Change and Knowledge Management: Keep AI Work Up to Date

Change and knowledge management work together by turning an approved process change into current instructions that people and AI assistants can use. Record what changed, replace conflicting guidance, explain when the new process applies, and check a fresh handoff before calling the change complete.

fenbs can hold the person’s decision and Where we left off for the next assistant. It gives the change a shared work record, while the team remains responsible for updating source documents and checking actual behavior.

What is the practical sequence for a process change?

1. Describe the old behavior, the proposed behavior, and the problem the change should solve.
2. Identify who can approve the change and who owns the instructions it affects.
3. Record the approved choice, its effective point, and any exceptions.
4. Update the relevant guidance and mark earlier instructions as replaced.
5. Test an ordinary case, an exception, and a fresh assistant handoff using the new guidance.

Say a design studio changes client approval from an informal chat message to a review recorded by the project lead. The software change may be small. The knowledge change includes the submission checklist, the example handoff, and the instruction an assistant uses when deciding whether a draft is ready to send.

Keep the effort proportional to the consequence. A wording correction may need one owner and a quick check. A change affecting permissions or external publication needs the appropriate reviewer and a clearer test. Avoid requiring a new committee for every edit.

What must the decision say before instructions change?

An approved decision needs enough detail to settle the next action. Include the scope, the person who chose it, the reason, when it applies, and what it replaces. If approval is still pending, label it as a proposal rather than letting a polished summary make it appear settled.

For the studio, “get approval” leaves too much open. A useful decision says that the project lead must approve client-facing drafts before sending, including revised drafts, and identifies the record where that approval belongs. It also explains how work already in review should proceed.

NIST’s AI Risk Management Framework↗ (opens in a new tab) calls for clear responsibilities and ongoing review of AI-related policies. The workflow here is one practical way to apply that idea to changing project instructions; it is not a NIST certification checklist.

fenbs keeps Decisions and rules available to connected assistants, with Rules appearing first when they read the project context.

Where can old instructions keep influencing the work?

Look beyond the main handbook. People may reuse a template, a saved prompt, a checklist, a task plan, or an example from an earlier project. An assistant can receive the correct rule and a contradictory example in the same session, leaving it to guess which one takes priority.

Keep historical records useful without presenting them as current instructions. A previous decision can remain available as the reason work took a particular path. Make its replacement easy to identify so a reader does not revive an obsolete rule from a search result.

How do you introduce the change without creating confusion?

Tell affected people what they must do differently, where the current instruction lives, and how to raise a problem. A long explanation of why the change matters can accompany a short action message. Ask one person who did not write the guidance to explain it back using a realistic task.

Define the transition explicitly. If the new approval process applies only to drafts started after a cutoff, say so. If all active drafts need review, add that work to their tasks. Do not make an assistant infer the transition from the time a document was edited.

The NIST AI RMF Playbook↗ (opens in a new tab) offers voluntary actions that organizations can tailor to their own context. Use that flexibility: choose communication and review steps that fit the actual risk, rather than copying a complex rollout process wholesale.

How can you verify that the next assistant follows it?

Start a fresh session with the updated project context. Ask what rule applies, what the next action is, and which action still needs a person. For the studio, the assistant should prepare the draft and identify the approval requirement without treating its own quality check as approval.

Try a conflicting example from the earlier process and a task already in progress. Check whether the assistant notices the conflict and asks about an unresolved exception. Record the actual result, including failures; rewriting a prompt until it sounds convincing does not show that the work now follows the process.

If the test fails, distinguish missing guidance from failure to use available guidance. The first needs a correction to the record. The second may need better starting instructions, narrower permissions, or a person in the workflow. Save the outcome through the fenbs AI handoff workflow so the next session begins with the remaining check.

How to do this with fenbs

1. Create a free account and add a project for the process you are changing.
2. Add an update task with the affected instructions, proposed transition, and verification steps in its plan.
3. Record the person’s approved decision in Decisions and rules. Keep a pending proposal open until a person chooses.
4. Update the source guidance in its own system, then ask a connected assistant to read the fenbs context and help check a representative case.
5. Record the test outcome and save progress with the changed rule, remaining exceptions, and the next review step.

fenbs does not automatically update your handbook, notify every affected person, or enforce approval steps in external tools. It is not the right fit when you need a dedicated change deployment or document-control system; use it for shared project decisions and continuity alongside those systems.

Broader planning belongs in knowledge management strategy. For the surrounding work, see agile project management, project management methodologies, and the AI governance hub.