Insights

Track 3 is the only track that answers the audit question — and nobody is shipping it. Part 3 is the build: firing OpenLineage at the MCP boundary, modelling the agent as a new kind of lineage actor, and the two open RFCs already trying to standardise it.

Somewhere in your organisation today, an AI agent built and executed a data pipeline inside a chat session — and nothing logged the data flow. The industry is conflating AI governance and AI observability, exactly as we conflated data governance and data observability in the 2010s. I was in the room the first time. Same mistake, harder problem.

When an AI agent makes an MCP call to your data warehouse, you are on one of three instrumentation tracks. Track 1 captures nothing. Track 2 captures everything in OpenTelemetry — and still cannot answer the audit question. Track 3 is the one nobody is shipping.

Approximately 80% of enterprise data — emails, contracts, documents, AI knowledge bases — is invisible to most data management functions. For financial institutions, that is not a technical footnote. It is one of the most significant and least acknowledged risk exposures in modern financial services.

Most organisations' data management tools were built for structured data and cannot simply be extended to cover the 80% of the enterprise estate that is unstructured. This article sets out the strategic framework for closing that gap — and the evidence on why acting now delivers a measurable return.

Six articles. Six questions. One consistent answer. What the Consulting Has Changed series established, what it deliberately left unanswered, and what comes next — the Inverted Pyramid and Jana, BearingNode's AI consultant.

It didn't change on a single date. But there is a clear moment of inflection — and you can use it to test whether your consulting partner has adapted. Three forces converged. The old model stopped working. Here's the timeline.

AI hasn't just changed what consulting produces. It has changed who should be doing it — and who clients should hold accountable. Senior intelligence must now be present throughout delivery, not just at the pitch and the debrief. Here's what that means for every role in the delivery chain.

The centre of gravity in consulting has moved closer to the operating system of the business itself. The recommendation layer has been commoditised. Most firms haven't moved with it — because their economics require them to stay where the value used to be.

Traditional consulting scales by adding bodies. The AI era demands something different — observable outcomes, operational governance, and evidence by design. Seven questions to ask any consulting partner to find out whether their delivery model has genuinely changed, or just their pitch.

Traditional consulting is structurally incompatible with the AI era. That's not commentary — it's a design problem. The pyramid was built to solve a production constraint that no longer exists. Here's why the economics have broken, and why the firms most threatened are the ones currently selling AI transformation.

The economics of consulting have shifted. AI has commoditised the output layer — frameworks, slide decks, operating model templates. What clients now pay a premium for is what AI cannot reliably provide in regulated industries: accountable delivery, operational integration, and evidence.

A BearingNode series on the AI era transformation of management consulting. Six questions. One uncomfortable answer. Written for CDOs, CIOs, and senior leaders in regulated industries who are tired of transformation programmes that end with a governance framework no one owns.

Better decisions don't come from more dashboards, faster AI, or bigger datasets alone. Barry Green, Advisory Board Member at BearingNode and co-author of Data Means Business, argues that good decisions require whole-body thinking — integrating logic, experience, intuition, ethical judgement, and organisational context.

Traditional project management fails AI projects not because it's badly executed, but because it's designed for the wrong world. Learn why Mediocristan governance can't handle Extremistan uncertainty.

A common mistake in data conversations is to treat governance and observability as interchangeable. They are not. D/I O11y overlaps with governance, compliance, management and strategy — but it is distinct from all four.

Daniel Rolles participated in FIMA 2025 at The Kia Oval in London, sharing insights on the convergence of Data Governance and AI Governance with industry leaders from across Europe's financial services sector.

Gen AI projects reveal inadequacy of inherited frameworks. Learn why 95% of GenAI pilots fail not from technical risks, but from unvalidated assumptions about adoption, workflow fit, and business value—and how to navigate uncertainty rather than just manage risk.

Both F1 racing and modern data platforms rely on telemetry to prevent disasters. The difference? F1 cars cost £15 million and get 300 sensors. Your data estate? Worth billions, gets a quarterly audit.

It's encouraging to see open standards taking a prominent place at Big Data London. At BearingNode, we're invested in communities like OpenLineage and are thrilled to see the industry embracing open source solutions for next-generation data, analytics, and AI architecture.