Learn them one at a time, each tied to the question it asks of you. A governed workspace's Overview shows six numbers: decisions taken, refusals, waiting on you, depth in use, certified and ledger height. Reading them means knowing what normal looks like, what a change means and what action it prompts, then practising on real or simulated mornings.
What are the six numbers, and what does each ask of you?
| Number | What it shows | What it asks of you |
|---|---|---|
| Decisions taken | Every verdict the decision chain reached in the period | What drove a jump or a fall in volume? |
| Refusals | Actions the chain stopped, each naming the bound that decided | Which bound is doing the refusing, and did it change? |
| Waiting on you | Held actions waiting for a named person, with the oldest item's age | Is the queue growing, or is something stuck? |
| Depth in use | How far agents are acting, capability by capability | Which capability moved, and why? |
| Certified | Agent certificates currently in force | Is anything about to lapse, and is recertification planned? |
| Ledger height | Entries in the append-only record | Is it growing in step with decisions taken? |
The last column is the point. A number on its own is trivia; the question it prompts is the skill.
How do you learn to read them?
- Learn what each one counts. Start with the table above until you can say it without looking.
- Learn the bounds behind refusals. Every refusal names one of five bounds: role scope, certificate, sources, the business's own limits or the platform guardrails. A shift in which bound dominates matters more than the total.
- Learn the depth scale. Depth runs from 1 (read) through 2 (draft for a person to send), 3 (routine work within bounds), 4 (moving money or hard-to-undo actions) to 5 (coordinating other agents). A fall in depth in use usually means a certificate changed.
- Connect numbers to records. For any number that moves, open the underlying decisions. See how to learn to read an AI agent decision record.
- Practise on scenarios. Read a set of figures and say what happened before checking.
What does a reading look like in practice?
Imagine this Monday Overview. Waiting on you has doubled overnight, but the oldest held item is only two hours old. Refusals are flat. Depth in use for refunds has dropped from 4 to 1. Certified is one lower than on Friday. Decisions taken and ledger height have risen together.
The reading: the refunds agent's certificate lapsed or a run failed, which drops an agent to depth 1 the same day. Refunds it used to issue are now arriving as held items. Nothing is stuck, since the oldest item is young, and the record is intact, since ledger height kept pace. The action: tell the team lead, move a reviewer onto refunds and confirm when recertification will run.
Someone who has practised says that in a minute. Someone who has not spends the morning asking whether the agent is broken.
What mistakes do beginners make?
- Reacting to totals. Refusals rising may mean the business tightened a limit yesterday and the agent is doing exactly what it should.
- Ignoring the oldest item. A steady count with a rising oldest age means something is stuck, which is worse than a busy queue.
- Treating withheld actions as defects. Withheld is a platform decision, applied to every agent at every depth. The useful question is why people keep requesting that action.
- Skipping ledger height. If decisions rise and the ledger does not, something is not being filed. That is a question for the platform owner the same morning.
The same reading, taught as a team routine, is in how to train a team to read the six Overview numbers.
Where the Academy fits
Pryme Intelligence Academy covers this in the Operator module on reading the workspace, with labs in a free Build workspace where you can watch the numbers move as you act. Module tests are set in your own industry and role, so the cases look like work you already know.
Questions people ask
Do I need to read the numbers if I am only a reviewer?
Yes, at least "waiting on you" and "depth in use". They tell you why your queue looks the way it does, and an interviewer may ask you to read a sample Overview.
Should I learn the numbers before learning to review?
Learn the basics of review first; it is what you will do most. See what you should learn first about AI agents.
How long does it take to read them confidently?
A few weeks of daily practice is a reasonable expectation. See how long it takes to learn AI agent skills, and the guide to AI skills for job seekers for the wider skill set.