Every business has workflows that nobody owns. They cross team boundaries, live in shared inboxes and spreadsheets, and get done because someone happens to notice. Their cost is real, but it rarely appears on a report, because no report was ever built for work that belongs to nobody. This field note shows how to find these workflows, how to put a number on what they cost, and why they often make the best first mission for an agent — once someone owns them.
What an unowned workflow looks like
You can usually recognise one by its symptoms rather than its name.
- Work arrives in a shared inbox, a group chat or a spreadsheet, and whoever sees it first picks it up — or assumes someone else has.
- The question "who is handling this?" is asked more than once a week.
- Status is found by asking people, not by looking anywhere.
- Exceptions are resolved by whoever happens to know how, and the knowledge stays with them.
- The same problem is fixed again and again, because nobody is responsible for fixing its cause.
- Skilled people spend part of their day acting as routers: reading, forwarding, copying and chasing.
None of these is dramatic. That is why they persist.
Where the cost hides
| Cost | How it shows up | Who notices |
|---|---|---|
| Waiting | Work sits in a queue until someone picks it up | The customer, supplier or colleague who is waiting |
| Rework | Two people handle the same item; one item is handled twice | Rarely anyone |
| Error and risk | Steps skipped, checks missed, data entered twice and differently | Finance, audit, or the customer — later |
| Expertise drain | Senior people spend time routing and chasing instead of deciding | The people themselves, as frustration |
| Lost opportunity | A slow reply to a lead, a renewal not chased, a problem not escalated | Nobody, because it never happened |
Oracle note — The costs that matter most are the ones nobody measures. Waiting time and lost opportunity do not appear in any system, because the work that would have recorded them was never done.
The risk nobody signed off
Unowned workflows are not only slow. They are often where the riskiest actions happen without anyone having approved how. Bank details are changed on the strength of an email. Customer data is exported to a spreadsheet to answer a question and never deleted. Access is granted because someone asked in a chat channel. Each step is done in good faith by a helpful person. None of it is recorded in a way an auditor, or the business itself, could follow afterwards.
When something goes wrong in an owned process, there is a person to ask and a record to check. In an unowned one, there is usually neither.
How to find them: a one-week queue audit
You do not need new tools to find unowned workflows. You need a week and a spreadsheet.
- List the queues. Every shared inbox, request form, chat channel, tracker and "pending" folder where work arrives and waits.
- Count the volume. How many items arrive each week? A rough count from a sample day is enough to start.
- Time the wait. How long does an item sit before someone picks it up, and how long until it is done?
- Count the hands. How many people touch a typical item, and how many systems?
- Mark the ownership. For each queue: a named owner, a shared responsibility, or nobody. Be honest; "the team" usually means nobody.
- Ask what happens on holiday. If the person who usually handles it is away, what happens? If the answer is "it waits", you have found a single point of failure as well.
Expect to find more queues than you thought you had, and some with no owner at all.
Putting a number on it
A simple model is enough to compare queues and make a case. For each queue, estimate:
- Handling cost — items per week × minutes per item, converted into hours.
- Rework cost — the share of items handled twice, or by more than one person.
- Waiting cost — what the delay costs: late fees, missed discounts, lost sales, lost customers, or simply complaints.
- Error cost — the rate of errors and what each one costs to find and correct.
Illustrative example, not a client engagement. A shared inbox receives about 120 supplier queries a week. Each takes around 15 minutes to read, look up and answer: 30 hours a week. About one in five is picked up by two people before anyone notices, adding roughly 6 hours. Queries wait three days on average for a first reply, and twice a month a duplicate payment slips through. Before counting the waiting and the errors, the inbox already consumes the better part of a full-time role — and it has no owner.
The numbers do not need to be precise. They need to be good enough to rank one queue against another, and to give you a baseline to measure against later.
From cost to first mission
The most expensive unowned workflow is not automatically the right place to start. As the Guild's use-case library puts it, the best first missions combine meaningful volume, clear rules, accessible data and human review. Look for repeated work, context scattered across several systems, waiting time customers can feel, and a measure you can compare before and after. Choosing the First Automation turns a list of candidates into a scored decision.
But the first step is not an agent. It is an owner. Give the workflow one accountable person, a signal that tells them when work is waiting or going wrong, and a defined response. "Give every operation an owner, a signal, and a response" is how the Guild describes managed operations, and it applies before anything is automated.
Once the workflow has an owner, an agent can take on the parts that cost the most: reading and classifying what arrives, gathering the context from different systems, routing it to the right person, drafting the response, and recording what happened. The owner decides what the agent may do alone and what needs approval.
Oracle note — Measure the queue before the agent arrives. Volume, waiting time, handling time and error rate today are the only fair baseline for judging the agent later.
Questions to ask this week
- Where does work arrive and wait in your organisation, and do you have a list?
- Which of those queues has a named owner, and which belongs to "the team"?
- How long does a typical item wait before anyone touches it?
- How many people and systems does a typical item pass through?
- What happens to the queue when the person who usually handles it is away?
- Which queue would you most like to stop worrying about?
The answer to the last question is usually where the first mission is.
