New people practice on an AI customer, not on your pipeline.

A new rep practices in Spellit Academy by holding a full spoken sales conversation with an AI customer in the browser before any real customer is on the line. The AI customer opens the call, the rep answers out loud, and the customer replies in a synthesized voice. Reps can talk hands-free, hold a button while they speak, or type a line if speaking is not an option at that moment.

The AI customer can end the call if the conversation goes badly, as a real buyer would. When the rep presses Finish, a separate AI model scores the conversation, which usually takes 20 to 30 seconds. The breakdown quotes the rep's own lines and offers a better phrase for each weak spot. The rep can then run the same scenario again straight away, as many times as needed.

The AI customer in Spellit Academy is built from a scenario your team writes for sales onboarding: who the customer is, what they are after, how they behave and what a successful call looks like. Your methodologist edits these scenarios and the scoring rubric in the Academy admin area, so the customer raises the objections your market raises.

The instructions for playing the customer stay on the server. A trainee sees the customer's name, position and company, and nothing about what makes this customer open up or shut down, or how the call is scored. A rep who can read the answer key is no longer practicing, which is why the product hides it. If a scenario is missing or unpublished, the role-play does not start, so the AI never falls back to playing a generic chat partner.

Spellit Academy's script trainer removes the prompts one step at a time, so a new rep moves from reading a stage of the call to delivering it without help. Each stage has a Learn mode with a sliding hint: full text first, then anchors only, then just the goal, then nothing at all.

After that comes Practice, where the rep picks the best reaction to a customer line and gets feedback immediately. Stages that have their own scenario end with a voice role-play scored on that stage's rubric. A call map shows every stage as mastered, in progress or not started, with test and role-play results on the milestone stations. When the readiness card on a trainee's dashboard shows a weak area, its Practice this link opens that exact stage in Learn mode. The rep spends time on the step that is failing instead of replaying the whole course.

In Spellit Academy, a manager judges a new hire's readiness from that person's own test answers and scored role-plays, not from a course completion rate. The Users table lists every trainee with score, modules, tests, exam, trainings, diplomas, last sign-in and activity over 30 days, and a filter shows who has not passed the exam yet.

Opening a trainee's card shows the full breakdown of every role-play in the same form the trainee saw, with quotes and suggested phrases. Results also add up to a Client skills profile across five call stages: contact, discovery, presentation, objections and closing. By default, each stage score is the average of the last three scored role-plays, and a stage with no data shows as empty rather than as zero. From the same admin area, a manager can assign materials to a group with a deadline through a learning plan.

No real customer is involved while new reps practice in Spellit Academy, because every onboarding role-play runs against an AI customer in the browser. Academy does not listen to live sales calls or score them. A rep can fumble the hard objection twelve times at eleven at night, and nobody outside your team hears it.

Real conversations can still feed the material. An administrator can add call transcripts as source content when building a course, so the practice reflects how your buyers talk. Reviewing live calls is the job of Spellit Actions, which shows what a team needs to practice based on its real calls. Turning a gap found there into an Academy scenario is a step your team takes itself. The rep's first live customer is then no longer their first rehearsal.

A Spellit Academy role-play counts as passed when the scored conversation reaches your workspace threshold, which is 70 out of 100 by default. The score comes from a weighted rubric that your methodologist can edit, and the editor checks that the weights add up.

For each line of the rubric, the scoring model has to quote what the trainee said, explain what worked or failed, and give an example of a better phrase. Anti-patterns cut the score by two points each, up to six points in total. In the default sales rubric they include interrupting, long monologues, internal jargon, pressure and not closing. Tests and the final exam have their own thresholds, both 80% by default. All thresholds are set per workspace, so a team can raise the bar for live calls without rewriting content. A passed final exam unlocks a PDF diploma with a unique number.