Why Two People at Your Company Answer the Same Question Differently

A customer asks what your refund window is. One person says 30 days. Someone else says 30 days unless it was a custom order. Both have been here for years. Both are describing something real.

Key takeaways

• This is not a training problem. Both answers were correct at some point, and the difference is which version each person learned.
• 42% of institutional knowledge is unique to the individual who holds it, so a lot of what your company knows exists in exactly one head.
• People whose jobs are mostly communication and coordination spend nearly 20% of the workweek looking for internal information, according to McKinsey.
• Wikis fail on upkeep, not on setup. The page is accurate the day it is written and decays from then on.
• The durable fix is an answer that updates from the work, rather than one that waits for someone to remember to edit it.

The answers diverge because the policy did

Nobody in this story is wrong. The refund window really was 30 days flat. Then, some time later, someone made an exception for custom orders because a supplier could not take them back, and that exception became the practice without ever becoming a policy.

The person who learned the rule before that change answers one way. The person who learned it after answers the other. Neither has any reason to doubt themselves, so nobody escalates it, and the customer gets whichever they were told.

Every company accumulates these. The refund window, whether a discount needs approval, which supplier you use for rush jobs, who signs off on a contract over a certain value. They are all decisions that were made once, communicated to whoever was in the room, and never written anywhere durable.

Why so much of it lives in one person’s head

42% of institutional knowledge is unique to the individual who holds it. Not documented anywhere, not known to a colleague, just in one person’s memory of a conversation.

That is a background risk while everyone stays. It becomes an acute one when they leave, and people leave more often than they used to: median tenure is 3.9 years, the lowest since 2002. Knowledge loss is continuous rather than occasional.

Meanwhile the searching costs real time. McKinsey put the figure at nearly 20% of the workweek spent looking for internal information, and inefficient knowledge sharing costs a large business $47 million a year. Employees waste 5.3 hours a week waiting for information from someone else.

That $47 million is a large-business figure and it describes nothing you will ever see on your own books at twenty people. What you see at that size is smaller and more specific: answering a routine customer question meant interrupting the one person who knows.

Why the wiki did not fix it

Most teams respond by writing things down. That instinct is right and the usual execution has one flaw: it treats documentation as a project with an end, when the thing being documented keeps moving.

The wiki is accurate the day it is written. Then the refund exception happens in a Slack thread. The new supplier is agreed over email. The approval threshold changes in a meeting. None of those reach the page, because updating it is a separate act of remembering, performed by someone who already knows the answer and therefore does not need the page.

So the page ages into something worse than nothing: a document that looks authoritative but is actually wrong in three places. People learn not to trust it, go back to asking a colleague, and you are where you started with an extra thing to maintain.

The real test for any knowledge tool is month seven, when the person who set it up has moved on to something else.

What actually keeps an answer current

The decisions are not lost. They happened in writing, in the threads where the work happens: the Slack conversation where the exception was agreed, the email confirming the new supplier, the message where someone approved the discount.

The useful question is why the record of a decision sits somewhere other than the place the decision was made. “How do we get people to document more” has been asked for thirty years and answered badly.

So: answers should be assembled from the messages that contain them, and updated when a later message contradicts an earlier one, so that being current is the default rather than a chore.

Where we are on this

We build this+that.

The Brain is our attempt at exactly this problem: a shared place for what the company knows, that your team’s messages feed rather than something a person has to keep topped up. Workflows can write back to it when they learn something while running, so a page can be updated by the work rather than by a reminder.

The limits, in a category full of overclaiming. It answers well from what you put in it and from what your workflows write back. The version that maintains every page entirely on its own, with no setup, is where we are heading rather than where we are. And it has a cold-start problem: on day one it knows nothing, and it gets useful as messages accumulate.

If someone tells you their tool solves organizational knowledge outright, check what it does in month seven.

Frequently asked questions

Why do employees give customers different answers to the same question?

Usually because the policy changed and the change was communicated verbally or in a thread rather than recorded. Each person is accurately repeating the version they learned.

How much company knowledge is undocumented?

42% of institutional knowledge is unique to the individual who holds it, meaning it exists in one person’s memory rather than in any shared record.

Why do internal wikis stop being accurate?

Because updating them is a separate act from doing the work. The people who know about a change already know it, so they get no benefit from writing it down, and the page decays while looking authoritative.

What does inconsistent internal knowledge cost?

McKinsey puts information-seeking at nearly 20% of the workweek. Separately, employees lose 5.3 hours a week waiting on information held by someone else.

How do you keep company knowledge current without nagging people?

Derive it from where decisions actually get made, which is your email and chat, rather than maintaining a separate document that depends on someone remembering to edit it.