Track 07 · Data Academy

AI Governance

Most AI systems do not fail technically. They fail at the point where somebody has to sign off an answer they cannot trace, or explain a decision they cannot reproduce. This track is about the controls that get an AI system out of the pilot and keep it there.

Read time ~14 min Covers 8 topics and 6 frameworks For data, risk and engineering leads

01 The syllabus

What this track covers

Eight topics, in the order they have to be done. Each one is a capability rather than a document, and each depends on the one before it.

01

The use-case register

A single inventory of every AI system in use, including the ones bought inside other software. You cannot govern a portfolio you have not enumerated.

Shadow AI is the default state
02

Risk classification

Tiering each use case by the consequence of it being wrong, so oversight is proportionate rather than uniform — the structure every modern AI regime shares.

Consequence, not sophistication
03

Data provenance

Knowing what the system was trained on, or what it retrieves from, with the same lineage discipline applied to any reported number.

An agent inherits every unverified hop
04

Evaluation

A test suite with expected answers, run on every change, covering the failure modes as well as the happy path. Without it, "it seems better" is the release criterion.

Evals are regression tests for behaviour
05

Human oversight

Deciding where a person must review, approve or be able to intervene, and making that possible in the interface rather than in a policy document.

Oversight has to be actionable to be real
06

Transparency and disclosure

Telling people when they are interacting with an AI system, and being able to explain what informed a decision that affected them.

Explanation is a data problem
07

Monitoring and drift

Watching quality, cost and behaviour in production, because the inputs change even when the model does not.

The world drifts; the model does not know
08

Incident response and rollback

A defined path for when an AI system is wrong at scale: how it is detected, who can stop it, and how you revert.

Deployability requires reversibility

02 The frameworks

The frameworks, and what each asks you to be able to show

These converge more than they differ. All of them assume you can enumerate your AI systems, justify their inputs, evidence their testing and demonstrate oversight.

FrameworkWhat it is concerned withWhat you must be able to show
EU AI ActA risk-tiered regime for AI placed on the EU market, with the heaviest obligations on high-risk usesClassification of each use case; data governance over training and input data; technical documentation; logging; human oversight; accuracy and robustness appropriate to the tier
NIST AI Risk Management FrameworkA voluntary US framework structured as Govern, Map, Measure and ManageA mapped inventory of AI risks per use case, measured with defined metrics and managed with named owners
ISO/IEC 42001A management-system standard for AI, in the shape of ISO 27001Documented objectives, controls and evidence that the AI management system operates as described
Model risk management (financial services)Long-standing supervisory expectations for models used in decisionsIndependent validation, effective challenge, documented limitations and ongoing performance monitoring
GDPR, on automated decisionsDecisions producing legal or similarly significant effects, made solely by automated meansA lawful basis, meaningful information about the logic involved, and a route to human review
Sector regimes (medical, safety-critical)AI embedded in regulated products and clinical or safety decisionsEvidence under the existing product regime, which generally exceeds anything the AI-specific regimes require

Summarised at the level of the obligation each framework creates for data. For the binding text, and for how it applies to you, take advice — the requirements are stable, the detail and the timelines are not.

03 The controls

The controls that make an AI system deployable

Each of these is a data capability before it is an AI one, which is why organisations with governed data ship AI faster and not more slowly.

ControlWhat it doesWhat it depends on
Use-case registerEnumerates every AI system, its owner and its risk tierA process that catches AI bought inside other software
Input governanceControls what the model may be trained on or retrieve fromClassification, entitlement and purpose binding on the underlying data
Retrieval lineageShows which records produced a given answerLineage captured through the retrieval layer, not just the pipelines
Evaluation suiteTests behaviour on known cases before and after every changeA governed set of questions with agreed correct answers
GuardrailsBlocks actions and disclosures outside what is permittedPolicy expressed as data the runtime can evaluate
Action loggingRecords what an agent did, with what inputs, on whose behalfImmutable logging at the tool and data boundary
Human-in-the-loopRequires approval for defined categories of actionRisk tiering, so review is applied where consequence justifies it
RollbackReverts to a previous version quickly and completelyVersioned prompts, models, retrieval sets and configuration

04 Sequence

From pilot to production without a governance programme first

Governance here is not a gate you build before starting. It is the set of things that make the second use case cheaper than the first.

0101Register the use cases you already have, including the AI inside tools you bought. The list is always longer than expected.
0202Tier them by consequence. Most will be low-risk and need almost nothing; the few that are not deserve the effort.
0303Govern the inputs for the first real use case: what it may read, under what basis, with lineage captured.
0404Build the evaluation set with the business, not the data team. Agreed correct answers are the whole asset.
0505Ship with logging, guardrails and a rollback path. Reversibility is what makes approval possible.
0606Monitor quality and cost in production, and re-run the evals on every change to model, prompt or retrieval set.

05 What goes wrong

The four failures worth naming

Failure

Pilot purgatory

A working demonstration that never deploys, because nobody can approve an answer they cannot trace. The blocker is almost always governance of the inputs, not model quality.

Failure

An ungoverned retrieval corpus

Documents of unknown provenance, permission and currency. The model faithfully reports what the corpus says, including the parts that are three years out of date.

Failure

Evaluation on the happy path only

Tests that confirm what already works. The failure modes that matter — ambiguity, missing data, adversarial input — are the ones not covered.

Failure

No owner for the whole system

The model has a data scientist, the prompt has an engineer, the retrieval set has nobody, and the behaviour is the product of all three.

06 Diagnostics

Six questions to ask your own team

If two of these cannot be answered without a manual investigation, those two are the plan for the next quarter.

  • Can you list every AI system in use, including the ones inside software you bought?
  • For your highest-consequence use case, can you show what data informed a specific answer?
  • Do you have an evaluation set with agreed correct answers, and does it run on every change?
  • Who can stop an AI system in production, and how quickly?
  • Is the retrieval corpus governed — owned, permissioned, dated — or is it a folder?
  • Would you be able to explain a decision this system made, to the person it affected, six months from now?

08 Questions

Frequently asked

Does governance slow AI deployment down?

The opposite, after the first use case. What blocks deployment is almost never model quality — it is that nobody will sign off an answer whose inputs cannot be traced. Governed inputs and an evaluation set are what make approval a decision rather than an argument.

What is the minimum viable AI governance?

A register of use cases, a risk tier for each, governed inputs and an evaluation set for anything above the lowest tier, and a rollback path. That is proportionate for most organisations and defensible under every framework in the table above.

How does the EU AI Act change what we build?

Mostly it changes what you must be able to evidence, and only for higher-risk uses. The engineering consequence is documentation, logging, data governance over inputs and demonstrable human oversight — all of which are good practice regardless of jurisdiction.

Who should own AI governance?

The same function that owns data governance, with risk and legal as partners rather than owners. Splitting them produces two registers, two policies and two sets of lineage, none of which agree.

See these controls running

Classification, entitlement, lineage and policy applied to the data itself rather than to each report.

See Data Governance