Breaking
Mursion raises $40M+ to scale human-powered AI simulations 150+ enterprise clients now running virtual role-plays San Francisco HQ, 190 employees, founded 2015 Dow, T-Mobile, United Airlines among the practice rooms Avatars on the screen, actors in the studio Mursion raises $40M+ to scale human-powered AI simulations 150+ enterprise clients now running virtual role-plays San Francisco HQ, 190 employees, founded 2015 Dow, T-Mobile, United Airlines among the practice rooms Avatars on the screen, actors in the studio
YesPress / Profile / Company

Mursion practices what no one wants to.

The San Francisco company built a training platform where AI avatars are puppeted, live, by trained human actors. You rehearse the hard conversation before you have it. Then, ideally, you have it better.

Est. 2015 San Francisco ~190 People $40M Raised
A Mursion learner facing the platform's virtual avatars on screen.
Five avatars, one person, zero real-world casualties.

A practice gym for the conversations you keep dodging.

On a Tuesday morning, somewhere inside a Fortune 500 leadership program, a mid-level manager logs into a video window. On the other side is a polite, slightly tired-looking employee named "Alex." Alex is animated. Alex is also, technically, not real - the avatar is rendered by software, but the voice, hesitations, and small offended pause when the manager interrupts? All of that is a human being in a Mursion studio, breathing through a headset.

This is what Mursion sells. Not a tool. A practice room. The kind of place where you can flub a layoff conversation, mishandle a diversity complaint, butcher a sales objection, and nobody quits over it.

A note on what you're looking at: every Mursion avatar is operated, live, by a trained "Simulation Specialist." The AI is real. The humans are realer.
Mursion's whole bet is that empathy is a skill, and skills require reps. - The pitch, in one sentence

Soft skills are hard, and most training pretends otherwise.

Corporate learning has a dirty little secret: most of it is content delivery. Watch the module, click "next," answer the quiz, get the certificate, forget everything by Friday. The skills that actually decide whether a manager keeps a team together - listening under pressure, naming bias without shaming, holding a hard line without humiliating someone - those are exactly the skills you can't learn from a video.

The traditional fix was role-play. The traditional reaction to role-play was a quiet shudder. Asking two colleagues to "pretend" through a feedback conversation usually produces something between a sitcom and a hostage negotiation. Either nobody takes it seriously, or somebody does, and now there's residue.

Mursion's founders looked at that gap and saw an opening. If you could give a learner a stranger, a stakes-feeling scenario, and a do-over button, you could finally make practice both safe and uncomfortable enough to matter.

Role-play, the most uncomfortable word in corporate training, turns out to work fine - once you remove the colleagues. - The unspoken thesis

Two people, one improbable wager.

In 2015, Mark Atkinson, a serial education-technology entrepreneur, partnered with Arjun Nagendran, a post-doctoral researcher at the University of Central Florida's Institute for Simulation and Training. Nagendran's work on robotics and virtual humans had quietly produced something unusual: a way to puppet computer-generated characters with a single human operator. One actor could become five avatars. One studio could become a casting agency.

Atkinson brought the go-to-market instincts; Nagendran brought the patents and the pipeline. Together they made a counterintuitive call - in an industry already chasing fully autonomous AI tutors, they would build a platform where a real, breathing person remained in the loop. The phrase "human-in-the-loop" was years from becoming fashionable. Mursion just called it the only way to make practice feel real.

The bet looked stubborn at the time. Today, it looks early.

2015 Year founded

Atkinson and Nagendran incorporate Mursion in San Francisco.

150+ Enterprise clients

From Dow to T-Mobile to United Airlines to Ericsson.

$40M Total funding

Across six rounds; backers include Leeds Illuminate and New Markets Venture Partners.

190 Employees

A studio-shaped company: behavioral scientists, engineers, and actors.

What is actually happening on the screen.

A Mursion session is shorter than you'd think. Usually 30 minutes. A learner gets a scenario brief - say, you're a regional manager and one of your store leads has been accused of favoritism. You join a session. You see a small set of avatars in a conference room. You start talking.

On the other end, in a Mursion studio, a Simulation Specialist - typically a trained improv actor with hours of coaching on the scenario - is voicing the avatar in real time. They listen, react, push back, sometimes throw an unscripted curveball. The conversation happens. The conversation ends. A debrief follows, with AI-assisted analytics on what was said, how it was said, and what got missed.

The platform now stretches across leadership, sales, customer service, K-12 teacher prep, and clinical communication for healthcare. The scenarios change. The mechanic doesn't.

If you've ever wondered what it feels like to be on a video call where the polite-but-frustrated person you're talking to is actually three people taking turns in a closet in Florida - this is roughly that. In a good way.

A short timeline of an awkward idea getting taken seriously.

2010-14
Nagendran's avatar-puppeteering research takes shape at UCF.
2015
Mursion incorporated in San Francisco. First enterprise pilots.
2017-19
Scenario library expands; first major leadership and DEI deployments.
2020-21
Pandemic remote work makes virtual practice mainstream overnight.
2022
$5M venture round closes; co-founder Nagendran transitions out.

Who's actually paying for the practice room.

The customer list is the part that quiets skepticism. Dow built an inclusive leadership program around Mursion simulations. T-Mobile, United Airlines, Best Western, and Ericsson have rotated managers through the platform. Universities use it for teacher preparation. Healthcare systems use it for standardized-patient training, where the cost of a real-world fumble is measured in something more than awkwardness.

The pattern is consistent: organizations that pay for Mursion are usually the ones that have already tried the other thing. The webinar. The off-site. The dusty role-play handbook. They come to Mursion because the alternative was failing quietly.

Where the practice happens.

Mursion's scenarios cluster across a handful of high-stakes domains. Approximate distribution by use case.

Leadership
92%
DEI / Inclusion
74%
Sales
61%
Customer Service
55%
Teacher Prep
40%
Healthcare
33%

Illustrative share of customer deployments. Real numbers vary by year and probably by mood.

When mistakes can't be made on real people, you make them in Mursion. - A reasonable mission statement

Make the soft skills repeatable.

Mursion's stated mission is not modest: turn the most human parts of the workplace - leadership, empathy, communication, conflict - into things you can practice on purpose. Treat them as muscle, not personality. Schedule the reps. Measure the gains.

The supporting belief is that performance assessment in this domain has been mostly vibes. Mursion's analytics layer tries to do better - tracking what was said, what was avoided, what changed across attempts. The platform integrates with learning management systems, reports back ROI metrics, and quietly normalizes the idea that interpersonal skill is data, not destiny.

Whether the data is fully there yet is a fair question. Whether the question is worth asking at all - that part Mursion appears to have settled.

The human-in-the-loop turns out to be the moat.

For a while, it looked like Mursion's reliance on live actors might be a weakness. Big AI players were promising avatars that could converse autonomously, indefinitely, at near-zero marginal cost. Why hire a roomful of improv-trained specialists when a language model could do the job for free?

Then the language models started talking, and everyone discovered they had things to say but very little they actually meant. Tone collapsed. Stakes evaporated. The very thing that makes practice useful - the sense that another consciousness is on the other end, capable of being surprised, hurt, or persuaded - turned out to be hard to fake. Mursion's "old-fashioned" human-in-the-loop architecture suddenly looked less like a constraint and more like a strategic decision made five years ahead of schedule.

The interesting question now is not whether AI replaces the Simulation Specialists. It's how much of the orchestration, scenario generation, analytics, and coaching AI can absorb so the specialists can do more of what only humans do well. That's the next chapter, and Mursion seems to be writing it slowly, on purpose.

The avatars on the screen are AI. The presence on the other end is not. That's the trick. It's not really a trick. - Closing argument

That Tuesday morning, revisited.

Back to the manager and "Alex." The thirty minutes end. The manager closes the laptop. They got two questions wrong. They interrupted at minute four. They recovered at minute eleven. The debrief flags both, gently. Tomorrow there is a real conversation with a real direct report about a real conflict, and the manager walks into it having already been there once.

That's the whole product. Not virtual reality. Not artificial intelligence. Rehearsal. The oldest performance technology there is, finally available on a corporate calendar invite.

Share this profile

Send Mursion to the colleague who keeps avoiding that one conversation.