In 12 minutes we run our own sales training platform end to end. We build a course from documents in Notion, take its AI roleplay as a learner, score a bronze, and open the analytics that show whether behavior changed.
One on one coaching teaches people to do the job but doesn't scale. Watching content scales and teaches almost nobody. Measuring whether either one changed behavior costs enough that most teams skip it and hope. This 12-minute walkthrough runs our own sales training platform on our own product, from building a course out of documents we already had to opening the analytics that say whether any of it worked.
Building a course from documents you already have
We point the course agent at product documents we keep in Notion. It reads them, asks the follow-up questions a subject matter expert would ask, and writes the outline, the lessons, the quizzes, and the diagrams. Everything it produces stays editable, so the agent handles the first draft and a person handles judgment. Once it's published, learners self-enroll or a manager assigns it, and the course view carries the content, the quizzes, and generated diagrams in one place.
Practicing against a simulated buyer
The course includes an AI roleplay. We take it as a learner and score a bronze on the first attempt, which is the useful thing to show, because a demo where everything goes well teaches nobody what the feedback is worth. The grading names each criterion, quotes the moment it was won or lost, and offers the phrasing that would have scored higher. Attempt limits and a required score decide when somebody is finished, so practice has a finish line instead of running until people lose interest.
Seeing which skills a team is short on
Scenario analytics show attempts, best scores, and the lift between a learner's first try and their best one. Skill health aggregates the same evidence across everybody and pulls from scored customer calls as well as practice, so a manager can tell whether objection handling is weak across the team or weak in three people. From a gap, the AI Analyst will build the roleplay or the guide that addresses it.
Where courses, practice, and coaching each fit
Courses carry knowledge that has to land the same way for everyone, like onboarding and compliance. Roleplays carry the conversations that only improve with repetition. Cohort programs run structured engagements where a human coach is part of the design, and certifications record who has been signed off. Each part works on its own, and the walkthrough runs them in sequence because that's how a workspace tends to grow rather than because you need all of it.
Connecting it to the systems you already run
Exec reads from the tools your material already lives in, including Notion, Google Drive, Guru, and uploaded files, so a course is built from current documents rather than a copy somebody pasted last quarter. Call recordings come in from Fireflies, Gong, Recall.ai, or Microsoft Teams. If your company delivers training through an existing LMS, a roleplay can be launched from inside a course there and the score passed back to the gradebook, so learners never manage a second login. Directory sync and single sign-on keep the roster current without anyone maintaining a spreadsheet of who joined and who left.
Common questions
How long does building a course take? Generation runs in minutes from documents you already have. The time you spend goes into reviewing and editing rather than authoring from a blank page.
Do learners need training to use it? No. A learner opens an assignment, takes it, and gets feedback. Setup and configuration sit with whoever administers the workspace.
Can we start with one part instead of all of it? Yes. Courses, roleplays, call scoring, the Knowledge Hub, and coaching each stand alone. Most workspaces start with one and add the next when there's a reason.
What do the analytics actually measure? Score movement between attempts, performance on each evaluation criterion, and skill proficiency across people, sourced from both practice and scored live calls.