Keeping automated processes under control: reports, exceptions and the monthly review
Automated processes take work off your team's plate. What they don't take off your plate is the responsibility for making sure the work is done correctly. An automated process that nobody looks at anymore runs fine until something changes: a new kind of letter, an access that expires, a volume that's suddenly twice as large. That's why, with us, automation doesn't end with the sign-off. This article is about what comes after: the report for each run, the way exceptions reach your team, the monthly review of all processes, the question of when a reviewing role makes sense, and the emergency brake.
Why an automated process needs leading
A person on the team usually notices on their own when something has changed. They see that a form suddenly looks different, and they ask. An AI employee asks too, but only if it recognizes the case as unknown. That's the escalation path. Anything it believes it knows and still sorts wrongly only gets noticed when someone looks.
These are the quiet mistakes. A notice is in the wrong folder, an invoice is filed under the wrong client, and the report still says everything is done. Mistakes like these are often only noticed from the outside, when someone looks for something or asks about it.
Leading here means looking at three fixed points, without redoing every run: after every run in the report, with every exception through the escalation path, and once a month at all processes together.
The report for each run
With us, a process is done when it starts without a nudge, runs through to the end and sends you a report: what's done, what's still open, where there's an exception.
Without a report, you end up following up yourself, and you're back to managing every step.
For leading the process, one thing about the completion email counts above all: after a short read, you have to know whether you need to do something. That's why the numbers are at the top, right in the subject line, and below them, in a fixed order, what's done, what was left open, which exceptions there were and what needs a decision. If there's nothing for the last point, it explicitly says "none". If you notice that you have to read the whole email every time, its structure gets sharpened. What the completion email looks like in detail is covered in Autonomy Level 2, Autopilot: the process starts on its own and reports back at the end.
When the report doesn't show up
A missing report is a message too. If the completion email doesn't arrive, something didn't run, and someone takes a look. For some cases, we also define that the AI employee reports immediately instead of waiting until the end, for example when a run doesn't even start or an access is missing.
Exceptions: the escalation path to your team
If an AI employee doesn't recognize a case, it doesn't process it. It sends the case to a person on your team and lists it as open in the report. In a feedback session, the AI employee is trained on the new variant, and it's saved as a rule.
For leading an automated process, the exceptions are the most valuable part. They show where the rules still have gaps. Two things help here:
- One fixed person. The job description of every AI employee names a contact person on your team. The exceptions go to that person, not to everyone.
- An eye on repeats. If the same exception comes up several times in one week, that's a sign a rule is missing. Then it gets added in the feedback session, instead of every case being answered one by one.
That's how an automated process keeps learning over time. Every exception that gets answered is one variant fewer that goes to a person next time.
The monthly review
In everyday work, you see the process that's currently stuck. How all of them are doing, you only see if you deliberately look at all of them at once. For every process, you can see which level it's on. Once a month, we look at it together and decide what comes next. Which few metrics you need on one page for this is covered in KPI dashboard for a business with a team: a few metrics you actually read.
For every process, we distinguish between three outcomes:
- Up: The process has cleared the bar for the next level. From Level 1 to Level 2, that's four clean runs, then monitoring side by side, then running on its own with a final report.
- Hold: The process runs at its level, but something is still missing for the next one. Then what's missing gets written down.
- Back: Runs failed, or a person had to step in because results were wrong. An escalation email doesn't count here. It's the intended path. Then the lower level gets entered.
Only a level that has actually been reached gets entered, not one that feels like it has. That sounds strict. But it's the reason you can trust the monthly review.
You can do this without a program too. Once a month, take the list of your automated processes and ask for each one: which level is it at today? What held it up this month? And which process is up next?
The month's exceptions
One part of the monthly review is especially worthwhile: the month's exceptions side by side. Answered one at a time, every exception looks like a special case. Side by side, you can see which ones come up several times, which ones go to the same person and which rule is missing. That tells you what gets trained in at the next feedback session. And sometimes it also shows that a process has a variant that belongs more to thinking than to doing, and should therefore stay with a person.
When a reviewing role makes sense
One or two automated processes can be led well through their completion emails. Every additional process brings another email. At some point, you only skim the reports, and what you skim, you don't check.
Then the next step is Autonomy Level 3, Audited: an AI manager reviews the work and bundles the reports, so you only see the exceptions. In the program, that's a goal after the first process: the AI manager bundles the reports into a morning report and flags exceptions. In Kevin's business, this role is still on trial.
With us, the reviewing role works like this: it doesn't do any of the actual work, doesn't approve anything and doesn't change other employees' results. It reads, compares and reports. What that looks like exactly, and why this level only comes after Level 2, is covered in AI governance in SMEs: who checks the AI when you no longer do? Whether your business already needs a role like this, and which simpler measures work first, Kevin describes on his blog: Does Your Business Need an AI Lead Role?
The emergency brake
Every automated process comes with an answer to the question of what happens if something goes wrong after all. With us, that answer is fixed before the first run, in three parts. The AI employee stops on its own when it doesn't recognize a case, and hands it over. In front of every step that can't be taken back, the process waits for a person's explicit yes, and there's a deputy for that. And if a process fails, it's settled who carries it on by hand. The three parts are described in detail in AI governance in SMEs: who checks the AI when you no longer do?
For leading your processes, that means it's best not to wait for an emergency to test the emergency brake. We recommend that the monthly review also asks whether the deputy for every approval point is still right, and whether the person who would take over a process by hand is still on the team and knows about it.
Behind this is our house rule: we keep a human involved when decisions are about people, about money or about anything with legal consequences. That applies at every level, including for a process that has been running cleanly for months.
What stays with you as the owner
How much you still look at yourself depends on the level of the process:
- At Level 1, Assisted: Your team checks every run. You see what your team reports to you.
- At Level 2, Autopilot: You, or the person on your team who looks after the AI employees, read the completion email. You decide for each process who gets it.
- At Level 3, Audited: You get a report in the morning and only see the exceptions.
At every level, the decisions only a person can make stay with you, and so does the monthly review. What changes is the kind of work you do: less following up, more deciding where you're needed.
What this looks like for us
At a tax advisory firm, two automated processes are running: filing digital tax assessment notices and handling the notices that arrive on paper. For the paper notices, a second AI employee checks the inbox every 30 minutes, spots the emails with paper notices and hands them over to filing. Both processes are at Autonomy Level 2, Autopilot (as of September 2026).
They're led through exactly the points described in this article. Kevin gets completion emails. Anything the AI employee can't match with certainty goes to the office team by email, followed by a feedback session in which the new variant gets trained in. So the office team hasn't gone anywhere. It gets the exceptions and keeps the rules up to date.
We haven't counted how many exceptions arrive there each month, so we don't give a number here.
In Kevin's own business, 14 AI employees work, each with a personnel file, a fixed role and clear limits. An AI manager reviews their work; that role is still on trial. Kevin tries new levels in his own business first, before they come to yours.
The monthly review is a fixed part of Owner Autonomy. We don't have documented results from these monthly rounds yet. And Autonomy Level 4, the week-away test, is a goal for us. No process of ours has gotten that far yet.
Runs without me
Every two weeks, one process that runs without the owner: what came in, what the AI did, where a person decided. Two minutes to read.
Worksheet
Process Potential Check
With five questions and a first process list to fill in, you find the process that could run on its own first. Comes as a PDF by email once you confirm your sign-up.
Strategy session
In a 20-minute strategy session, you tell us where things get stuck. We'll tell you honestly whether an AI employee is a good fit for you.
Book a strategy sessionFrequently asked questions
How much time does managing automated processes take?
That depends on the number of processes and exceptions. We don't give a fixed number. The completion email is structured so that after a short read, you know whether you need to do something, and the monthly review bundles the rest.
What happens to the processes when I'm on vacation?
The processes at Level 2 are built to keep running and reporting. The approval points have a deputy, and the exceptions go to the contact person on your team. Whether a process runs through a whole week without you is the test for Level 4. That's what Autonomy Level 4, Autonomous: the week-away test is about.
Can a process be moved back down a level?
Yes. If runs fail or a person has to step in again, the lower level gets entered the next month. The case gets trained in, and your team watches alongside again until the runs are clean again.
How does a process get this far in the first place?
Through the process map, choosing process no. 1, the process profile, training and sign-off. The whole road is described in How to automate business processes when you have a team: the path in six steps. What Autopilot means exactly is covered in Autonomy Level 2, Autopilot: the process starts on its own and reports back at the end.
Your next step
Automated processes need a report, a path for exceptions and a fixed monthly review. If you want to know what that would look like in your business, we'll look at it together on a call.