Data and knowledge
AI tools and agents are only as useful as what they can find and trust. In most organizations that means years of documents with no consistent metadata, AI-generated drafts piling up next to the real thing, and data spread across systems that were never meant to talk to each other.
These posts cover the foundations: metadata and content lifecycle, a data architecture for agents worked through for wealth management, and the same ideas applied to a Microsoft-centred organization. The post on duplicated solutions is listed here too, because reuse depends on being able to find what already exists.
4 posts, in suggested reading order
Metadata is what makes knowledge findable
Most organizations already have the answers and can't find them. A small schema, AI-suggested tags and rules for AI-generated clutter fix most of it.
6 min read · Updated September 2026
Why teams rebuild what already exists
Organizations pay for the same solution again and again because nobody can find the first one. AI agents will do it faster unless reuse gets easier.
From Experimentation and adoption · 5 min read · Updated September 2026
Data architecture for AI agents in wealth management
Wealth firms run on overnight batch files. Before agents can act on that data, it needs layers that are provable, reconciled and modelled for use.
6 min read · Updated September 2026
AI readiness on the Microsoft stack, without a migration
If you run on Microsoft 365 and Azure, you can make scattered knowledge usable by AI agents without moving it all. How I'd sequence Purview, Fabric and search.
6 min read · Updated September 2026