Skip to main content
AI agents

Make what your longest-serving staff know findable before they leave

A Company knowledge assistant finds answers already written in your documents, and cites the source. What it can capture, and what it can't.

Reenva

6 min read

Picture Robyn. She's made up, but you may know someone like her. Robyn has worked in dispatch for nineteen years.

She knows the Cairns run leaves forty minutes early in the wet season. She knows one customer's site won't take deliveries on a Tuesday. She knows the "special pricing" flag means something different on accounts opened before 2014.

None of it is written down, and all of it keeps the business running. Ask an owner what happens when Robyn retires, and you'll get a nervous laugh. In many growing businesses, a few people carry this load, and some are close to retirement.

Knowledge kept in people's heads quietly costs paid hours

The dramatic version, where someone leaves and things break, is real but rare. The everyday version costs more, because it never shows up in a review.

A customer service rep spends twenty minutes on a product question that has been answered in writing four times. The answers sit in an email thread, a PDF on the file server, an old ticket and an unfiled supplier bulletin.

A new estimator prices a job without the margin history on that client, and leaves margin on the table. A new branch manager takes a year to get up to speed. The role isn't that complex; the manual just doesn't exist, and the answers live in six people's heads.

None of these incidents is big enough to escalate. Across a fifty-person business, they add up to a lot of paid time spent hunting for answers you already own.

Why an intranet or wiki rarely fixes it

Many businesses have tried to fix this once, usually with SharePoint or a wiki. It often goes the same way: a keen start, a few dozen good pages, then decay. Within a year or two, the pages are out of date, and everyone asks Robyn again.

The cause isn't laziness; it's economics. Most knowledge-base projects ask busy people to write things down and keep them current. The benefit goes to someone else, later, maybe.

That trade rarely lasts. The people who know the most have the least time to write it down.

A plan that starts with "first, everyone documents what they know" is likely to fail. Much of the knowledge is already written down, in quotes, job notes and supplier documents. The problem is finding it.

A Company knowledge assistant finds the answer and shows you the source

A Company knowledge assistant fills this gap. It's worth being precise about what it does, because this kind of tool attracts inflated claims.

Nobody has to write anything new. It searches the documents you choose: SharePoint and file shares, job notes and supplier documents. It answers questions in plain English and cites the documents it used.

Ask "What's the lead time on that pump for a WA delivery?" You get the answer, the supplier bulletin and the past quotes that mention it. That takes seconds, not twenty minutes.

Three things decide whether staff trust the assistant or give up on it:

  • Sources shown. It is set up to name the documents behind each answer. Staff can open the source and check it, so trust comes from checking, not faith.
  • "I don't know" beats a confident guess. When your documents don't hold an answer, it is set up to say so. An assistant that guesses smoothly is worse than none, because it turns doubt into confident errors.
  • Permissions first. It follows your SharePoint permissions, so staff only get answers from documents they can already open. Check those permissions before anything is indexed, so a loosely shared payroll folder stays private.

What it can't capture: judgement that was never written down

An assistant can only find what was written down at some point. Robyn's judgement was never in a document, so it won't be in the index. That includes her feel for when a customer is about to leave, or when a job is quoted too thin.

What the assistant captures is the findable share of her nineteen years: facts, precedents and answers written once and lost. That can be a large share. It isn't all of it.

So use each for what it does best. Let the assistant do the finding. Spend people's time on the judgement that needs a person: pairing, handover and conversations before the retirement lunch.

Start with one team's documents

Projects that work tend to start narrow. Pick the area where it hurts most, such as supplier documents for the counter team or client history for estimators. Index those documents, check permissions, and put the assistant in front of real users.

It earns its place in one corner of the business before it gets the rest. A company-wide knowledge programme tends to fail, for the same reason the intranet did. Its size sinks it before people start using it.

Robyn is still going to retire. The question is whether nineteen years of answers retire with her.

+ Your next step

Who's your Robyn?

Book a 30-minute call. If there is a fit, a two-week fixed-fee review maps where your knowledge lives, checks permissions and scopes a first assistant for one team.

+ Your next step

Want to know what this would mean for your business?

Book a 30-minute call →

If there is a fit, a two-week fixed-fee review follows, ending in a clear go or no-go.