheriting unlimited trust?
- 02
How does a system prove that reality still matches the original intent?
- 03
Where should probabilistic reasoning end and deterministic control begin?
Current work and research
What I am doing, and what has been reviewed.
- Current practiceLendable · 2026
Agentic workflows and data integrity
Working on internal AI workflows and large-scale reconciliation, with correctness, lineage, and operational clarity as the real constraints.
- ResearchJORS · Under review
Teaching Assistant Allocation at Scale
First and corresponding author on a MILP study spanning formulation, implementation, experiments, analysis, and writing.
About
I care about the handoff from a good decision to a system people can trust.
I am an AI & Infrastructure Analyst in London and an MSc AI graduate from King's College London. My work connects AI systems, reconciliation, optimisation, and backend engineering.
Read the full profile