Skip to content
Zevqi
AI Engineering

Bus Factor in the AI Era: How to Find Single Points of Knowledge in Your Codebase

AI agents can create code nobody understands and accidental single owners. Learn how to measure blind spots, find single points of knowledge, and raise your bus factor on purpose.

By Neha SinghPublished November 16, 20266 min read

The bus factor is the number of people who would need to disappear before a project stalls. A bus factor of one means a single person holds critical knowledge. AI coding agents were supposed to make this problem smaller. In practice, many teams have made it worse, in a new way: the code exists, but the knowledge of how it works never formed in anyone’s head.

What changes in the AI era

Traditionally, the bus factor risk was concentrated in senior people who wrote and remembered the hard parts. With agents, a different pattern appears:

  • Hidden zero-owner code. Files nobody on the team has really read, because an agent wrote them and the review was a skim.
  • Accidental single owners. One developer happened to prompt the agent for a feature, so they alone have context.
  • Lost reasoning. The conversation that explained the decisions is gone, so even the prompter’s understanding fades.

How to measure it in a real codebase

You do not need a perfect number. You need a useful map. Three views are enough:

  1. Blind spots: changed files with no person who has demonstrated understanding.
  2. Single points of knowledge: files with exactly one person who understands them.
  3. Coverage by area: for each directory, what share of changed files has at least one owner.

Weight everything by risk. A single owner on a login flow matters more than a single owner on a marketing page. Our dashboard ranks files by a simple risk score that combines how recently the file changed, how often, and whether the path looks sensitive (authentication, billing, infrastructure and so on).

Ways to raise the bus factor on purpose

  • Rotate lessons. When an AI change lands in a single-owner file, ask a second person to walk through it.
  • Write short explanations at the time. Two paragraphs beside the code beat a wiki page nobody finds.
  • Pair on the scary parts. Billing, auth and data migrations are worth the time.
  • Review ownership monthly. Put the list of blind spots and single owners in your planning meeting.
  • Onboard through real changes. New hires learn fastest from recent diffs with explanations.

Do not turn this into a scoreboard

If measuring understanding becomes a way to rank individuals, people will game it or resent it. Use the numbers to find gaps in the system, not gaps in people. Quiz scores that are self-reported are a signal, not an audit.

Where Zevqi fits

Zevqi turns each AI-written change into a short lesson and quiz, then rolls results into exactly the views above. Open the demo dashboard to see blind spots, single points of knowledge and coverage trends on sample data, or start a free trial on your own team.

Frequently asked questions

What is a good bus factor?

For critical code, at least two people should understand each file well, preferably more. For low-risk code, one may be fine.

Can version control history tell me the bus factor?

It shows who changed a file, but with AI-written code the author of a commit is not necessarily someone who understands it. Understanding needs its own signal.

Related: what is comprehension debt?

Next: how Zevqi helps your team understand the code AI writes.

Written by Neha Singh

Neha Singh is the founder of Zevqi. She builds and tests the tools on this site and writes these guides from hands-on use. Follow on X (@heynehasingh) or read more about Zevqi.

Try it now

Ready to put this into practice? Zevqi has the tools built in — free, no signup.