AI Implementation
My AI build playbook, part 1: the model is the easy part
Part one of the playbook I use to build AI systems: the principles behind reliable outputs, and a stack you add to only when a build needs it.
Applied AI researcher · Business analyst · Scotland
Applied AI, evidence first.
I write about what actually works in AI, and what is only a good demo: evaluating agents, multi-agent systems, voice agents, and putting AI into real organisations.
Latest
All posts →Research
Co-authored work on multi-agent evaluation and causal reasoning
- 2026Published
Persona-Driven Feedback Agents for High-Stakes Documents: A Multi-Perspective Evaluation of Global Talent Visa Submissions (opens in new tab)
IEEE AAIML 2026 (International Conference on Advances in Artificial Intelligence and Machine Learning), pp. 367–372
Co-authored research prototype: persona-based AI agents assess high-stakes application documents. Evaluated on 20 synthetic applications; no real-world outcome validation.
- 2026Accepted · awaiting publication
Adversarial Causal Deliberation among Autonomous Agents for Reliable Consumer Behaviour Explanation
ICSRS 2026, Rome
Co-authored simulation study: agents with competing roles must challenge a causal claim before it is accepted. Accepted, not yet published.
- 2026Accepted · awaiting publication
CausalAgentBench: Reliability-Aware Benchmarking of Causal Reasoning, Coordination and Explainability in Multi-Agent Consumer Intelligence
ICSRS 2026, Rome
Co-authored benchmark that scores the handoffs between components of a causal analytics pipeline, including corrupted inputs and abstention. Accepted, not yet published.
What I write about
- Evaluating AI agents
- Multi-agent systems and causal reasoning
- Building voice agents: design, guardrails and failure modes
- AI implementation in real organisations, through the business-analysis lens
- UK and Scottish AI: policy, ecosystem and what it means in practice
About
I'm a business analyst, applied AI researcher and builder, based in Scotland. I care less about whether a system produces an impressive answer and more about whether anyone can show the answer is right.
More about me →