The second service
One screen across every system — and who actually logs into it
A command center is one oversight screen over every connected system: the live state of the operation, the exception queue holding every decision the layer refused to make alone, every decision it did make with the rule it followed and the source it believed, and controls over what it may act on without a person.
What the command center is
Your people keep working in the tools they already use. The command center is for the ones who need to see across all of them — the live state of the operation, the exceptions the layer refused to decide alone, and the reasoning behind every decision it did make. Not another queue for everyone else to live in.
Spelled command center throughout, including for readers who would write command centre. It is a product name, and forking it splits the thing being named.
Why there is a screen at all, when the layer acts inside your tools
Because the two jobs are different. The layer acts inside the tools your people already have open, which is where the work happens and where a result has to land to be acted on. Oversight is a different job, with different users.
So this is a screen somebody logs into, and it would be dishonest to say otherwise. What it is not is a queue the whole company lives in.
Live state, joined across every connected system
Live state of the operation, joined across every connected system.
1 screen showing the whole operation — joined rather than tabbed, which is what an operations lead is usually asking for when they ask for one pane across ERP, CRM and ITSM.
What is an exception queue in automation?
An exception queue holds every decision the layer refused to make alone. Anything outside the tolerance your team set — a price mismatch, an unmatched line, a missing document — waits there for a person rather than being approved automatically. It sits in the command center, beside the live state and the audit trail.
What decides whether something waits there is a written tolerance, not a score:
Anything outside tolerance goes to escalation-queue. It is never auto-approved, regardless of confidence score. An operator decides.
ops/vendor-onboarding.md
The tolerance rules that decide what escalates — unit price ±2%, quantity exact, freight separate — are published in full.
What is human-in-the-loop automation?
The exception queue — everything the layer would not decide alone.
The phrase is defined everywhere and implemented almost nowhere. The implementation is the sentence above: a confidence score does not buy an approval, and no threshold quietly lets one through.
The audit trail: every decision, the rule it followed, the source it believed
Every decision, with the rule it followed and the source it believed.
Two things get recorded that most systems do not keep: which source was believed and why, and who changed the rule afterwards. The field list is on audit trail and autonomy controls for automated decisions.
Autonomy controls: what it may act on, what always waits for a person
Controls: what it may act on, and what always waits for a person.
Set per workflow before anything runs, and visible in the same screen as the decisions they govern. A control living in a configuration file nobody in operations can open is not a control anybody is exercising.
No named tier framework is published here yet. Tiers describe how engagements are actually configured, and that description has to be written and signed off first — inventing a ladder of autonomy levels to fill a heading is inventing product capability.
Correcting the layer from the command center
Correctable. A reviewer edits the rule. The next run behaves differently. No redeploy, and the change is attributable. The rule itself is the markdown note an operator edits; the mechanism is the intelligence layer that acts inside your existing tools.
Questions
Straight answers.
What is an exception queue in automation?
An exception queue holds every decision the layer refused to make alone. Anything outside the tolerance your team set — a price mismatch, an unmatched line, a missing document — waits there for a person rather than being approved automatically. It sits in the command center, beside the live state and the audit trail.
What does an AI command center do?
It shows live state across every connected system, the exception queue of everything the layer refused to decide alone, every decision it did make with the rule and source behind it, and the controls governing what it may act on without a person.
What is human-in-the-loop automation?
Automation that stops and asks. In practice: a written tolerance decides what escalates and a confidence score never overrides it, so anything outside the band waits for an operator rather than being approved because the system felt sure enough.
How do you keep a human in control of automated decisions?
Three things together — written tolerances deciding what the system may act on, an exception queue where everything outside them waits for a person, and an audit trail recording the rule followed and the source believed. Controls are set per workflow before anything runs.
What is a command center for operations?
One oversight screen over every connected system, for the people who need to see across all of them at once: live state, the exception queue, the audit trail and the autonomy controls. Everyone else keeps working in the tools they already use.
Is the command center something everyone in the company has to use?
No. It is for oversight, not for doing the work. Your people keep working in the tools they already have open, and the layer writes results back into those. The command center is for the few who need the whole operation on one screen.
Start here
Bring us one workflow.
Tell us the process that crosses the most systems. You get a scope, a measure and a delivery plan back — and a straight answer if we think it is not worth building.
Get in touch