Built by an evaluator who understands the tools.
Thomas Crossman has a background spanning private-sector data analytics, humanitarian MEL, and security-sector programming. He has led and contributed to evaluations across fragile and conflict-affected settings for clients including Mercy Corps, UNICEF, IOM, ACTED, UNMAS, and the British Council, working against donor frameworks from FCDO, AFD, BHA, ECHO, UNTFHS, JICA, KOICA, and GAC.
The frustration that led to Meridian was straightforward: the same analytical bottlenecks on every assignment (manual document review, manual coding, manual drafting), regardless of the question or the deadline. The tools to compress that work existed, but the sector wasn't using them. Meridian was built to change that, bringing the data engineering and AI tooling side of evaluation practice to the front, not as an afterthought bolted onto traditional workflows.
Prior to evaluation, Thomas worked as an analytical lead in the private sector, building teams and analytics infrastructure from scratch. That technical foundation now underpins the AI-augmented workflows Meridian delivers.
He's very sharp, unusually good at making sense of messy, fast-moving contexts, and fundamentally honest about what the evidence really shows, which is rarer than it should be.
Despite the complexity of the program, he was able to quickly grasp its broader dynamics, operational gaps, priority needs, and diverse achievements. This allowed him to deliver an evaluation of high technical quality that was both insightful and relevant to our context.
[Tom's] work was thoughtful, thorough, and actionable, and resulted in insights that have supported strong learnings for our programming. Thomas has a deep and nuanced view of the humanitarian sector, and it shows in his work.
The risk in this sector isn't that evaluators adopt AI. It's that they adopt it uncritically, outsourcing interpretation to a tool that has no stake in being right, and no obligation to the people the programme is supposed to serve.
This practice is built on the opposite premise: AI as leverage for the evaluator's judgement, not a substitute for it. Transparency about tooling. Refusal of extractive data practices. Active scepticism about anything a model produces that can't be traced back to source. And a working assumption that the people whose lives are being evaluated deserve evidence that holds up under scrutiny, not evidence that arrived faster.
Setting the standard, not waiting for one.
There is no sector-wide standard for how AI should be used in evaluation. No agreed disclosure framework, no minimum transparency requirements, no shared understanding of where machine output ends and evaluator judgement begins. In the absence of that standard, most practitioners are using these tools quietly and hoping nobody asks.
Meridian's position is the opposite: document everything, disclose by default, and design outputs so that AI-generated content is identifiable and auditable. Not because regulation requires it (it doesn't, yet) but because the people whose lives are being evaluated deserve evidence they can trust. We are publishing our own transparency protocol because we believe it should become the norm, and someone has to go first.
Based in Nairobi. Working globally.
Active across East Africa, the Horn, the Sahel, West Africa, the Great Lakes, MENA, and Southeast Asia. Working language: English. Experienced field partners for French, Arabic, Swahili, Amharic, and Somali assignments.
Credentials & clearances.
- — MSc Analytics
- — BPSS Security Clearance
- — HEFAT trained
- — UNICEF Safeguarding (PSEA, Child Safety)
- — Professional Certified Data Analyst