The accumulated gap between what an AI system actually knows about a project and what it would need to know to work on it reliably. For agencies, it is what happens when AI workflows get built fast and nobody writes down what the model was told.
Context debt is the AI-era cousin of technical debt. It builds up when work gets done through AI agents or copilots without anyone documenting the context those tools were given: which brand guidelines, which prior decisions, which constraints. Each shortcut works fine in the moment, but the gaps compound, and eventually no one, human or agent, can fully reconstruct why a piece of work looks the way it does.
The term became common in 2026 alongside “vibe coding” and “vibe marketing,” both of which describe letting AI handle execution based on loose, conversational direction. That speed is real, but it’s borrowed against documentation that usually doesn’t get written, which is exactly what context debt describes.
Agencies are especially exposed to context debt because accounts change hands, and an AI workflow’s context often lives in one person’s chat history instead of the account files.
It makes handoffs expensive. When a strategist leaves or rotates off an account, the context their AI tools had built up, brand nuances, past client feedback, what’s already been tried, often leaves with them unless it was deliberately captured somewhere durable.
It compounds quietly. A skipped documentation step rarely breaks anything the week it happens. The cost shows up months later, as rework, when a new team member’s AI-assisted output contradicts a decision the client made in March that nobody wrote down.
It has a straightforward fix that requires actual discipline. Treating an agent’s context, brand guidelines, approved claims, prior client feedback, as a maintained asset, not a one-time input, is the difference between AI workflows that get more reliable over time and ones that get harder to trust.
An agency’s paid social team has been running a client’s ad copy through an AI assistant for months, feeding it fresh examples and corrections in each session but never updating a shared brief. When the lead strategist goes on leave, her replacement’s AI-generated copy reintroduces claims the client rejected back in Q1, because that rejection only ever lived in the original strategist’s chat history. The agency spends a week rebuilding a written context document that should have existed from day one, and makes updating it part of the account’s weekly routine going forward.
The automations and agents module of the workshop teaches you how to build AI workflows that compress the busywork without taking the craft out of the studio.