Platform · AI Simulations

Real workplace scenarios that verify and develop skills

30–45 minute immersive simulations. Voice calls, chat, code, document work — with AI actors who behave like real colleagues, clients, and stakeholders. Behavioral evidence, not subjective ratings.

300+
Training simulations
30–45
Minutes per simulation
5
Task types
3
Modalities: hire · assess · train
What is an AI Simulation

Not a test. A real work scenario.

A candidate or employee takes on a specific role — project manager, sales rep, engineer — inside a realistic company context. They work through actual tasks, make decisions, and interact with AI-powered characters. Performance is evaluated against defined skills, not against the effort of a human reviewer.

Real role, real stakes

Players take on a specific job role and receive a briefing. The scenario is built around real workplace situations — not abstract puzzles.

AI actors, not scripts

Up to 4 AI characters per simulation — colleagues, clients, managers. Each has a personality, a backstory, and a defined level of difficulty.

Mixed modalities

One simulation can include voice calls, chat, code challenges, document analysis, and collaborative editing — mirroring how real work actually happens.

Objective evaluation

Every simulation is scored against defined skills and criteria. Pass/fail checks. Competency levels 0–5. No subjective ratings, no interviewer bias.

Task types

Five ways people demonstrate skill

A single simulation can mix task types — reflecting how skills are actually used at work. Not one format for everything.

Voice call

Real-time voice conversation

Live voice interaction with an AI actor. Discovery calls, coaching sessions, negotiations, incident escalations. The player speaks; the actor responds in real time. Measures communication, presence, and judgment under pressure.

Chat

Async written interaction

Text-based conversation with an AI actor simulating a colleague, client, or stakeholder. Tests written communication, clarity, and professional judgment — at a realistic pace.

Code

Live coding environment

Full code editor with syntax highlighting and test execution. Debugging, architecture review, code quality challenges. Supports Claude Code, Copilot, and Codex integration for AI tooling proficiency scenarios.

Document

Review, create, analyse

Players work with real-format files — PDF, DOCX, PPTX, XLSX — to complete analysis, drafting, or review tasks. Supports structured output evaluation against defined criteria.

Collaborative doc

Shared document — player and AI work together

A live Markdown workspace both the player and AI actors can read and edit in real time. Built for co-authoring, joint planning, shared analysis — scenarios where writing is collaborative, not solo. The final state of the document is evaluated just like a code submission.

Three modalities

Hire, assess, or train — same simulation engine

The same AI Simulation infrastructure powers all three use cases. What changes is tone, difficulty calibration, and the purpose of the feedback.

Hiring

Assess external candidates

Test whether a candidate can actually do the job — before making an offer. Replace or complement early-stage interviews with evidence from a realistic job scenario.

Output: objective skill scores per candidate. No interviewer bias. Comparable results at scale.
Assessment

Verify your internal workforce

Map the real skill levels of your existing employees — not what they claim to have, but what they demonstrate under realistic conditions. Feed the AI Readiness Score and skills intelligence layer.

Output: verified skill profile per person. Feeds directly into Workforce Intelligence and AI Readiness reporting.
Training

Develop skills through practice

Training simulations guide employees through skill development with more structured hints, checkpoints, and feedback. They practice; the system measures and reinforces. Learning with evidence of outcome.

Output: demonstrated skill acquisition with measurable competency progression — not just course completion.
AI actors

Characters designed to challenge — never to block

Each simulation has up to 4 AI actors. They push players to demonstrate real skills — but always leave a path to success.

Friendly

Supportive and collaborative

Responsive, helpful characters. Good for lower-stakes scenarios or training contexts where psychological safety matters.

Neutral

Professional and realistic

Business-as-usual characters. Neither helpful nor obstructive. Most simulations use neutral as the default — it reflects real work.

Difficult

Challenging but fair

Characters with strong opinions, skepticism, or resistance. They require players to demonstrate real skill — but always respond to good reasoning.

Evaluation

How every simulation is scored

Scores are built from defined criteria — not impressions. Every data point is traceable to a specific behavior observed during the simulation.

  • 1
    Skills selected (3–4 per simulation)
    Each simulation targets a focused set of skills — enough to produce a meaningful profile without diluting the assessment.
  • 2
    Evaluation criteria per skill
    Each skill has multiple criteria defining the specific behaviors that demonstrate competency in this scenario and context.
  • 3
    Score checks (pass/fail)
    Essential checks that must pass, plus non-essential checks that contribute to the overall score. Binary, traceable, auditable.
  • 4
    Competency level (0–5)
    Raw scores map to a 0–5 competency level. Every person’s result ties directly to the 60K+ skill taxonomy — feeds the Workforce Intelligence layer.
Competency scale
0
Cannot perform
1
Needs supervision
2
Works independently on routine tasks
3
Delivers consistent results, trains others
4
Owns complex projects, mentors
5
Drives innovation, sets standards
Where it fits

Phase 3 of the 5-Phase Methodology

AI Simulations are the verification layer — where claimed skills become demonstrated skills.

Phase 1
Auto-mapping
CV · HRIS · LinkedIn · external sources
Phase 2
Guided self-evaluation
Confirm what people think they can do
Phase 3 ← you are here
AI Simulation
30–45 min behavioral verification — the evidence layer
Phase 4
AI Interview
Qualitative signal — sentiment, context, aspiration
Phase 5
Integrated analysis
AI Readiness Score — per person, team, and org
Common questions

Frequently asked

What HR leaders and engineering teams ask before deploying AI Simulations.

How long does a typical AI Simulation take?
Most simulations are designed to take 30 minutes. The maximum is 45–60 minutes for complex, multi-task scenarios. Simulations are not timed tests — the session ends when the player completes the tasks, within the designed scope.
Do players need any special setup or software?
No. Simulations run in a web browser. Voice calls use the built-in microphone. No software to install, no VPN, no special hardware. Players receive a link and start immediately.
How many simulations are available out of the box?
Anthropos includes 300+ training simulations across tech, commercial, finance, HR, and operations roles. Hiring assessments have a separate library. You can also build custom simulations using Anthropos Studio — no engineers required.
Can we test AI tooling proficiency — like Copilot or Claude Code?
Yes. Coding and GenAI simulation templates support live AI tool environments. Engineers work on real coding scenarios with Claude Code, Copilot, or Codex active — the same tools they use in production. Results feed the AI Readiness Score.
How does scoring connect to the rest of the platform?
Every simulation result maps directly to skills in the 60K+ taxonomy. Scores feed the Workforce Intelligence layer — updating each person’s skill profile and AI Readiness tier in real time. No manual data entry or reporting pipeline.
AI Simulations

Ready to see an AI Simulation in action?

Book a 30-minute demo and run through a simulation yourself. The best way to understand what candidates and employees actually experience.