What it is
Second Nature is a dedicated AI role-play platform. A rep talks to an AI buyer persona, the platform scores the conversation straight after on messaging, product knowledge and delivery, and managers track who has passed which certification. Personas are built from a scenario description or from uploaded material such as a script, a deck or a playbook. Around that loop sit course authoring, certification tracking and manager dashboards. Sessions run as text, voice or, since July 2026, a video avatar.
The company is based in Tel Aviv and was founded by Alon Shalita, formerly a lead engineer at Facebook, and Ariel Hitron, formerly of Kaltura. It raised a $22M Series B on 16 October 2025, led by Sienna VC with Bright Pixel, StageOne Ventures, Cardumen Capital, Signals VC and Zoom. That round took total funding to $38M. Zoom is an investor and a customer. The Series B release names Oracle, Adobe, Teleperformance, Check Point and GoHealth as customers alongside Zoom.
It is the incumbent of the dedicated role-play category. It holds G2 Winter 2026 Leader and Momentum Leader badges, sits at about 4.7/5 across roughly 400 G2 reviews, and was a 2026 Webby honoree in Sales and Revenue AI.
Where a dedicated role-play vendor fits
The question this page answers: Hyperbound is the AI-native upstart and Mindtickle is the suite, so where does Second Nature fit?
It fits where the buyer is training, not RevOps. Its customer list leans toward large sales-and-service floors: a contact-center outsourcer, an insurance distributor, enterprise software. The product reflects that. It supports 30+ languages according to its product page, up from the 20+ cited at the Series B. It connects to the LMS systems L&D teams already run, with Cornerstone, Docebo, SuccessFactors and Adobe Learning Manager plus SCORM, and it handles identity through SAML SSO, SCIM and Microsoft Entra ID. The same platform trains service agents, CSMs and sellers, so one contract can cover every customer-facing role.
What it is not is a system for real calls. Hyperbound and Mindtickle both score recorded customer calls. Second Nature scores practice. If your mandate is visibility into live deals, you are looking at the wrong category.
Pricing reality
Quote-only. There is no public rate card, and Capterra lists it as contact-the-vendor. Three things are known from outside the company, none confirmed by it:
- Third-party trackers put it near $30-40 per seat per month with a floor around $20K a year. That band appears in no vendor source.
- Billing is described as monthly-active-user credits. A credit is consumed for each month a licensed user is active, so 150 users active across two months uses 300 credits. Reviews say a credit costs the same whether the user runs one session or thirty. This description comes from competitors’ review pages, so confirm it on the order form.
- Video avatars came with a trial allowance. The July 2026 launch post says every customer gets 6,000 free minutes.
Worked estimate, marked as one: 150 reps at $35 a month for 12 active months is about $63,000 a year. The same 150 reps active for only a four-month onboarding window is about $21,000, which lands near the reported floor. If the credit model holds, the contract rewards bursty cohort training and charges full price for always-on practice. Hyperbound and Mindtickle charge per licensed user whether or not that user practices.
Best for
An enablement or L&D leader at a 200+ seat sales or service organization that trains in several languages, where the job is certifying every new hire on the same scenarios and reporting completion into an existing LMS. The scoped case where it wins outright: a global onboarding program for sellers and service agents across multiple countries, run by a training team rather than by sales managers.
Not for
Teams under 30 reps, where the reported floor buys more governance than anyone will use. Also the wrong pick for RevOps teams whose goal is coaching on real deals, and for teams already paying for a suite with role-play included.
Versus the alternatives
- Mindtickle — the best-funded independent readiness suite, at about $281M raised. Role-play is one of seven agents in ElevateOS, launched on 21 April 2026, next to certification, call scoring and a readiness index. Pick it when practice scores and real-call scores must feed one number. Pick Second Nature when role-play is the whole purchase and the suite’s call-recording layer would duplicate Gong. See Mindtickle.
- Allego — the other large suite, built video-coaching-first. Pick it for asynchronous video practice and peer review. Pick Second Nature for live, spoken simulation at volume. See Allego.
- Hyperbound — the fastest-growing entrant, launched in January 2024 with an $18.3M total raise. Pick it when RevOps owns the program, the scorecard has to match your sales methodology, and Practice has to connect to real-call scoring. Pick Second Nature when the rollout spans service roles and languages. See Hyperbound.
- The bundled options — Zoom Revenue Accelerator shipped Sales Roleplay on 22 July 2026, paid in Premium AI credits, and Gong announced Dry Run for July 2026. Pick a bundle when pre-call rehearsal is the need and nobody runs a certification program.
Watch-outs
- Its own investor now sells the feature. Zoom backed the Series B and then shipped Sales Roleplay inside Revenue Accelerator nine months later. Gong’s Dry Run was slated for the same month. The category’s standalone budget is being absorbed by recorders you may already pay for. Guard: before signing, price role-play from Zoom or Gong at your seat count and write down what Second Nature does that they do not: languages, LMS reporting, service roles. If the list is empty, buy the bundle.
- Reviewers call the personas robotic. The complaint recurs across G2, Capterra and Software Advice, and one G2 review titled “This Platform needs an overhaul” described the voice as unresponsive. Scripted paths bend less than you expect when a rep goes off-script. Guard: in the pilot, have three of your strongest reps try to break the persona with unscripted objections, and score realism before you score anything else.
- The avatar is newer than the product. Video avatars launched English-only and single-persona, with multi-persona and scripted role-plays still to come. Guard: if non-English or multi-stakeholder video is why you are buying, require it working in the trial, not on a slide.
- The credit meter changes what you pay. If billing counts active users per month, an always-on practice program costs roughly three times a four-month onboarding window. Guard: get the credit definition and the per-credit price in writing, and model the invoice against your real practice schedule before you agree to a seat count.
- The outcome numbers are vendor-reported. The Series B release cites sales up more than 20% after about 30 minutes of practice per trainee. Its AppExchange listing claims a 28% higher win rate. Neither names a control group. Guard: hold one onboarding cohort out for a quarter and compare time-to-certification and first-quarter attainment against the trained cohort.
Related: ramp time defines the metric this purchase should move, and sales enablement places role-play inside the wider program.