Ryan Hanley joined the AI InterConnect podcast for an episode on deterministic AI and enterprise governance. The conversation starts with the problem that sent him into this space: regulated organisations spending heavily on compliance that produces documents rather than decisions, while the facts those decisions depend on sit scattered across email, spreadsheets and the heads of the people who have been there longest.
From there the discussion moves through the choices behind Taskd. Why vector search is the wrong tool once an answer has to be defended to a regulator. Why a structured graph, built once from the source records, lets every subsequent question be answered by walking the data rather than asking a model to guess. And what that looks like in practice, using the public-records demonstration Taskd built for the Government of Alberta-hosted hackathon, where rules over a graph of charities, grants, contracts and directorships flagged $861 million of related-party funding flows for expert review.
In the episode
- 0:00 The path to deterministic AI
- 0:50 Ryan’s background in blockchain, crypto and compliance
- 3:30 The problem: regulatory theatre and messy data
- 5:30 Defining Taskd
- 6:45 The limit of vector databases
- 8:30 Knowledge graphs and the advantage for professional-services firms
- 10:20 Identifying $861M in Alberta’s public records
- 14:15 Adoption curves and the coming wave of agents
BoundaryThe Alberta figures discussed in the episode come from a public-records demonstration, not a customer deployment. The flags it produced were prompts for expert review, not findings of wrongdoing.
AI InterConnect publishes conversations with founders and operators on automation and the future of business. The episode was released on 4 June 2026.