What started as a training engagement has become something more like embedded consulting. Hundreds of employees across both organizations, from frontline teams all the way to executive leadership, trained through single and multi-day sessions built entirely around their existing workflows.

The sessions didn't just teach people how to use AI tools. They uncovered where automation actually made sense, and where it didn't. Out of the coursework itself, teams began building agents that are projected to save 2,000 to 3,000 hours annually. But the bigger outcome was what nobody planned for: new value-adds beyond just automation, places where AI could improve decision quality, not just speed.

Forty-plus senior executives needed something that most AI training doesn't offer: an honest conversation about when to invest in AI and when the answer is no. The engagement was built around strategic decision-making, not tool demos. We focused on helping leadership develop the judgment to evaluate AI investments against real operational needs, cutting through vendor narratives to get to what actually matters.

Executives left with a practical operating framework for AI investment decisions, including risk thresholds, evaluation criteria, and ownership models across business and technology teams.

Working with Armetour's defense and intelligence teams to integrate AI capabilities into operational workflows. This engagement draws on the same zero-defect thinking that comes from working on systems where failure isn't an abstraction. When the stakes are this high, computational orchestration matters: knowing which layer each problem belongs on, whether that's AI, traditional code, human judgment, or a decision not to build it at all.

VigilOre is the system that came out of this work: a multi-agent compliance platform for DRC mining operations. Four specialized agents handle document intake, aggregation, comparison, and reporting across multilingual regulatory frameworks. The entire product, from architecture to the demo video, was designed and produced by Eduba.

An AI ethics and governance briefing for ambassadors and international affairs experts, focused on how AI tools are reshaping soft power and influence measurement across 25+ countries. The workshop brought together researchers from the British Council, the British Foreign Policy Group, Ben Gurion University, and other institutions to discuss what happens when the tools we use to measure influence start changing the influence itself.

Presented the Ethics Engine research alongside ongoing work with ICR and the British Council on international soft power, connecting the psychometric evaluation of AI models to the broader question of how governments and institutions should think about the tools they're adopting.

Training delivered to IAG's innovation division, helping one of the world's largest airline groups think through where AI fits within development, thought leadership, and customer experience operations. The kind of environment where getting it wrong isn't just expensive, it's felt by millions of passengers.

"A pleasure working with Jake and his team at Eduba." — Lissa Pritchard, Head of Innovation: Development, Thought Leadership & Customer Experience, IAG

Most organizations choose AI tools based on marketing. We built a methodology to evaluate them using validated psychological instruments — the same psychometric frameworks applied to humans, now applied at scale to large language models. The Ethics Engine gives organizations something they've never had before: an empirical, repeatable way to assess the behavioral tendencies of the models they're trusting with their work.

Being able to evaluate these models at all was the hard part. Doing it at scale is what makes it useful. This research was also presented at UKICER 2025 under "AI Workflows: Honest, Transparent, Reusable Work" and formed the basis of the AI governance briefing at the Academy of International Affairs NRW.

Applying machine learning and NLP to map how ideas transform as they cross borders, for British Council-commissioned research on international soft power. The work uses AI to do what would take a research team months: trace influence patterns across languages, institutions, and policy frameworks.

The engagement went well enough that ICR made a decision they hadn't made in over a decade: they wanted to produce a hard copy of the study. That's not a metric you find on a dashboard, but it tells you something about the quality of the work. When a research organization breaks a ten-year pattern to put your findings in print, the work did its job.

Before the Fortune 500 engagements, before the global consultancies, the work started here. After using AI to produce research that was accepted to Flagler's Saints Academic Review, faculty started asking how it was done. That turned into contracted AI workshops for the college, teaching professors how to think about these tools in the classroom.

It was the first time the tables turned from student to teacher, and it proved something that still holds: the people most skeptical of AI often become its most thoughtful adopters, if someone takes the time to connect it to their actual work. The work continued at the University of Edinburgh, where Eduba Labs ran experimental workshops at the Edinburgh Futures Institute for faculty exploring AI in education.