AI Provider & Model Selection

IntuitivePM supports AI features through company-approved AI services. Most workspaces use built-in cloud providers such as Claude, OpenAI, Gemini, and Mistral. Company administrators may also approve optional non-US/non-EU providers such as GLM/Z.ai or Kimi/Moonshot, or configure a company-managed OpenAI-compatible service such as an internal AI gateway or on-prem model endpoint.

Every AI feature in IntuitivePM, including Project Assist, Epic Assist, AI chat, and Virtual Colleagues, uses either your personal AI preference or your company’s default AI service.

The AI Assistant can answer questions about IntuitivePM data that you are allowed to access, such as listing the projects visible to you in the current workspace. It uses IntuitivePM tools for company data and stays scoped to your active company permissions.

Where to Configure Your Preference

Your AI preference is managed from your profile:

Click your avatar or name in the sidebar.
Select Profile.
Open the AI Credits tab.
In Preferred AI Service, choose Use company default or one of the approved services.
If you selected a specific service, choose an available model.
Click Save Changes.

Use company default is recommended for most users. It follows the default service and model chosen by your company administrator.

Your preference is saved per workspace. If you belong to multiple companies, changing your AI preference in one company does not change your preference in another company.

If your company has disabled AI services, the selectors are unavailable until an administrator enables AI again.

Available Services

The services you see depend on what your company administrator has approved.

Service	Typical Models	Best For
Claude	Opus, Sonnet, Haiku	Complex reasoning, long-form writing, nuanced task decomposition
OpenAI	GPT flagship, standard, and lightweight models	General-purpose work, speed-sensitive workflows
Gemini	Pro, Flash, Flash Lite	Long-context documents, multimodal inputs, large codebase analysis
Mistral	Premier, standard, and fast model slots	European-provider workflows, general AI assistance, and cost-conscious work
GLM/Z.ai	GLM model slots	Optional coding and agentic workflows when your company explicitly approves non-US/non-EU providers
Kimi/Moonshot	Kimi coding model slots	Optional coding and agentic workflows when your company explicitly approves non-US/non-EU providers
Custom OpenAI-Compatible	Company-defined model IDs	Enterprise internal cloud or on-prem AI services

Custom OpenAI-compatible services are configured by company administrators and are available only on Enterprise plans. Regular users cannot add endpoints or API keys themselves.

Mistral, GLM/Z.ai, and Kimi/Moonshot only appear when they are available for your workspace and approved by your company administrator. GLM/Z.ai and Kimi/Moonshot are explicit opt-in services because their provider jurisdiction differs from the standard US/EU built-in choices.

Latest And Previous Model Slots

For built-in providers, IntuitivePM shows curated model slots instead of long provider-specific model ID lists. Your company may approve slots such as:

Claude Opus Latest or Claude Opus Previous
Claude Sonnet Latest or Claude Sonnet Previous
Gemini Pro Latest or Gemini Pro Previous
OpenAI Standard Latest or OpenAI Standard Previous
Mistral Standard Latest or Mistral Budget Latest
GLM Coding Latest or Kimi Coding Latest when your company has opted in

Latest follows the newest IntuitivePM-validated model in that family. Previous keeps the prior validated model available for companies that need extra stability before moving to the newest version.

The selector still shows the exact provider model ID in the model details or tooltip. Use that exact ID if your team needs to research provider release notes or compare behavior between versions.

Choosing a Model

There is no universally correct choice. The right model depends on what you are doing:

Project scaffolding and epic generation: use a stronger reasoning model such as Claude Opus, a flagship GPT model, Gemini Pro, or an approved Mistral model.
High-volume Virtual Colleague tasks: use a faster, lower-cost model such as Claude Haiku, a lightweight GPT model, Gemini Flash/Flash Lite, or a fast approved Mistral slot.
Long documents or large codebases: Gemini Pro, GLM, Kimi, and other long-context models can be useful when the AI needs to reason over a lot of information at once.
Jurisdiction-sensitive work: ask your company administrator which providers are approved for your organization’s policy. European-provider options and non-US/non-EU opt-in providers are governed separately.
Operationally sensitive workflows: choose a Previous slot if your team needs time to test the newest provider model before using it for important daily work.
Company-controlled environments: if your company provides a custom AI service, use the guidance from your administrator. Model names may be specific to your organization.

Company AI Defaults

Company administrators manage AI availability from AI Agents -> Advanced setup -> AI Providers. Admins can:

Enable or disable AI services for the company.
Approve or remove built-in model slots from the company pool.
Set the company default service and model.
Add and manage Enterprise custom OpenAI-compatible services.

Company AI settings use a page-level save pattern. Changes such as disabling Gemini, changing the company default, or disabling AI are not applied until the administrator clicks Save Changes on the AI Provider page.

Administrators still use Profile -> AI Credits for their own personal provider preference and credit/usage view. Profile includes a link to the AI Provider settings page when the user has permission to manage company AI settings.

Built-in services such as Claude, OpenAI, Gemini, Mistral, GLM/Z.ai, and Kimi/Moonshot cannot be deleted by company administrators. Their model slots can be removed from the company pool by unapproving them.

GLM/Z.ai and Kimi/Moonshot require explicit company opt-in before users can run workloads on them. This extra approval is separate from ordinary model-slot approval and is intended to make provider-jurisdiction decisions deliberate.

When AI is enabled, the company must have at least one approved service and a valid default service/model. If no service is approved, AI features remain unavailable until an administrator approves a service or disables AI.

Custom OpenAI-Compatible Services

Enterprise companies can use a custom OpenAI-compatible AI service. This can be useful when your organization has its own AI gateway, private cloud model endpoint, or on-prem model service.

Important rules:

Only company administrators can create or manage custom services.
The service URL must use HTTPS.
API keys, if required, are stored securely by IntuitivePM.
The administrator decides which models are available.
The service must pass a connection test before it can be activated.

If your company uses a custom service, it appears in the same Preferred AI Service dropdown as built-in services once it has been activated and approved.

AI Credits

Built-in cloud providers consume IntuitivePM AI credits. Credit usage depends on the selected provider, model tier, input size, and output size.

In general:

Flagship models consume more credits and are best for difficult or high-value work.
Balanced models are good everyday defaults.
Fast/lite models consume fewer credits and are best for frequent automated work.

Credit rates, balances, and recent usage are shown in Profile -> AI Credits.

Enterprise custom OpenAI-compatible services may not consume vendor-token credits in the same way as built-in cloud providers because your company may operate the model endpoint directly. Your administrator can explain how your company accounts for that usage.

How AI Features Use Your Selection

AI-powered features read your preference at the time a request is made:

If you selected Use company default, IntuitivePM uses the current company default.
If you selected a specific approved service/model, IntuitivePM uses that selection while it remains available.
If your selected service or model slot is later removed from the company pool, IntuitivePM falls back to the company default when possible.
If company AI is disabled, AI requests stop before a provider is contacted.

AI Project Assist

When you create a project using AI Project Assist, IntuitivePM sends your project description to your selected service and asks it to generate epics, tasks, and milestones. See the Creating Projects guide for the full walkthrough.

AI Epic Assist

AI Epic Assist uses your selected service to break down a high-level epic into individual tasks with descriptions, suggested estimates, and acceptance criteria.

Virtual Colleagues

Virtual Colleagues use the AI service configured for their run context. If they perform frequent automated scans or reports, choosing an efficient company default helps control credit usage.

Viewing And Managing Providers

For regular users, Profile -> AI Credits remains the place to view and change your personal AI preference.

For company administrators, AI Agents -> Advanced setup -> AI Providers is the place to manage company AI services. This page includes:

Company AI enabled/disabled status.
Approved built-in provider model slots.
Enterprise custom OpenAI-compatible providers.
Company default service/model.
Provider test and activation actions.

The AI Provider page lives under AI Agents -> Advanced setup -> AI Providers and is also reachable via global search with terms such as ai, model, provider, custom OpenAI, Claude, OpenAI, Gemini, Mistral, GLM, or Kimi.

For workspaces using managed internal agents, company administrators also see the Runtime Capacity tab. This tab shows the current managed-agent rollout limit, runtime online count, work-slot usage, company active-work cap, and per-agent active-work cap. Use it when an internal agent cannot be selected, appears to be waiting for capacity, or when planning how many managed internal agents a workspace can safely run.