Data Governance Has Been Smacked in the Face by AI
Eight truths from Experian’s Data Governance Masterclass panel
Experian recently brought together four practitioners who have spent decades trying to make data governance stick inside large organisations. Andrew Andrews from EDMA. Kylie Kirkby from Bendigo and Adelaide Bank. Tim Moon from DGX. Belinda Toniolo from RACV. The panel was unusually candid about what is working, what is broken, and where data governance has to go next. Eight insights stood out.
1. AI has smacked data governance in the face
That was Andrew Andrews’s framing, word for word. AI is the biggest opportunity data governance has ever had. It is also exposing every crack in the foundation. AI proof of concept failure rates are high. Not because the AI is bad, but because the metadata is incomplete, the data quality is poor, the catalogues are empty, and nobody owns the business processes feeding the models.
The same pattern shows up across industries. A team gets excited about a use case, runs the pilot, then discovers the underlying data has no defined owner, no documented lineage, and no quality benchmark. The pilot stalls. Trust evaporates.
2. Stop policing, start advocating
Belinda reframed ownership as advocacy. Her example landed. Her husband works on a Cadbury factory floor where IoT sensors track caramel temperature and line throughput. Frontline workers engage with data quality the moment it is tied to their actual productivity. Not compliance. Not policy. Results.
The takeaway: embed data obligations into position descriptions matched to role level. Broad KPIs are too blunt. A call centre agent and an executive general manager have completely different obligations to the data.
3. Define what you will die on a hill over
Belinda also described the shift from a four-month RFP process to a two-week test-and-learn cycle. That speed forces a decision most governance teams avoid making. What actually matters? What can you live without?
Govern everything with maximum rigor and you become the black hat. Excluded from conversations. Blamed for slowing delivery. Minimum viable governance is not laziness. It is survival. In practice that means picking three or four non-negotiables (PII handling, lineage on regulated reports, a clear escalation path when data quality breaks) and letting everything else operate lighter until it earns more.
4. Stewards without a garden are just decoration
DGX’s Tim Moon shared a story that captures the steward problem in a single image. He worked with a client that had 25 data stewards. The roster was cut to 12. The other thirteen had no real domain to steward. His metaphor: a gardener weeds, fertilises, prunes. If there is no garden, there is no gardener.
Stewardship bolted onto a marketing manager’s day job, with no defined data domain and no time allocated, is theatre. If an organisation cannot point to the domain a steward is responsible for and the time they have to govern it, they are not a steward. They are a name on a spreadsheet.
5. Influence is the skill nobody is building
Tim was direct. If a data governance leader is more than two levels from the C-suite, they may as well pack up and go home. Influence was named the single biggest skill missing from data governance leaders at last year’s industry conference. The problem is structural. When a message has to pass through three layers of management before it reaches a decision-maker, it arrives diluted, reframed, or dead.
Proximity to power is not a perk. It is a prerequisite.
For leaders without that proximity, the workaround is the same as it has always been: find a sponsor who has it, and feed them what they need to make the case.
6. Be a therapist, not a salesperson
Andrew offered a memorable framing. Data governance leaders need to be therapists. Get into the executive’s head. Understand what is keeping them up at night. Then turn that stress into a data lens.
Tim pushed it further. If you have to convince someone that data governance matters, forget about it. Let go of the tug of war, and they often tell you the real problem themselves. The organisations making progress are not the ones with the best pitch deck. They are the ones asking better questions.
7. Trust comes from visibility, not perfection
Kylie Kirkby has built her data governance function from scratch over twenty years at Bendigo and Adelaide Bank, so when she talks about trust, it is worth listening. Organisations do not need gold-standard data across the board. They need people to know which data sets are gold and which are bronze, and to use them accordingly.
Andrew reinforced it. Business culture is king, and data culture has to mirror it. The fastest way to destroy trust in a data team is to be caught polishing a number that was never reliable in the first place.
8. How to start
Three things any organisation can do this week.
- Map your top AI use cases back to their data foundations. Document the metadata, quality, ownership, and lineage situation honestly. Most teams skip this step.
- Pick one business problem an executive already owns and reframe governance work as the solution to that problem. Same initiative, not a separate one.
- Audit your stewards. If someone has no defined data domain, no time allocated, and no real accountability, please either assign one or remove the title.
The throughline across the panel was the same. Data governance fails when it lives in a silo, speaks in jargon, and waits to be invited. The organisations closing the gap are treating governance as a change-management problem first and a technical problem second.
———
Wrestling with these challenges in your organisation? DGX helps regulated enterprises rebuild their data foundations, design steward operating models that hold up under pressure, and equip data governance leaders with the influence they need at the executive table. Get in touch to talk through what good looks like in your environment.
