❓ Problem

Hiring is still a guessing game dressed up as a process. We screen CVs, run three rounds of "tell me about a time when..." interviews, and then act surprised when 46% of new hires fail within 18 months. The US Department of Labor pegs the cost of each bad hire at a minimum of 30% of first-year earnings, and SHRM puts the replacement cost at 50-200% of annual salary. On an $80k role, one mis-hire is a $40-160k mistake.

Here is the part the industry keeps ignoring: we have known the fix for decades. Schmidt and Hunter's famous meta-analysis of 85 years of selection research found work sample tests sitting at the very top of the validity hierarchy for predicting job performance, while unstructured interviews rank far down the list. In other words, watching someone do the job predicts whether they can do the job. Asking them to describe doing the job, much less so.

So why does everyone still interview? Because real work samples were expensive to build. You cannot hand a candidate your live Zendesk and your actual customers. Until now, the only options were toy take-home tasks or generic aptitude quizzes that test an abstraction of the role rather than the role itself. The moment building software became trivial, that constraint disappeared.

✅ Solution

A platform that generates a 1:1 sandboxed replica of the software stack a person would actually use in a role, populated with AI-simulated customers, prospects and colleagues, for interviewing candidates and training employees.

• Employers describe the role and connect their stack. The platform clones the working environment: a support rep handles AI-simulated angry customers inside a fake Zendesk, an SDR works AI prospects in a replica of the company's CRM, an executive assistant triages a chaotic simulated inbox and calendar.
• AI plays every other character. Simulated customers escalate, go quiet, change their minds and get things wrong, the way real ones do. Voice and chat agents are now good enough that the candidate genuinely has to think on their feet rather than pattern-match a scripted exercise.
• Every session produces a structured performance report. Speed, accuracy, judgement under pressure, tone and written communication, all scored against a rubric the employer sets. The hiring manager gets a real work sample, and the candidate gets an honest preview of the job before they sign.
• The same simulations become the onboarding and training layer. Once a company's environment is built, every new hire ramps inside it, and new processes, products or tricky scenarios get rolled out as training modules in the sandbox instead of "lunch and learn" decks.

The wedge is hiring assessments for high-volume operational roles. The platform underneath, a simulation engine for any company's stack, expands into the entire learning and development budget.

📊 Key Numbers

Market size

• The talent assessment market is worth roughly $25 billion in 2025, growing at ~9.4% a year, with pre-employment testing software alone at ~$3.1 billion.
• The real prize is training: the global corporate training market was valued at ~$413 billion in 2024 and is heading towards $800 billion+ by the mid-2030s.
• The pain is quantified on both ends: bad hires cost at least 30% of first-year earnings, and it takes around 8 months for a new hire to reach full productivity, which structured onboarding can cut to 3. A product that improves both numbers sits on two budget lines at once.

ARR potential

• Bottom-up on hiring: 1,500 mid-market companies on a ~$750/mo platform fee plus per-candidate credits (say ~$25/session, 40 sessions a month) is roughly $32M ARR. Assessment budgets already exist, so this is substitution, not new spend.
• The training expansion changes the ceiling: a company paying $9k/year to assess candidates will pay $30-50k/year when every new hire onboards in the same simulations. 1,000 accounts at a blended ~$35k is a $35M ARR business from training alone.
• Realistic path: $5-10M ARR as a focused assessment tool for two or three role types, $50M+ once the training layer lands, with a much larger ceiling if it becomes the default way operational roles are hired and ramped.