Problem
Two employees find two different answers
The customer has a straightforward question: “How much notice do I need to give to change my booking?” One employee says 24 hours. Another says 48. The first employee used a saved reply. The second checked the company’s policy page. Both thought they had found the right answer.
Now someone has to explain the difference to the customer, decide what applies to their booking, and correct whichever answer was wrong. The business updated its policy last week. Updating the places people use to answer questions is still unfinished work.
This happens with cancellation terms, expense rules, delivery promises, warranty information, and routine procedures. The approved document changes. The old answer remains in a template, a downloaded PDF, a training guide, or a message someone saved because it was useful. For employees, a simple question becomes another interruption. They ask a colleague, wait for a manager, or give the answer they used last time. For customers, it looks as though the business cannot agree with itself.
Examples
Follow one policy change through the business
Consider a fictional service business with 30 employees. It changes its booking policy from 24 hours’ notice to 48 hours. The approved change applies to new bookings made from October 1. Earlier bookings retain the previous terms. The manager updates the official policy page and sends an announcement.
But the front desk has a saved reply containing the old wording. A confirmation email still includes it. The training guide has a screenshot of the previous policy. An internal assistant searches both versions when answering questions. The employees who read the announcement remember the new rule. Others encounter the old one while doing their normal work.
There is also a customer whose booking was made before October 1. For that booking, the old rule is still the applicable one. Choosing the newest document alone would give that customer the wrong answer.
Solutions
Define which answer applies
Start by defining what makes an answer applicable. The business needs to know which policy is approved, when it takes effect, who or what it covers, and whether earlier arrangements follow different terms. Someone needs responsibility for resolving questions the policy does not answer.
APQC’s guidance on trustworthy organizational knowledge identifies outdated material, contradictory guidance, and missing context as problems that AI can amplify. It recommends approved sources and a way for people to check the source behind an answer.
Find where people use the answer
For the fictional booking policy, the next step is to identify where the answer appears. Start with the places employees and customers actually encounter it:
- The official policy page.
- The booking confirmation email.
- The front desk’s saved replies.
- The relevant training instructions.
- Any internal search or assistant that uses those materials.
Prepare specific updates for review
This gives an approved policy change a defined set of follow-up work. When the policy owner approves a new version, an automation could prepare a list of those connected materials. It could compare known templates with the changed wording, flag passages that need attention, and prepare proposed updates for review.
The reviewer sees the specific change and its scope. “Update the cancellation wording for bookings made from October 1” is much easier to act on than “Please review the handbook.”
Some steps may already be supported by the business’s existing tools. A shared reply template can avoid multiple personal copies. A confirmation email can link to the applicable terms. A policy page can show its effective date and the version it replaces.
Show what has actually been updated
Where automation adds value, it should show what has actually been updated. A changed policy page does not mean the email template has changed. A successful document update does not mean a search index has refreshed. If a connected system cannot be checked, that work remains visible.
Help the employee verify the answer
For an internal assistant, the answer should help the employee verify what applies. In this example, it could show the notice requirement, the booking date used to select it, and a link to the relevant approved version.
If the booking date is missing, the assistant should ask for it. If two approved sources disagree, it should show the conflict to the policy owner. Employees also need a clear route to ask about circumstances the rule does not cover.
Keep access appropriate to the person asking. An answer about a routine expense rule should not expose somebody else’s claim or a manager’s confidential notes.
Test the questions people really ask
Test the change with the questions people really ask. Try a booking made before the new rule, one made after it, a question without a booking date, and a request outside the policy’s scope.
Check the answer, its source, and the next action. A fluent explanation is useful only when it reflects the applicable rule.
Older versions may still be needed to explain earlier decisions or commitments. Mark which version applies to which circumstances, and manage retention through the business’s separate records policy.
Check whether the answers become more dependable
The improvement should be visible in everyday work. Count contradictory answers, repeat clarification requests, and corrections to customer messages. Measure how long an approved change takes to reach the connected materials. Include the time people spend reviewing proposed updates and correcting automated suggestions.
Then ask the employees a simpler question: Can you answer this confidently now?
A useful workflow means the front desk can find the right rule, the manager spends less time resolving avoidable confusion, and the customer gets an explanation the business can stand behind.
How Zoevin helps
Bring one old answer your team keeps finding
At Zoevin, I build scoped automations in a business’s own environment and hand them over on an agreed date.
If your team keeps finding an old answer after a policy change, bring one example. We can follow where that answer appears and look at which updates and checks could be made easier the next time the policy changes.
- Follow one approved policy change through the materials people use.
- Identify a possible automation scope and the updates your team reviews.
Zoevin has no delivered automation projects yet. Discovery names the seam. If I build, I hand it over on a written date. You get the workflow, the logins, and the write-up. I don't stay on to run it. You buy the software.
Evidence
Sources
Primary reporting first. Open the sources yourself rather than taking this account alone.
primary source
APQC — How can AI use organizational knowledge that is accurate, current, and trustworthy?Practitioner guidance on approved sources, content currency, contradictory guidance, source links, access restrictions, and human review. It does not measure small-business prevalence or prove automation outcomes. The 30-person service business, booking terms, and proposed workflow are fictional teaching examples, not client results or documented product capabilities.