AI versioning needs a change record. An AI system’s behaviour depends on more than its code: the model weights, the prompt template, retrieval sources, and configuration parameters can each change independently, and any one of them can shift the output.

Google’s MLOps guidance documents model, data, and pipeline versioning as distinct tracked artifacts, and general software configuration management practice applies the same discipline to the prompt and parameter layers unique to AI systems.

AI operations is the discipline of running deployed AI systems the way any production service is run: with named owners, tested rollback paths, cost visibility, and evidence that the system still does what it was approved to do. It sits after the build phase and before the system is forgotten about.

Teams that treat a model launch as the finish line tend to discover the real work only after something breaks: a quiet accuracy drift, an unexplained cost spike, or a change nobody tracked. Operations work is what prevents that discovery from happening in front of a customer.

Keep the operating record close to the system, not buried in a slide deck. A reviewer six months later should be able to reconstruct what was decided, why, and what evidence supported it.

Treat AI systems as living services. Usage patterns shift, upstream providers change models without notice, and the data feeding a system evolves. An operating model built for a static deployment breaks quickly against that reality.

Avoid confusing activity with control. A busy Slack channel about a model is not the same as a defined process with an owner, a trigger, and a record of what happened.

On this page

Version every layer that affects output

Track the model version, prompt template, retrieval source, and key parameters as separate versioned artifacts, not as one combined release.

A change to any one layer can shift behaviour. Versioning only the application code leaves the layers most likely to cause a regression untracked.

  • List every layer that affects output.
  • Version each layer independently.
  • Record which layers changed in each release.

Record the reason for every change

Attach a short rationale to every version change: what problem it solves, what was tested, and what risk was accepted.

A version history with no rationale becomes a list of dates nobody can interpret months later when trying to understand why behaviour shifted.

  • Require a rationale on every change.
  • Link the rationale to test evidence.
  • Keep the record searchable by version.

Make rollback a first-class version operation

Treat reverting to a prior version as a normal, tested operation rather than an emergency improvisation.

Teams that only ever move forward through versions discover during an incident that reverting is untested and slower than expected.

  • Test rollback as a routine operation.
  • Keep prior versions retrievable and deployable.
  • Time the rollback and record the result.

Connect versions to their evaluation results

Store the evaluation score and test conditions alongside each version so a regression can be traced to the exact change that caused it.

Without this link, a team investigating a regression has to guess which of several recent changes was responsible.

  • Attach evaluation results to each version.
  • Store test conditions with the score.
  • Use the record to isolate regressions quickly.

Assign a named owner

A AI versioning practice only works when one accountable person can explain the current state, not when the responsibility is spread across a channel nobody checks.

Write the owner into the runbook itself, next to the review cadence and escalation path. Rotate ownership deliberately, with a handover record, rather than letting it drift when someone changes teams.

  • Name the accountable owner.
  • Record the review cadence.
  • Define the escalation path.

Keep the evidence, not just the dashboard

A dashboard number is a claim. The evidence behind AI versioning is the log, the test result, or the approval record that a reviewer can check independently.

Store evidence close to the decision it supports, with a timestamp and the person who reviewed it. Delete evidence on a defined retention schedule rather than an indefinite pile nobody prunes.

  • Keep raw evidence, not summaries alone.
  • Timestamp every record.
  • Set a retention and deletion rule.

Review after material change

Ai versioning decisions age. A model version change, a new tool integration, a new data source, or a usage spike can invalidate a decision made months earlier.

Pair a scheduled calendar review with change-triggered reviews. The calendar catches slow drift; the trigger catches the event a calendar would miss entirely.

  • Set a fixed calendar review.
  • Define change triggers.
  • Log what changed and why it mattered.

Make the failure path explicit

Most AI versioning programs are designed around the happy path. Test what happens when the process is skipped, delayed, or overridden under pressure.

Record the degraded-mode behaviour and who is allowed to invoke it. An undocumented exception becomes the normal path the moment the team is busy.

  • Test the skip and override case.
  • Name who can approve an exception.
  • Log every exception used.

Connect the metric to a decision

A AI versioning metric earns its place on a dashboard only when a defined action follows a defined threshold.

State the denominator, the period, and the owner for every number. A metric with no attached decision is decoration, not governance.

  • Define denominator and period.
  • Attach an action to the threshold.
  • Retire metrics nobody acts on.

Avoid the common early mistakes

Most teams new to AI versioning repeat the same few mistakes: treating it as a one-time setup task, assigning ownership to a group rather than a person, and building the process around whatever tool was easiest to install rather than the risk it needs to cover.

These mistakes are cheap to fix early and expensive to fix once the practice is embedded across many systems. A short review against this list before the first production rollout catches most of them, and repeating the review after the system has been live for a full quarter catches the rest, since some gaps only become visible once real usage patterns diverge from what was assumed during design.

  • Do not treat it as a one-time setup step.
  • Assign a person, not a group, as owner.
  • Build the process around the risk, not the easiest tool.

Operating rule: A control only counts once a named owner, a review trigger, and stored evidence all exist for it.

Ai versioning is an operating discipline, not a one-time setup task. Keep ownership, evidence, and review cadence visible so the system stays explainable as it changes.

Revisit the decision after every material change and keep a record a new team member could follow without asking around.

Decision table

Area Question to answer Evidence to keep
Layers What actually changed? Model, prompt, retrieval, parameters
Rationale Why was it changed? Problem, test evidence, accepted risk
Rollback Can it be reverted quickly? Tested procedure, retrievable prior version
Evidence What proves the version worked? Evaluation score, test conditions

Related Global Tech Insights reading

FAQ

What should be versioned in an AI system beyond code?

The model artifact, prompt templates, retrieval sources, and key configuration parameters, since each can independently change system behaviour.

Why record a rationale for every version change?

Without it, a later investigation into a regression has no way to understand what problem the change was solving or what was tested beforehand.

How is AI versioning different from standard software versioning?

It tracks additional layers, model weights and prompts, that do not exist in traditional software and that can each independently shift output.

Should rollback be tested before it is needed?

Yes. An untested rollback procedure discovered during a live incident is slower and riskier than one rehearsed in advance.

Who should own AI versioning?

One named accountable person or role, even when several teams contribute. Shared ownership without a single accountable owner tends to leave gaps nobody notices.

What is the first operating task for a new AI system?

Name the owner, define the review cadence, and record the rollback path before scaling usage.

How often should the operating decision be reviewed?

On a fixed calendar plus every material change to the model, data, tooling, or usage pattern.

What counts as evidence rather than a claim?

A log, test result, or approval record a second person can independently check, not a summary or a dashboard screenshot alone.

Conclusion

Ai versioning works when it is owned, evidenced, and reviewed. Treat the deployed system as a living service, not a finished project, and keep the operating record close to the decision it supports.

Sources

Previous post AI Evaluation Needs a Test Set Owner
Next post Human Review Loops Need a Bounded Queue