Concept trainers: the deck nobody reads, as a conversation instead

Update, October 2026: AI Tutors are now called AI Trainers. This post uses the new name.

‍

‍

An enablement lead has a 40-slide deck on the new plan structure. She sent it to 60 people in March. In June, six of them can explain the difference between a copay and coinsurance, and the escalation queue is full of customers who were told the wrong one. The deck was accurate. Nobody learned from it.

‍

Concept is one of three AI Trainer types in Outdoo, and it is the one for material a person has to understand before they can do anything with it. You attach the document, and your team learns it by being asked about it.

‍

Why this matters

Most training content fails at the same point. It explains something once, in writing, to someone who is not currently trying to use it. Reading is a poor test of understanding, because a reader cannot be wrong out loud. They finish the deck, feel informed, and discover on a live call that they had two similar terms confused the whole time.

‍

A Concept trainer inverts that. It knows the material, and it asks the learner to explain it back. When the answer is wrong, it says so and corrects it, which is the one thing a PDF cannot do. Nobody has to guess whether the content landed, because the session is a record of whether it did.

‍

This is the same argument behind everything else in Outdoo. Practice built from your own material, not from generic scenarios, and coaching that closes the loop instead of ending at delivery.

‍

How it works

Open Roleplay, then AI Trainer, then Create AI Trainer. Set the Type to Concept, give it a title that names the subject rather than the format, and attach your document under Resources. At least one resource is required, because that document is everything the trainer knows.

‍

Resources accepts PDF, DOC, DOCX, PPT and PPTX. You can attach more than one, and pairing works well: a slide deck carries the claims, a written explainer carries the reasoning, and the trainer teaches from both. Our own security overview is built this way, with the deck alongside a document explaining what SOC 2 and AES-256 actually mean.

‍

Then write Roleplay Instructions, up to 500 characters, shown to the learner just before the session starts. The instructions that work name the behaviour you want. "Explain each term in one plain sentence, then walk the arithmetic in the order the plan applies it" gets you a better session than "this covers plan basics."

‍

Persona is who the learner meets: name, gender, language, voice, avatar, and the opening line. Save, then start the session from the AI Trainer list. Preview shows you the session screen but does not run it, so save first and test from the list.

‍

What this unlocks

An enablement lead who just received the new pricing bands can have a trainer running on them the same afternoon. Attach the document, write two lines, pick a voice, save. The people who need it practise explaining the bands out loud before they quote them to anyone.

‍

A compliance officer with a new SOP has a way to confirm it was understood rather than distributed. The session shows who talked it through and what they got wrong, which is a different kind of evidence from a read receipt.

‍

A support manager onboarding a winter intake can build one trainer per product line and let new starters work through them in their first week, in whatever order suits them, without booking anyone's afternoon.

‍

The bigger picture

Understanding is not a smaller version of practice, it is a prerequisite for it. Putting someone into a roleplay on material they were never actually taught measures how well they improvise, and then files the result as a skill gap.

‍

The deeper problem is that our industry has spent twenty years measuring delivery because understanding was too expensive to measure. Completion rates, attendance, time in module. All of them describe what was sent, not what was learned, and every enablement team knows the difference and reports the number anyway.

‍

What changes when a trainer can ask someone to explain the material back is that understanding becomes observable. That is useful, and it is also uncomfortable, because the first thing it tends to show is how much of what a company already calls training never landed. Teams that look at that honestly usually find the problem was never the content. It was that nobody was ever asked to say it out loud.