Literacy without practice is a slide deck

A course made of slides can teach people the words: depth, held, refused, withheld, provenance. It cannot teach them what to do when a held item lands in their queue and the customer is waiting. AI literacy that never puts people in front of a working agent is a slide deck with a quiz attached, and it fades within weeks.

What goes wrong with slide-based literacy?

Most AI literacy programmes follow a familiar pattern. A set of slides or videos explains what AI is, what agents do, why governance matters. A short multiple-choice test checks that people were paying attention. A completion certificate is issued. Everyone is counted as literate.

The problem is not that the content is wrong. It is that knowing a definition and recognising the thing in front of you are different skills. A person can correctly define "withheld" on a quiz and still, faced with a withheld action in a live workspace, log it as a fault and look for a workaround. They have the word but not the reflex.

Slides also flatten the hard parts. On a slide, every held item has a clear reason. In a real queue, the reason is visible but has to be read, the source of a key fact has to be opened, and the person has to decide while three other items wait. That experience is the skill, and slides do not provide it.

What does practice actually mean here?

Practice is making decisions in a working system and seeing what happens. For people who will work with governed agents, that means:

• Working in a real workspace, or a free build workspace that behaves like one, with agents producing actual verdicts.
• Reading held items and deciding, with the reason and sources in front of them.
• Meeting a refusal and following the correct route instead of working around it.
• Seeing a withheld action and understanding that it is blocked at the platform floor for every agent, by design.
• Correcting a draft and seeing how that correction is recorded.
• Reading the record afterwards, to see their own decision filed with their name.

None of this needs to be high-stakes. Labs can run in a sandbox with sample data. What matters is that the person's hands are on the controls.

How does this look in specific sectors?

Take a clinic practice manager whose team is adopting a booking and billing agent. Slides can explain that the agent will hold appointment changes for patients flagged as needing extra support. Practice shows the manager what that hold looks like, what information it carries, and how long a held change can sit before a patient is affected. Without that, the first time the manager meets a real hold is with a real patient. The question of whether clinic practice managers are accountable for an agent's mistakes gets much harder to answer well if their only preparation was a deck.

In wealth management, literacy for staff who oversee client-facing agents is about recognising when a draft message strays toward advice the agent is not permitted to give. AI literacy for agents in wealth management sets out what that involves. It is a judgement that only develops by reading real drafts.

In the public sector, where fairness and transparency duties are heavier, staff need to practise explaining an agent-assisted decision to a member of the public; AI literacy for agents in the public sector covers that context.

Does practice matter for senior people too?

Yes, though the practice looks different. A board member does not need to review held items, but they do benefit from seeing a workspace, watching an agent refuse something, and reading a verdict record. A director who has done that asks better questions than one who has only seen a summary. The piece on what AI literacy boards need makes the case that even oversight literacy improves with a short hands-on session.

Does practice need to be repeated?

Yes, because the system moves. The agent's own certificate lapses after 90 days and must be renewed; its depth on a capability can change; new verdict types and sources appear. A person who practised at launch and never again is working from a memory of a system that no longer quite exists. Building a short practical refresh into each recertification cycle, as described in preparing clinic practice managers for 90-day recertification, keeps the reflex current.

What is the trade-off?

Practice is more expensive than slides. It needs a workspace, realistic scenarios, and time for people to work through them. It takes longer to complete, and it is harder to deliver to large numbers at once.

There is also some risk. A sandbox that behaves differently from production teaches the wrong reflexes, and labs built on unrealistic data can make the real thing feel unfamiliar.

The answer is to keep practice proportionate. Not every member of staff needs a two-week practicum; most need an hour or two of hands-on labs alongside the reading. Reviewers and approvers need much more. And labs should run in an environment that behaves like the real product, set in the learner's own industry, so the reflexes transfer.

The alternative, slides alone, looks cheaper but delivers almost nothing that survives contact with a real queue.

What should you do differently?

Audit your current AI literacy programme for practice. If every module ends with a quiz and none ends with a lab, change that. Make sure people meet each verdict type in a working system before they meet it at work. For approvers, require observed practice on real decisions before granting rights. Refresh with hands-on work, not just new slides. The guide to AI literacy and certification for people who work with agents explains how literacy and certification connect.

Where the Academy fits

Every Pryme Intelligence Academy lab runs in a free Build workspace, so learners act on a working system rather than reading about one. Lessons, labs and tests are free, and lessons can be read or listened to as audio. For reviewers, the Operator certificate goes further, with a two-week practicum of real review decisions signed off by a Pryme assessor.