For all the money companies pour into corporate training, the return often disappoints. Employees click through modules, watch explainer videos, and check the completion box. Weeks later, little has changed in how they actually talk to customers, handle objections, or lead difficult conversations. Synthesia, the startup known for its synthetic avatars and voice cloning technology, is now betting that the fix is not better video. It is replacing the video with the conversation itself.
The company is rolling out Roleplay Sessions, the first product in its new Sessions platform. The move signals a sharp strategic turn. Synthesia no longer wants to be the tool that merely generates training content. It aims to become a performance-management layer where employees practice, fail, measure, and improve.
From Watching to Doing
Corporate learning has long struggled with a basic physics problem. Video beats text. It holds attention better and scales cheaply across thousands of workers. But it is still a broadcast medium. The learner absorbs; the content does not adapt. A salesperson can watch a dozen tutorials on negotiating discounts and still freeze when a real prospect pushes back on pricing.
Synthesia built its reputation by making that broadcast medium cheaper and more personalized. Its proprietary AI avatars and voice technology let enterprises spin up training videos without booking studios or hiring actors. That was genuinely useful. It just was not sufficient.
Roleplay Sessions are designed to close the distance between knowledge and action. In these simulations, an employee speaks directly with an AI avatar built to behave like a customer, colleague, or direct report. The avatar does not follow a fixed script. It reacts. It raises objections. It gets frustrated. It changes tone based on what the user says. Someone practicing a performance review might face an avatar that deflects initial feedback, forcing the manager to reframe the conversation. A customer-service trainee could encounter an avatar that escalates from annoyed to angry unless the agent hits specific de-escalation cues.
This matters because soft skills are conversational, not informational. You do not learn them by viewing slides. You learn them by stumbling through the dialogue, feeling the pressure, and adjusting in the moment.
The Feedback Layer
What makes this more than a chatbot with a face is the scoring infrastructure underneath. Synthesia is integrating reasoning intelligence into the product. The platform uses OpenAI’s models to handle the underlying language logic, but that is only the foundation. On top of it, Synthesia has built a proprietary system of rubrics, performance data, and analytics that turns raw conversation into structured coaching.
Here is how the difference shows up in practice. A standard AI conversation might tell a user, "You handled that well." Synthesia’s layer can score the exchange against a company’s own criteria. Did the employee open with empathy? Did they ask a discovery question before pitching? Did they establish next steps? The platform captures these signals and returns granular, actionable feedback. Managers do not get a vague impression of who is "good with people." They get precise maps of capability across an entire sales force or leadership cohort.
Victor Riparbelli, the company’s CEO and co-founder, has emphasized that this data-centric approach is where the real value sits for executives. When hundreds or thousands of employees run through role-play scenarios, the organization accumulates a level of insight that was previously impossible to gather at scale. You can spot which teams struggle with objection handling, which managers rush through feedback, and where skill gaps cluster before they hit quarterly numbers.
Who Is Using It, and Where It Goes Next
The early customer list suggests the pitch is resonating inside large, complex organizations. Synthesia has signed on one of Europe’s top three companies by market cap, a Fortune 100 giant, and a global leader in recruitment. The initial use cases center on high-stakes soft skills for sales and leadership teams, the exact domains where poor conversations carry heavy costs.
Przedsiębiorstwa mogą dostosowywać symulacje, dostarczając platformie własne dokumenty szkoleniowe, transkrypcje rozmów i kontekst. Scenariusze są szyte na miarę, bez konieczności angażowania zespołu inżynierów oprogramowania do budowania ich od zera. Ta równowaga między personalizacją a wdrożeniem gotowego rozwiązania jest kluczowa dla adopcji w działach szkoleń korporacyjnych, które działają ostrożnie.
Synthesia nakreśliła również jasną ścieżkę ekspansji. Firma planuje rozszerzyć Sessions na rozmowy kwalifikacyjne i weryfikację kandydatów. Ostatecznie, gdy koszty wnioskowania spadną, produkt wyjdzie poza segment przedsiębiorstw. Małe firmy, klienci profesjonalni oraz instytucje edukacyjne są uwzględnieni w mapie drogowej. Uniwersyteckie centrum karier mogłoby wykorzystać ten sam silnik, aby pomóc studentom ćwiczyć rozmowy behawioralne. Mała firma konsultingowa mogłaby przeprowadzić swoich partnerów przez ćwiczenia negocjacyjne bez konieczności zatrudniania aktorów.
Szersza zmiana w obszarze AI dla przedsiębiorstw
Ten zwrot warto obserwować, ponieważ oddaje on kierunek, w którym zmierza dojrzałe AI dla przedsiębiorstw. Pierwszą falę szumu wokół generatywnej sztucznej inteligencji zdominowały dema