Ask a financial services firm who owns its AI strategy and you’ll usually get an answer, even if it’s a contested one. Ask who owns its data strategy and the answer is normally the Chief Data Officer. Records retention belongs to records management or compliance. The systems belong to IT.
Now ask who owns the job of making thirty years of historical information usable by AI. In my experience, that’s the question that produces a pause.
Everyone owns a piece of the past. Almost nobody owns the whole of it.
The present is getting an owner
The ownership of AI and data is, at least, being worked out. A Gartner survey of 504 data and analytics leaders found that 70% of chief data and analytics officers now have primary responsibility for building their organisation’s AI strategy and operating model[1]. Not every firm is there yet. Cynozure’s 2026 survey of senior data leaders found that 80% of organisations assign data strategy to a CDO or Head of Data, but only 28% do the same for AI, while 17% report no clear AI owner at all[2]. That’s a small sample of 60, but the direction is clear enough.
What’s striking is that the debate is almost entirely about current data and future AI. Historical information – the claims files, policy wordings, advice records, correspondence and decisions accumulated over decades – rarely features in it at all.
Everyone owns a piece
That isn’t because nobody is responsible for historical information. It’s because several people are, each for a different part of it.
IT owns the legacy systems and the repositories where old content lives. Its objectives are usually stability, security and cost, including the cost of eventually decommissioning those systems.
Records management or compliance owns retention: What must be kept, for how long, and when it can be destroyed. Its measure of success is defensibility.
The data team owns data strategy, but in most firms that strategy centres on structured, operational data: Platforms, pipelines, warehouses and analytics. Decades of unstructured documents tend to sit outside it.
The business owns the knowledge. The claims, underwriting, advice and legal teams understand what the records mean, but curating them isn’t their job.
And in many firms, the offsite paper archive is managed as a storage contract by facilities or procurement, where its main attribute is what it costs each year.
Each of these functions is doing its job properly. The gap is between them. Gartner’s own research with CDAOs identifies the intersection of IT and data governance as a challenge, particularly in clarifying ownership and collaboration[3]. For historical information, the overlap is wider still.
Why the gap matters now
Before AI, having nobody responsible for historical information as a whole carried a limited cost. Records were retained to meet obligations and retrieved occasionally by people who knew how to find them. Their value was mostly defensive.
AI changes that, because it raises decisions that cut across every one of those functions. Which historical information should AI be able to use? Which should be kept but ring-fenced? What should be destroyed before it’s ever connected to anything? What’s worth digitising, and to what standard? Who decides which version of a policy or procedure is current?
Each of those decisions involves at least two of the owners above, and often all of them. Without someone with authority across them, the decisions don’t get made. They default. Usually they default to whoever is building the AI tool, who connects whatever is easiest to reach. That’s how historical information ends up in AI systems by accident, or is left out of them entirely.
A hypothetical: The migration that did everything right
Here’s a hypothetical. An insurer decommissions a legacy claims platform to reduce running costs. IT runs the migration. Open claims and the last seven years of closed claims move to the new system. Everything older is exported to a low-cost archive store as document images with minimal indexing, which fully meets the firm’s retention requirements. Records management signs off. The platform is switched off and the saving is booked.
Eighteen months later, the AI team starts building an assistant to support long-tail liability claims. It needs exactly the older claims history that went to the archive. But in the export, the documents were separated from the structured fields that gave them meaning: Claim type, cause, outcome and how each case was resolved. Reconnecting them now is slow and expensive. During the migration, keeping that context would have cost relatively little.
IT did its job. Records management did its job. Nobody was asked whether the past might be worth more than the cost of storing it.
A useful regulatory test
UK financial services already has a framework built around exactly this kind of question. The Senior Managers and Certification Regime is designed to make it clear who is responsible for what in a firm, and each Senior Manager has a Statement of Responsibilities[4]. It’s a useful test to look at your own responsibilities map and ask where historical information appears. It will probably feature under record-keeping and compliance. It’s much less likely to appear anywhere as an asset someone is responsible for making usable.
What an owner actually does
This doesn’t necessarily mean a new role or a new hire. It means one named person who is accountable for the firm’s historical data strategy and has the authority to work across the functions that each hold a piece of it.
In practice, that person would be responsible for:
- Knowing what the firm holds: Across legacy systems, current repositories and physical archives, in enough detail to make decisions about it.
- Deciding what happens to it: Use case by use case, what is made AI-ready, what is kept but ring-fenced, and what is destroyed.
- Having a voice before value is lost: Being consulted before legacy systems are decommissioned, archives are migrated or storage contracts are retendered.
- Setting standards for context: So that AI can tell what is current, what has been superseded and where each record came from.
- Connecting the past to the AI plan: Making sure whoever owns AI strategy knows what historical information exists and what it would take to use it.
Where that responsibility sits will vary. In many firms, the natural home is an extended remit for the CDO, bringing historical and unstructured information into the data strategy. In others, it may sit with the COO or the head of information governance. Who it is matters less than the fact that it’s someone, and that they’re not simply asked to hold the responsibility without the authority to act on it.
How Dajon approaches it
At Dajon, we’ve spent nearly thirty years working with every one of the functions that owns a piece of the past: IT teams on migrations and legacy content, records managers on retention and destruction, facilities and procurement teams on storage, and business teams on retrieving what they need. That gives us a clear view of where the gaps between them lie.
We help firms bring those pieces together: Building a single view of their historical information across physical and digital estates, applying retention decisions consistently, and creating a practical route to making the valuable parts usable for AI. That gives whoever owns the firm’s historical data strategy something concrete to own.
Your organisation already owns its past. The question is whether anyone is responsible for it.
Can you really have an AI strategy without a historical data strategy?
References
- Gartner Survey Finds 70% of CDAOs Are Responsible for AI Strategy and Operating Model Gartner[↩]
- AI Adoption Surges, but Only 15% of Organisations Can Prove Financial Impact Business Wire[↩]
- Top 3 Priorities for CDAOs in 2025 Evanta, a Gartner company[↩]
- The Senior Managers and Certification Regime FCA[↩]