AI Assist changes for Enterprise plans
Last updated on
Starting November 9, 2026, AI Assist in Screen and Interview comes in packages, each with its own AI models, features and usage limits.
Action needed before November 9
Review the AI Packages and decide which one fits best for your team.
This page explains what changes for Enterprise plans, what happens when a session reaches its limit, and how Frontier Pro and Frontier Max keep your most demanding interviews on frontier models.
What's changing
- A limit of 5M AI tokens per test or interview on the included Frontier package.
When a test or interview reaches the 5M limit
The session doesn't stop, but it switches to a lighter model for the rest of the test or interview.
- Interviewers see a notice that the model changed.
- Candidates keep working, but with a less capable model for the rest of the session.
- In long or agentic interviews, this can change what you observe in the final stretch. If your scorecards weigh how candidates work with AI late in a session, account for it.
Where to see your AI usage in Interview
You can check how much AI a pad has used at any point during an interview, so you can keep an eye on long or agentic sessions as they approach your limit of 5M per test or interview.
- In an Interview pad, open the AI Assist panel.
- Find the usage indicator, the circle in the top-right corner of the panel, next to the model icons.
- Hover over the indicator to see how many tokens the pad has used so far, for example “You’ve used 48.2K tokens.”
Compare Frontier, Frontier Pro and Frontier Max
Frontier is included with every Enterprise plan. Frontier Pro and Frontier Max are new packages for the interviews that push AI hardest, with more AI usage and more agents at once.
| Feature | Frontier (included) | Frontier Pro | Frontier Max |
|---|---|---|---|
| Models | Every model, incl. frontier | Every model, incl. frontier | Every model, incl. frontier |
| AI usage per test or interview | 5M tokens | 20M tokens | 40M tokens |
| AI agents at once | 3 | 4 | 4 |
| Claude Code and Codex in the pad | Included | Included | Included |
| Best for | AI-assisted interviews with frontier models | Agentic interviews and long take-homes | The longest, most demanding multi-agent interviews |
Choosing a package
- Frontier if you assess how candidates use AI to write, debug and explain code in a live interview.
- Frontier Pro if you assess how candidates build with AI agents: planning a feature, directing agents through implementation and testing, and reviewing what the AI produces.
- Frontier Max if you assess orchestration: candidates coordinating several agents across a large codebase or an extended take-home project.
Before November 9
- Identify which interview types use AI most heavily: agentic, multi-agent or long take-home formats.
- Decide whether those interviews should stay on frontier models from start to finish.
- Talk to your account manager about which package fits.