AI-assisted development can 10x output-and 10x the rate at which technical debt accumulates if left unmanaged. VP-level leadership must set the balance. If you only measure feature velocity, you'll optimize for speed at the expense of long-term health; if you only measure debt, you'll slow down delivery. The art is in making both visible and allocating capacity explicitly to both.
Debt as a strategic variable
Not all debt is equal. Categorize by impact: system-critical vs. localized, security vs. maintainability. Allocate a percentage of capacity to debt reduction and make it visible in planning and exec reviews. Many teams use a 70/20/10 or similar split: most capacity to delivery, a meaningful slice to debt and reliability, and a slice to experimentation. Adjust for your context, but make the split explicit so it doesn't get squeezed to zero when deadlines loom.
Create a simple debt register: system or area, type of debt, estimated effort to address, and owner. Review it in your leadership sync and in board or exec updates so that "we're paying down critical debt in X" is a normal part of the story. When you defer debt, document why and when you'll revisit so it doesn't become invisible.
AI and debt creation
Codegen can produce more boilerplate, duplication, and inconsistent patterns. Counter with strong linting, shared libraries, and architecture decision records. Use AI to help refactor and document, not only to generate new code. Set expectations in your engineering standards: AI-generated code must meet the same review and test bar as hand-written code, and reviewers should flag patterns that increase debt (e.g., copy-paste across modules, missing error handling).
Invest in shared patterns and templates so that AI-generated code starts from a good baseline. If your lint rules, test patterns, and ADRs are clear, both humans and AI can produce more consistent output. Periodically run debt-spotting exercises: use static analysis and manual review to find the biggest sources of duplication or complexity, then prioritize refactors that unlock multiple teams or reduce incident risk.
Metrics that matter
Track DORA-style metrics alongside debt indicators: build times, test coverage of critical paths, incident frequency. Report both delivery and health so the board and peers see the full picture. If you only report velocity, leadership will assume health is fine; if you only report incidents, they may not see the delivery wins. Combine them in a single narrative: "We shipped X and reduced critical debt in Y; our deploy frequency is up and our MTTR is down."
Set targets for debt reduction (e.g., "reduce build time by 20% this half" or "eliminate the top 5 legacy hotspots") and track progress. Celebrate when the team hits these goals so that paying down debt is seen as valued work, not a side project. Over time, this builds a culture where sustainability is part of how you ship, not an afterthought.
Speed without sustainability is a short-term win. Own the narrative: we ship fast and we keep the foundation strong.