A data leader asked a question this week that stayed with me: we've spent years talking about data as an asset. What does it cost when data becomes a liability?
He named three costs, and every one of them is familiar to anyone who has run an insurance operation.
The first is the single point of failure. A critical process depends on one person who knows how everything really works. We call that experience. Until they leave — or become so overutilized that every important decision has to flow through them. Then we discover it was institutional knowledge with a single point of failure. In a carrier, that's the underwriter who knows how the rating really works, the claims handler who knows which report to trust, the person who knows why the bordereaux never quite reconciles. The knowledge is real. The fact that it lives in one head is the liability.
The second is the storage tax. We keep years of data because we might need it someday, and keep paying to store it, secure it, migrate it, and manage it. Lifecycle FOMO has a price tag — his phrase, and a good one. Most of what a carrier stores has never informed a decision. It is kept because deleting it feels riskier than paying for it. That is a strange way to run a balance sheet: paying rent on an asset nobody uses, in a building nobody visits.
The third is the swivel tax. People spend their days moving between systems, spreadsheets, and reports, piecing together an answer the technology was supposed to give them. None of it shows up on a dashboard as data debt. It just quietly makes the business harder and more expensive to run — in the hours of your best people, spent stitching.
None of this is new. What is new is what AI does to it.
What AI does to it
An agent does not clear data debt. It reprices it. Give an agent a messy estate and it will retrieve the mess faster, reconcile it more confidently, and present it with more polish. The confident wrong answer arrives sooner. That is why the accountability question matters more than the model question: can you say which document, which version, which person stands behind the answer? A ledger needs lineage. Data debt is what makes lineage expensive to build — and the agent expensive to trust.
This is also where the business case flips. We usually try to prove what better data could unlock. Sometimes the stronger case is simpler: understand what the debt is already costing you to live with. That cost is already being paid. It is in the hours, in the single points of failure, in the storage bill. You do not need a projection to find it. You need to look at what your people already do every day.
From the operator's chair
Name the single point of failure. For every critical process, write down the one person whose departure breaks it. If there is a name, there is debt, and it has a face.
Price the swivel. Count the hours your best people spend stitching answers across systems. That is the debt's interest payment, and it is due every week.
Stop paying the storage tax. If no decision has needed a data set in two years, archive it and stop securing it like it is live. Deleting is a decision too — and usually the cheaper one.
Give every critical data set one named owner. Data with no owner has no one to ask, no one to blame, and no one to fix it. A named owner is the cheapest control in the book.
Data is still an asset. But an asset with no owner, no lineage, and a single point of failure is a liability wearing an asset's name. The ledger knows the difference. So should we.
The line-by-line version of this piece — the three taxes in every line of business, from SOVs to bordereaux — is published alongside it: Data Debt, Line by Line.
— Krishna