Company knowledge becomes more useful when the connections survive beyond the task that uncovered them. Which records refer to the same organisation? Which rule applies? What evidence supports the answer? Keeping those connections gives the next person or AI workflow something to build on.
That is the direction behind Taskd, and the subject of co-founder and CTO Alex Dollery’s appearance on The Connected Startup, Neo4j’s podcast about the problems that lead founders to build with graphs. Alex joined host Brian O’Keefe for Episode 4, broadcast on 17 September 2026.
Watch the full episode on YouTube or listen on Spotify. The conversation runs for about 35 minutes.
Give AI the connections behind the facts
Alex explains how concept graphs represent meaning and relationships alongside the underlying records. Combined with rules, that structure can help a system follow connections, derive information and show the evidence behind an answer.
He illustrates the approach with Taskd’s Alberta public-records demonstration, then previews a research engine answering questions from news articles. In the Alberta example, connections between organisations and transactions produce flags for human review. These are prompts to investigate, not findings of wrongdoing. The public-records demonstration and research preview are separate from Taskd’s production deployment.
Build intelligence you keep
Taskd is building a maintained representation of company knowledge: facts, relationships, rules and reviewed outcomes connected to their sources. The aim is to make changing knowledge reliably usable by people, models and agents, with less repeated work to reconstruct the context.
As models and tools change, that knowledge should remain useful to the business. The episode offers a look at the technical choices behind that ambition, including how graph structure and smaller models can work together. The research engine is still in development.
Jump into the conversation
- 7:01: Concepts, relationships and rules
- 18:47: Why a flag needs human review
- 21:44: A preview of the research engine
Thank you to Brian and the Neo4j team for the conversation and their support through the Neo4j Startup Program.