When the process lives with one person

Every business has work that “just gets done.” It rarely has a procedure anyone trusts. It lives with one person — the one who knows the exceptions, the suppliers who need a nudge, the spreadsheet that is actually the system of record.

That person is usually excellent at it. That is part of the problem.

As a Managed Intelligence Provider, we look for this pattern on purpose: repetitive work that sits with a single owner. Not because people are the problem — because undocumented, single-threaded work is where hours disappear, and where the business is most exposed when someone is on leave, overloaded, or gone.

What these processes look like

They are the jobs that move information from one place to another, apply the same judgement dozens of times a week, and keep things moving without anyone calling it strategy.

• Intake that arrives by email, PDF, or form and gets keyed into another system
• Chasing missing fields, certificates, or approvals until a case can move
• Matching, reconciling, or assembling a pack from sources that never quite agree
• First-draft replies, quotes, or summaries that always start from the same ingredients
• “Just ask Sam” knowledge — answers that exist, but only in one head

The tell is simple: if that person is out for a week, the work queues, quality dips, or someone invents a temporary workaround that nobody writes down.

How to spot them

You do not need a six-month audit. You need a few honest questions and a week of paying attention.

1. Follow a live case end to end. Pick one recent example and watch who touches it. Note hand-offs, waits, and rework.
2. Ask who covers whom. For each recurring job: who else can do this tomorrow without a briefing? A shrug means you have a candidate.
3. Listen for tribal language. “Only Julie knows the exceptions.” “It is in the shared drive — somewhere.” Those sentences are inventory.
4. Count volume and sameness. High volume plus the same rules most of the time is where systematizing — and later AI assistance — pays. Pure judgement work may stay human. That is fine.
5. Map the information path. Where does the truth start? Where does it get re-typed? Where does it wait for someone already doing five other things?

Write the shortlist down. A rough list beats the polite fiction that everything is already under control.

Why single-person ownership is risky

Key-person risk is usually reserved for executives. Operationally, it shows up quieter: the person who runs invoice intake, opens every supplier file, or rebuilds the board pack by hand.

When that work is repetitive and undocumented:

• Continuity breaks the moment they are unavailable
• Quality depends on memory, so cover becomes inconsistent
• Improvement stalls, because you cannot improve — or hand off — what you cannot see
• Heroics get mistaken for process, so the business buys overtime instead of fixing the information flow

None of that criticizes the person doing the work. Often they are why it still functions. The risk is organizational: the capability is not shared, measured, or durable.

What good looks like

Good is not “automate everything this quarter.” Good is making the work visible and transferable — then deciding what to systematize.

Document the real path

Capture how it is actually done, including exceptions. A short runbook beats a polished procedure nobody follows. Name the inputs, decision points, systems touched, and what “done” looks like.

Share ownership on purpose

At least one other person should be able to run the happy path. Cross-cover proves the process exists outside a single calendar.

Systematize the repetitive middle

Templates, structured intake, a single place for source documents, clear routing rules. Reduce re-typing and hunting. Put the rules where the work happens.

Only then apply AI to the work

Models help with reading, extracting, drafting, classifying, and triaging — after you know the process shape. Otherwise you decorate a fragile dependency with a chatbot. Consequential steps keep a human in the loop; uncertain cases go to an exception queue with a reason, not a guess.

How Kultiv thinks about this

Process intelligence starts with watching the real work — not reading the official version. We look for volume, cycle time, rework, and the exceptions nobody wrote down. The costed case comes before the build.

When a process fits, we put the smallest useful slice into production inside your environment, often shadowing humans first. Then we run it: monitoring, exceptions, accuracy, and a written monthly review. Managed operations means someone is still accountable when a supplier changes a layout or an edge case turns up at an inconvenient hour.

If your shortlist starts with “the thing only one person knows,” you are already looking in the right place. That is usually where hours are hiding — and where the business is more fragile than it feels on a normal Tuesday.