Corporate L&D
Your people practice the hardest conversations for the first time on your customers.
We build those conversations as simulations against a counterpart designed to resist. Everyone gets unlimited reps. Every session leaves a transcript scored against your rubric.
Research under review at the International Journal of Advanced Corporate Learning.
Session transcript — turn 7
Counterpart
That's the third time you've said "alignment." Tell me which of my two teams loses headcount, or let's stop.
Participant
Neither, today. I can't give you the number until Friday and I'm not going to invent one now.
Pressure point
Held the boundary without hedging. Scored 4/4 on candor, 2/4 on acknowledging cost to the other party.
Three conversations your people are avoiding.
Read these to a manager and watch which one they flinch at. That flinch is your training needs analysis. Each runs as a simulation built from your material, your language, and the standard you actually hold people to.
Managing people
The employee who won't hear it
Every specific you raise, she reframes as a resourcing problem. Name the pattern without litigating each instance.
Revenue
Procurement's last-minute squeeze
A tactic runs underneath the friendliness. See it, name it, and hold your price without blowing up the deal.
Change & presence
The team that's seen this before
Third system in five years. She isn't angry, she's tired, and she has receipts. Enthusiasm makes it worse.
These are the three managers flinch at most. The full library runs to twelve — from the promotion you have to deny, to the renewal that's already lost, to the comment in the meeting. Don't see yours? Every simulation is built from your situations.
See all twelveYou already run role-play. Here's what changes.
Role-play works. The problem was never the method — it's that a room, a facilitator and a willing colleague don't scale, and nothing survives the session except an impression.
| Role-play today | With a Friction Bot | |
|---|---|---|
| The counterpart | A colleague who doesn't want to embarrass anyone, playing difficult for ten minutes. | A character with a position, a temper and something to win. It doesn't soften because you're struggling. |
| Who gets reps | Whoever is in the room, on the day you could get everyone in the room. | Everyone. As many times as they want, without booking anything. |
| Consistency | Varies with who plays the part, how well they know it, and how late in the day it is. | Same pressure and same standard for the first participant and the four-hundredth. |
| What's left afterward | A facilitator's impression, a smile sheet, and whatever people remember. | The full transcript, every turn kept, reviewable months later by anyone you choose. |
| What it costs you | A facilitator's day, a room, and the calendars of everyone in it. | As much or as little of your team's time as you want to spend. |
Hands off
Let the Session Scorecard do it
Every session comes back scored against your rubric, dimension by dimension, with the reasoning attached. Nobody on your team has to read a transcript for this to work.
Hands on
Go straight to the Pressure Points
Every session is marked where it turned — the deflection, the pivot, the moment someone lost composure and the moment they got it back. Jump to turn seven instead of reading forty.
Where live role-play still wins
A skilled facilitator reading a room, body language, the weight of being watched by peers, and a debrief that adapts in the moment. We don't replace that day — we give everyone reps in between, and hand your facilitator a transcript to run the debrief from.
What comes back
Every session returns a scorecard with the reasoning attached.
Not a completion record and not a satisfaction score. Each attempt is graded against your rubric, dimension by dimension, with the evidence from the transcript that produced the mark. Below is a real assessment from a B2B sales scenario.
Session Scorecard
60/100
Handling objections
18/30
Identified the core objections — cost from Linda, reliability from Greg. Struggled to produce the math Linda demanded, and fell back on generic AI buzzwords when Greg pressed for specifics on error types.
Specificity and grounding
20/40
Used vague language — “constantly evolving,” “bulk of the work” — instead of the numbers the characters had already supplied: labor rates, report volumes. Never justified the $70k gap with an ROI calculation.
Stakeholder management
22/30
Looped Tony into the conversation to explain the workflow, which worked. Allowed Linda and Greg to form a united front of skepticism by not addressing the hidden costs and training concerns they raised.
Understands the basic sales mechanics but relies too heavily on marketing jargon, failing to provide the granular data and operational reassurance required to close a high-stakes industrial sale.
30-day pilot
Thirty days. One conversation your people keep getting wrong.
A scenario built from your own material, a cohort of twenty-five, and a readout of what they actually did under pressure. Not a login and a trial period.
Week 0
Build
You send materials. We build the counterpart, the scenario and the rubric around them.
Weeks 1–3
Run
Twenty-five people, unlimited attempts, on their own time. Nothing to schedule.
Week 4
Readout
Sixty minutes. What your cohort did, where they broke, and the transcripts worth reading.
Day 30
Decide
Continue on an annual per-seat license, or stop. Nothing auto-renews.
What you get
- One scenario built from your material — your language, your standard, your people's actual situation.
- Twenty-five participants, unlimited attempts each.
- A Session Scorecard for every attempt, scored against your rubric with the reasoning attached.
- Pressure Points marking where each conversation turned — and every transcript, kept in full.
- A sixty-minute readout of what happened across the cohort, not a usage dashboard.
What we need from you
- Your learning objectives and curriculum outline.
- Existing materials for the skill you're targeting.
- A plain description of the problem you're trying to solve.
- Twenty-five participant names.
- About two hours of one person's time, across the whole thirty days.
What counts as working — agreed before we start
In week 0 we write down what you expect to see. At the readout we hold the cohort's actual behavior against it. If it didn't do what we said it would, we will tell you that plainly — a pilot nobody can evaluate is worse than no pilot.
Which conversation would your managers flinch at?
Tell us that one. We'll build it, run it with twenty-five people, and show you what they actually did. Fixed fee, credited toward year one.
Pricing on request · nothing auto-renews
Questions we get
Before you ask.
We already run role-play workshops. Does this replace them?
No, and we would not try. A facilitator reading a room and adapting a debrief live is the hardest thing in training to replicate. What a workshop cannot do is give four hundred people fifteen reps between sessions, or hand you a transcript six months later. The honest question is how many of your managers actually practiced at the last one, and how many watched.
Can't we just prompt ChatGPT to be a difficult stakeholder?
You can, for about four turns. General-purpose assistants are trained toward helpfulness and drift back into agreement as soon as the user struggles — which is exactly the moment the exercise starts to matter. The engineering is not the persona. It is holding the resistance for the length of a real conversation and scoring what happened against your standard.
How do we know it works?
Partly, and we will not overstate it. We can show that participants rate the exercise as worthwhile and that subject-matter evaluators agree with the assessments — that is what is in the paper under review. What nobody in this category has is a longitudinal study tying simulated practice to behavior change years out. Anyone telling you otherwise is selling. Tell us what evidence you would need to justify a pilot and we will tell you honestly whether we can produce it.
Is it voice or text?
Text today. People ask for voice and we understand why, but voice is harder to build, harder to evaluate reliably, and a meaningful number of participants find synthetic voice unsettling at exactly the moment the exercise needs them present. We would rather ship the better instrument than the more impressive demo.
Do you integrate with our LMS?
A 30-day pilot does not need to. Participants get a link; nothing has to be installed or provisioned. If you continue past the pilot, tell us what system the records need to land in and we will scope it then rather than promise it now.
What does it cost?
The pilot is a fixed fee, credited toward year one, and it converts to an annual per-seat license only if you decide to continue. Nothing auto-renews. Pricing depends on cohort size and how many scenarios you want built — ask and we will quote it in the first conversation, not the fourth.
Data and consent
The transcript is the product, so who sees it matters.
We are asking your people to be bad at something on the record. That only works if everyone knows the rules before they start.
Participants know what is recorded
Before the first session, every participant is told that the full transcript is kept, who will read it, and what it will be used for. No silent scoring.
You decide who can read what
Access is set at the start of the pilot — participant only, participant and their manager, or the L&D team in aggregate. We agree it in week 0 and it does not change mid-pilot.
Your material stays yours
The scenarios, rubrics and transcripts built from your material are yours. We do not use them to train models or to build scenarios for anyone else.
Deletion on request
If you stop at day 30, you can have the transcripts exported and the data deleted. Tell us the retention window you need and we will hold to it.
Using simulations for evaluation rather than practice raises a different set of questions — about consent, about fairness, and about what a score should and should not decide. We will not sell you an evaluation deployment without having that conversation first.
