Open your project board. How many say in progress? How many actually are?

,
By Cheri Stockton, Chief Technical Therapist, Hot Hand Media.

Automation running unwatched is the reason your project board shows tasks as in progress when nothing is actually moving, because a status label reflects what someone set once and not what a system is doing right now. The board tells a story. The work tells another. Nobody reconciled them.

  • A status label is a claim, not a fact, and automation running unwatched turns every stale claim into operational chaos.
  • The gap between “in progress” and actually in progress is where quiet failures hide for weeks.
  • Automation that keeps running past the point it should have stopped is a governance problem, not a tooling problem.
  • Most project boards decay because no human and no system owns the job of reconciling label to reality.
  • You can catch the gap in two minutes by counting two numbers and comparing them.

What is automation running unwatched?

Automation running unwatched is any process you handed to a tool or an agent that keeps executing with no human and no check confirming it still does the right thing, so it drifts past its stopping point while the board that tracks it reports a status nobody updated. It is the task that says in progress because a trigger fired three weeks ago. It is the workflow still pulling data into a sheet nobody reads. The system looks alive. The value stopped.

Here is the line that started this whole thing for me. A client said it out loud, almost cheerful about it. “I don’t even know, Cheri, to be honest with you. It’s probably connected to Claude and doing my tasks.” Probably. Connected. Doing tasks. Three words of pure faith, zero visibility.

A status label reflects what someone set once, not what a system is doing right now.

That gap is normal, right? You set a card to in progress, you move on, the card stays where you left it. The automation behind it may have finished, failed, or looped forever. The card does not know. The card cannot know. It only knows what the last human told it.

Why does your board say in progress when nothing is moving?

Your board says in progress when nothing is moving because status in most tools is a manual field that someone updates by hand, while the actual work runs in a separate automated layer that never reports back, so the two drift apart the moment you stop babysitting the connection between them. No sync. No reconciliation. Just two systems telling two stories.

Think about where the label lives versus where the work lives. The label sits in your project tool. The work runs inside an automation on Make.com, or a scheduled job, or an agent connected to Claude or ChatGPT. Nobody built a bridge that says “when the work stops, change the label.” So the label stays frozen in the last state a human touched.

This is the quiet version of operational chaos. Not a fire. A slow drift. Cards pile up in a column that means nothing. You scan the board, you feel productive, and half those cards are ghosts.

The gap between what a board claims and what a system does is where quiet failures live the longest.

The two layers that never talk

Every automated operation has a reporting layer and a doing layer. The reporting layer is your board, your dashboard, your status column. The doing layer is the actual automation. When those two layers are not wired to each other, you get a board full of confident labels describing work that may have died days ago.

  • The reporting layer updates when a human clicks.
  • The doing layer updates when a trigger fires.
  • Nothing updates when the doing layer fails silently.

That third line is the whole problem. Silent failure does not move your card to a “broken” column. It just stops. And a stopped automation looks identical to a running one if nobody is watching the output.

What is the real cost of automation nobody is watching?

The real cost of automation nobody is watching is not the wasted compute, it is the false confidence, because you make decisions based on a board that says work is happening when the work stopped, and that confidence compounds into missed handoffs, dropped clients, and reports built on data that quietly went stale. You trust the dashboard. The dashboard is lying by omission.

Industry research backs this up. Gartner predicts over 40 percent of agentic AI projects will be canceled by the end of 2027, pointing to unclear value and weak risk controls. Translation: a lot of these systems look busy without being governed.

Forbes names the exact failure with a sharper term. The article describes a capability-deployment verification gap, where an agent works fine in a test but the business cannot verify or trust it once it runs on live systems. That gap is your frozen project card, scaled up.

Status label versus actual status

What you see What might be true
Card says in progress Automation finished and nobody moved it
Workflow shows active Last successful run was weeks ago
Dashboard looks green Output stopped, error was swallowed
Agent is connected Nobody checked what it did today

How do you catch automation running unwatched?

You catch automation running unwatched by forcing a reconciliation, meaning you compare the number of tasks your board claims are in progress against the number where work actually moved in the last seven days, and the gap between those two numbers is your exact measure of how much your system is lying to you. Two numbers. One truth.

Whoa, that sounds simple, and it is. The reason nobody does it is that the board feels authoritative. You built it. You trust it. But trust is not verification, and automation running unwatched survives entirely on the difference between those two things.

  1. Open your project board right now.
  2. Count the cards marked in progress.
  3. For each one, ask: did real work move on this in the last seven days?
  4. Count how many pass that test.
  5. Compare the two numbers.

If the numbers match, your board is honest and you can trust it. If they do not, you just found your operational chaos, and now you can name it instead of feeling it.

The word “automaton” comes from the Greek for “acting of one’s own will.” Fitting, given that the whole problem here is a system acting on its own long after you stopped paying attention to its will.

Almost every time I open a client’s project board, I find at least one column that functions as a graveyard. Cards marked in progress, untouched for weeks, describing automations that either finished quietly or died quietly. The owner always says the same thing: “oh, that’s still running.” Nobody checked. The label held, the work left, and the gap between them grew into a month of decisions made on fiction. The fix is never a fancier tool. It is one recurring human check that reconciles the claim to the output.

How do I know if my automation stopped working?

Check the output, not the status. An automation that stopped often still shows as active or connected. Look at the most recent file, row, or message it was supposed to produce. If the last one is older than the schedule says it should be, the automation stopped and the label never caught up.

Why does my project board say in progress when nothing is happening?

Because status is a manual field in most tools, not a live signal. Someone set the card and moved on. The actual work runs in a separate automated layer that never updates the card. The two drift apart the moment you stop manually reconciling them.

How do I audit automation nobody is watching?

Count two numbers and compare. First, how many tasks your board claims are in progress. Second, how many actually moved in the last seven days. The gap is your audit result. Do this weekly and the gap stays small instead of compounding into a mess.

What makes an automation keep running past when it should stop?

A trigger with no off-switch and no owner. The automation fires on a schedule or event and keeps going because nothing was built to tell it to stop, and nobody was assigned to watch whether it still should. Governance, not the tool, is the missing piece.

Is this a tooling problem or a process problem?

A process problem almost every time. The tools work. What is missing is the recurring human step that reconciles what the board claims against what the system actually produced. Better software without that step just produces confident lies faster.

How often should I reconcile my board?

Weekly for most operations. A seven-day window keeps the gap small enough to fix in minutes. Wait a month and the stale cards pile up until the whole board feels untrustworthy, which defeats the point of having a board at all.

Do not file this under “later.” The whole point was to interrupt your scroll, so interrupt it. Open the board, count the two numbers, and sit with the gap. That gap is the most honest metric in your operation today.

Go look right now. Post the two numbers. I bet they don’t match.

Add Hot Hand Media as a Preferred Source

… and see more of my posts in your Google results.

Your business should not run on chaos.

Tell me what you are trying to fix and I will tell you whether Hot Hand Media is the right fit. No pitch deck. No salesperson voice.