Context Debt

The claim

Technical debt has a name, a rough shape everyone in software agrees on, and usually a line item somewhere in a sprint plan. Context debt is the same phenomenon, wearing different clothes, and almost nobody has named it yet. It's the accumulated backlog of decisions, exceptions, and tribal knowledge that a company never wrote down, because it never had to.

What accumulates

Every organisation has knowledge that was never written down. Policy exceptions agreed in a meeting and never added to the handbook. The reason one customer gets special treatment, known only to the two people who've managed the account for years. The step in a process everyone follows and nobody documented, because whoever set it up thought it was obvious. None of this comes from a lack of discipline. It's what happens when knowledge lives in people rather than in writing, which is the normal state of most companies.

Why it was invisible before AI

It stayed hidden because people handle gaps differently from documents. A person who hits missing information asks a colleague, makes a sensible guess from the situation, or flags it to a manager. That absorbing happens quietly and costs nothing anyone can see. Companies have been able to ignore context debt for years because their people have been compensating for it every day, without drawing attention to it.

Why AI makes it visible

AI doesn't absorb gaps. A model given an incomplete picture doesn't stop and think "I should check with someone". It fills the gap with something plausible and carries on, and the output looks complete but is usually wrong. The context-brief demos on the Work page show the same failure in five unrelated places: a customer support reply, a financial audit brief, a legal clause review, a hiring memo and a clinical handover. It happens wherever the material a model is given is missing reasoning that used to live only in someone's head. The Context Debt Map shows what finding it looks like: twelve sources behind a fictional firm's HR and facilities knowledge, sorted by the kind of debt they carry and how urgently each needs fixing before an AI assistant goes near them.

What paying it down actually looks like

The first reaction to seeing the debt is usually to try to document everything. That doesn't work, for the same reason "write more documentation" never fixed technical debt: it doesn't prioritise, and it's never finished. Paying down context debt means finding the specific gaps that affect the work AI is doing and fixing those first: the exceptions that come up again and again, the history someone keeps having to re-explain, and the places where tone or judgement changes depending on who's doing the work. It's a focused, ongoing practice, not a one-off documentation project. If your team keeps typing the same background into AI tools, that's the place to start.

How context debt shows up breaks this down further: five recurring patterns, each traced back to real fragments from the Context Debt Map rather than described in the abstract.

The line item that doesn't exist yet

Most companies have a category for technical debt, even if they never fully fund it. Almost none have one for context debt, because the idea hasn't had a name for long enough to earn a budget line. That's starting to change. AI is the first tool most companies have used that can't work around the debt the way people do. The debt was always there; it just wasn't visible.