Cut cost on high-volume, routine work
If most of the workspace's traffic is simple questions or repetitive formatting, set the workspace model to a Fast tier and handle that same volume for a fraction of the credits a Frontier model would cost.
Model Selection
Every employee runs on a model you choose, not one fixed default. Switch models per employee from the AI Models page whenever a job calls for more reasoning power or a cheaper, faster response.
Every workspace runs on one model at a time, chosen from four tiers: Fast, Standard, Advanced, and Frontier. The picker lives on the AI Models page under Settings > Technical, and it applies to every employee in the workspace at once, not one employee at a time. The tier is a rough capability class, not a strict price ladder: a cheap Fast model and a cheap Advanced model can land close together in credit cost, so the picker shows the exact rate on each model, not just its tier.
A model change is not retroactive and never breaks continuity. It takes effect on each employee's next turn; everything an employee already knows, its skills, tools, and full conversation history, stays exactly as it was. Reasoning-capable models can be tuned further with a Thinking Depth selector (Auto, Off, Minimal, Light, Medium, Deep) that controls how hard the model reasons before answering, at the cost of more credits on deeper settings.
This is a workspace default, not a per-employee dial: the backend can store an override for one employee, but that control is switched off in the interface today, so picking a model here changes what every employee in the workspace runs on. Model Selection also sits next to, but is different from, Credits Saver: Credits Saver automatically routes simple conversational turns to a cheaper model behind the scenes and leaves your chosen model for real work, while Model Selection is the manual choice of which model that real work runs on in the first place.
Free and Pending workspaces stay locked to the platform default model, shown in the picker with a lock icon and an upgrade prompt. Builder plan and above unlock the full picker. Paid plans start on a Frontier-tier model as their default brain, so choosing a different model is an optional tune, not something you have to do to get a capable employee.
One Picker, Applied Workspace-Wide
The AI Models page shows every available model grouped into four tiers, with its provider, a 1-to-5 star reasoning rating, a brain icon if it streams a visible thinking process, and Recommended or Expensive badges where they apply. Whichever model you pick there is what every employee in the workspace runs on; there is no separate picker per employee in the interface today.
You can sort the list by cheapest, by provider, or by reasoning strength instead of by tier, and filter down to reasoning-only models. That matters because tier and price don't move in lockstep: check the credit rate next to the specific model, not just which tier it sits in, before switching the whole workspace to it.
Switching Never Loses Continuity
Change the workspace model and every employee's memory, skills, tools, and conversation history carry over untouched. The switch changes which engine answers the next turn, not who an employee is or what it remembers about your business, and it applies immediately with no downtime.
See the Effect, Not Just the Choice
The Activity Inspector on any past message shows which model actually produced it, so a model change isn't a guess: you can judge whether a switch made replies sharper or slower by looking at the deliverables that came back, not by trusting that a higher tier or a bigger number means a better result.
How It Works
One picker, per employee or company-wide, with the change taking effect on the next turn
Model Selection lives on the AI Behavior page under Settings > Technical, and on each employee's own Settings tab. The company-wide picker sets the default that new hires start on; the per-employee picker overrides it for one employee without touching anyone else's model.
Models are grouped by tier so the trade-off is visible before you commit: Fast, Standard, Advanced, and Frontier, each shown with its provider, its relative credit cost, and whether it supports reasoning. Reasoning-capable models can be tuned further with a Thinking Depth selector, which controls how much the employee reasons before answering.
A model change is not retroactive and never breaks continuity. It applies starting with the employee's next message: everything the employee already knows, its skills, its tools, and its full conversation history, stays exactly as it was. The Activity Inspector on any past message shows which model actually produced it, so you can judge a switch by the deliverables that came back, not by guesswork.
Use Cases
Get more rigor on work that needs it
Switch the workspace to an Advanced or Frontier model with reasoning enabled ahead of competitive research, long document analysis, or multi-step planning, so employees think through the problem instead of rushing a shallow first pass.
Judge a model change by real output
After switching, use the Activity Inspector to compare the deliverables produced before and after on the same kind of task. You see which model actually ran each piece of work, so you know whether the switch was worth it before you commit to keeping it.
Pair with Credits Saver instead of switching manually every time
Leave the workspace on a stronger model for real work and turn on Credits Saver, which automatically routes simple conversational turns to a cheaper model behind the scenes, rather than manually flipping the whole workspace back and forth.
FAQ
Do I have to pick a model, or is there a good default?
Every workspace ships on a strong default model, and paid plans start on a Frontier-tier brain. Model Selection is for when you want to tune it yourself, not a required setup step.
Can different employees run on different models?
Not from the interface today. The model you pick on the AI Models page applies to every employee in the workspace at once; there is no per-employee picker currently shown.
Does switching models lose an employee's memory or history?
No. Memory, skills, tools, and conversation history carry over. Only which model answers the next turn changes, and the switch is not retroactive.
Can I see which model actually handled a piece of work?
Yes. The Activity Inspector on any past message shows the model that produced it, so you can judge a switch by the deliverables it produced, not just by the tier name.
Is Model Selection available on every plan?
It's a Builder-and-above capability. Free and Pending workspaces stay locked to the platform default model with an upgrade prompt on the picker.
Where Model Selection fits
Model Selection is part of Things you hire or set up.
Browse a marketplace of pre-built AI agent roles and full teams, or build your own from scratch. Organize them into teams with leaders, set up your org chart, and onboard new hires in minutes. Every employee arrives configured and ready to deliver from day one.
- Things you hire or set up: A full workforce. Hired in seconds.
Read the guide
More in Fast Onboarding
- Hire an AI Employee: Browse the Marketplace and hire an AI employee for a specific role: a department specialist, a personal assistant, or a custom role you define from scratch. Hiring is free and instant. You land straight in the new employee's chat, and every skill, duty, and tool it starts with can be changed afterward from its dashboard.
- Interview Before Hiring: Chat with any employee template before you hire it, on the public marketplace pages or inside the app, to see how it thinks and communicates about your actual work. When you're convinced, say so in the conversation or click Hire; nothing is added to your team until you decide.
- Employee Builder: Employee Builder is the "Build Your Own" card in the Marketplace: instead of hiring a pre-built role, you start from a blank template and set the name, role, and persona yourself. The new hire starts with one skill, Research & Brief, and nothing else pre-assigned, so you add skills, duties, and tools from the Skills, Duties, and Tools tabs to match the exact job you need done. It is a Custom-plan feature; every other plan hires from the marketplace's pre-built roles instead.
- Personal Assistants: Every account starts with a Personal Assistant already hired, ready to talk before you decide who else to bring on. No marketplace trip, no setup, no decision to make first.
- Get Started Through Conversation: When you hire an AI employee, they land in your chat with a short greeting that names their role and what they handle, then ask what you need first. There is no setup wizard, no progress bar, and no settings panel to fill in before they can work. You customize name, persona, skills, and tools later from their dashboard, but none of that is required before the first message.
- Manage Employee Status: Pause, resume, or let go of an AI employee without losing anything. Suspend an employee when you don't need them right now and bring them back exactly as you left them, or terminate one and keep their full chat history and work record in your Alumni list. Terminated employees stop counting against your plan's employee limit, and you can re-hire the same role later.
- Hire a Full Team: Hire a complete pre-built team from the Marketplace in one click: every member configured, a leader assigned, and sprint planning ready to start. Or build a custom team from your existing hires, pick a leader, and set it up yourself. Either way, you talk to one leader and they coordinate the rest.
- Keep Former Team Members: When you let an employee go, they don't disappear. Sistava moves them into your Alumni list, where their chat history and work stay intact and readable. If you need them back, one click restores them to active with everything they knew still in place.
- Marketplace: The Marketplace is where you browse and hire AI employees: two personal assistants, a specialist for every department (Marketing, Sales, Support, HR, Operations), and a Custom AI Talent card if none of the pre-built roles fit. Click a card's photo to preview its skills, duties, tools, and persona before you commit, then click Hire to add it to your workforce and land straight in its chat. Hiring is free; you only spend credits when an employee does work.
- Employee Avatar Upload: Upload a custom profile photo for any AI employee straight from their Profile tab, replacing the default generated avatar. JPEG, PNG, GIF, and WebP files up to 5 MB are accepted, and the change takes effect immediately. The photo appears everywhere that employee shows up: chat, the org chart, the virtual office, notifications, and task cards. Any member of your workspace can change it, not only the person who hired that employee.
- Multiple Workspaces: Create additional workspaces under the same login, each with its own AI employees, conversations, training data, billing, and collaborators. Nothing crosses a workspace boundary by default, so a separate client, business, or department stays fully isolated. Your plan sets how many workspaces you can create, and you switch between them without signing out.
- Academy Live Sessions: Learn how to put an AI employee to work in a live session with an Applied AI Engineer, tailored to your business. Sessions run about an hour, one on one or in a small group, and walk through the same curriculum as the free Academy course. Free for founders, one hundred dollars per session otherwise.