AI is changing how projects get run. But a PMO can't hand its portfolio to a chatbot and hope for the best. Fluid's AI strategy rests on four commitments: meet people in the AI tools they already use, build your standards into the AI inside Fluid, keep a person in charge of every change, and never compromise on how your data is handled.

Fluid AI Connectors: bring the AI you already use

Your teams already work in Claude, ChatGPT and Codex, and in Microsoft Copilot. Fluid AI Connectors bring Fluid to them. They are built on the Model Context Protocol (MCP), the open standard AI assistants use to reach business systems.

Once connected, an assistant can work with live Fluid data on the user's behalf:

• Project managers can ask "what changed on my projects this week?", review risks and issues, update weekly status, and adjust schedule dates and assignees.
• Board and delivery teams can find, create, move and update cards across their boards.
• PMO and portfolio leads can see portfolio health, spot projects that need attention, and draft summaries for steering meetings.

Every connector is scoped to the person using it. The assistant sees exactly what that person can see in Fluid, and nothing more.

Enterprise-ready sign-in, coming October 2026

The October 2026 release adds OAuth sign-in. Each Fluid tenant gets its own connector address, https://yourcompany.fluid.work/mcp. Add it to Claude, Codex, Copilot or any MCP-compatible assistant, then sign in with your normal Fluid account.

• No tokens to copy, no software to install. People sign in with their Fluid account, and the standard OAuth flow does the rest.
• Clear consent. Before connecting, people see exactly what the assistant is asking to do, such as read projects or update status. Write and delete permissions are separate choices.
• Admin control. Administrators turn connectors on, choose which tools are available, and decide who gets access. That can start with a small trained group and widen from there.
• Easy to revoke. People can disconnect an app at any time from their own settings. Administrators can see and revoke every connection in the tenant. Access also expires on its own after 90 days.

The new sign-in launches as a Beta. Teams that need a connector today can still use personal access tokens.

Embedded FluidAI: your PMO standards, built in

Connectors serve people who live in their AI assistant. FluidAI serves people who live in Fluid. It sits inside the Project Workspace and the Project Dashboard, so answers arrive where the work is done.

Ask in plain language: "What are the key risks on this project?", "What are the next steps?", "Is anything blocked?" FluidAI answers from the project's own data: schedule, RAID, financials, resources and status history.

Generic AI gives generic advice. FluidAI works within the way your organisation runs projects:

• Guidance at the point of need. Project managers get help with the task in front of them, such as writing a status update, reviewing a risk or preparing for a gate.
• Your standards, not someone else's. Your PMO's governance rules, reporting standards and stage-gate criteria shape the guidance. A new project manager gets the same advice an experienced PMO lead would give.
• Governance that keeps itself current. Fluid Governance checks every project against your rules. FluidAI helps people understand the gaps and fix them before the steering meeting, not after.

Human in the loop, by design

In project delivery, accountability can't be delegated to a model. In Fluid, AI drafts, suggests and summarises, and a person stays in charge.

• AI acts as the user, never above them. Every connector call runs with the signed-in person's own Fluid permissions. It can't reach a project, board or portfolio they can't already see.
• Changes are deliberate. Reading data and changing it need separate permissions, and deleting is a separate permission again. Users grant each one explicitly at sign-in.
• Administrators set the limits. They decide which tools are available in the tenant, which people can connect, and whether deleting is allowed at all.
• Everything is attributable. Changes made through an assistant are recorded against the person who made them, just like changes made in the Fluid app. Your audit trail stays intact.

Trust: your data stays yours

AI is only useful if security and compliance teams can sign it off. These commitments are the foundation of Fluid's AI strategy, not an afterthought.

We don't train models on customer data. Your projects, portfolios, risks and status reports are never used to train or fine-tune AI models, ours or anyone else's.

Data residency. FluidAI runs on private AI models hosted in Fluid's own regional Microsoft Azure environment. Questions and answers stay within Fluid's services and are never sent to third-party AI services. Your data stays in the region your Fluid tenant is hosted in.

Data handling you control. Connectors are off until an administrator turns them on. When someone connects an external assistant such as Claude, Codex or Copilot, only the data they choose to work with goes to that assistant, and it is governed by your organisation's agreement with that provider. Administrators decide whether connectors are allowed, which tools are available and who may use them.

SOC 2 Type 2. Fluid's controls are independently audited under SOC 2 Type 2. Our AI capabilities run inside the same controls as the rest of the platform: secure sign-in, encryption, access control, monitoring and audit.

In short:

• Embedded FluidAI runs on private models in Fluid's Azure environment, sees only what the user can see, never trains on your data, and sits inside Fluid's SOC 2 Type 2 controls.
• AI Connectors (MCP) run in the assistant you choose, under your agreement with its provider. They need administrator approval, user sign-in and consent, see only what the user can see, and Fluid never uses the data for training.

What's next

OAuth sign-in for Fluid AI Connectors arrives in the October 2026 release, starting as a Beta. Over the coming releases we will add more tools across projects, boards and portfolios, and connect FluidAI guidance more deeply to your governance framework.

Our aim stays the same: AI that saves your teams hours every week, that follows your standards, that keeps people accountable, and that your security team is comfortable approving.

Getting started: existing customers can ask their Fluid account manager or support@fluid.work to switch on AI Connectors for an evaluation group. New to Fluid? Book a demo and see it working on real portfolio data.