Information metabolism
Data goes in. Decisions come out.
Most organisations ingest enormous quantities of information and convert almost none of it into action. Infobolism is the practice of fixing that conversion — governed data foundations, real synthesis, and AI implemented end to end rather than bolted on.
in·fo·bol·ism /ˈɪnfoʊˈbɒlɪzəm/ · noun · the conversion of raw information into usable intelligence
Take it in
Pipelines, lakehouses and zero-copy federation. Getting to the data is rarely the hard part — getting to data you can defend is.
Make it mean something
Semantic layers, certified metrics, lineage and ownership. The step almost everyone skips, and the reason their dashboards disagree.
Spend the energy
Agentic analytics and decision intelligence — insight delivered into the workflow, with audit trails intact.
The idea
Why metabolism is the right metaphor.
Metabolism isn't storage. It's conversion — the set of processes that turn what you consume into energy you can actually spend.
Enterprises have spent fifteen years optimising ingestion. Warehouses got cheaper, pipelines got faster, lakes got bigger. Consumption went up and almost nothing changed about the rate at which decisions improved.
That's a metabolic problem, not a storage problem. The bottleneck sits in the middle — in synthesis. In whether anyone agrees what a metric means, who owns it, where it came from, and whether it can be defended in front of a regulator.
AI has made this urgent rather than academic. An agent acting on unsynthesised data does not fail politely. It fails confidently, at scale, with a plausible explanation attached.
So the sequence matters. Govern before you automate. Define the semantics before you let a model speak on your behalf. Establish lineage before you ask anyone to trust an answer.
Do that, and the agentic layer becomes genuinely useful. Skip it, and you have automated your existing problems and given them a confident voice.
Solutions
What we build.
Four engagements, each scoped to stand alone. Start anywhere — though the order below is the one that tends to work.
Governed data foundations
The substrate everything else depends on. Ownership, stewardship, quality rules, lineage and classification — wired into the platform rather than written into a policy nobody opens.
- Target-state data architecture
- Catalogue, glossary and lineage rollout
- Policy-as-code and active metadata
- Regulatory and audit readiness
Synthesis & semantic layers
One definition of every number that matters. Certified metrics, a shared semantic model and reporting people stop arguing with — the layer that turns storage into meaning.
- Metric standardisation and certification
- Semantic modelling for BI and agents
- Governed self-service enablement
- Legacy estate migration and rationalisation
Full AI implementation
Agents that answer from your data, not from a general model's imagination. Grounded, evaluated and shipped into the workflow — with the guardrails specified before the demo, not after.
- Agent design: topics, actions, guardrails
- Retrieval and grounding over governed data
- Evaluation harnesses and regression testing
- Human-in-the-loop and escalation design
AI governance & assurance
Govern the AI before using AI to govern anything. Model risk, transparency, bias, third-party exposure and the evidence trail an auditor will ask for.
- Responsible-AI control framework
- Model inventory and risk tiering
- Explainability and audit evidence
- Regulatory alignment and review gates
Writing
Notes from the middle layer.
Technical writing on governance, synthesis and applied AI — opinionated, and grounded in work that actually shipped.
Governance that executes itself
The shift from documented governance to active, metadata-driven governance: data discovered automatically, sensitive fields classified in real time, lineage tracked continuously, audit evidence generated inside the workflow.
An agent is only as good as the data beneath it
Why grounding is a data-governance problem wearing an AI costume — and what "trustworthy enough to act on" actually requires in practice.
Stop shipping dashboards. Start shipping decisions.
Infrastructure gravity, a semantic trust layer in the middle, agentic analytics on top. Why "which BI tool" is last decade's question and stack composition is this decade's.
Govern your AI before you use AI to govern data
AI can detect anomalies, classify, summarise lineage and draft audit evidence. What it cannot do is own accountability — and why that distinction decides your operating model.
The five pillars, and the one everybody skips
People, data, data stores, processes, governance. A working framework — and an honest account of why the people pillar is where programmes quietly die.
Grounding agents on a governed lakehouse
A build log: harmonisation and identity resolution feeding an agentic layer over warehouse data, and the guardrail patterns that survived contact with real users.
Posts are placeholders wired to the contact
section — swap the href on each card as articles go live.
Stack
Tools we actually run.
Cutting edge where it earns its place, boring where boring is correct.
Platforms & compute
Governance & metadata
Analytics & semantics
AI & agentic
Who's behind it
Written and built by one practitioner.
Start a project
What's stuck in your middle layer?
Whether it's a governance programme that stalled, dashboards nobody trusts, or an AI pilot that can't get past legal — a first conversation is thirty minutes and free.
[email protected]