Cross-team workflows

When AI connects silos

The most transformative use cases don’t live inside a single team; they span teams. Within a department, everyone already shares the same vocabulary and files: gains are real but bounded. As soon as a process crosses multiple departments, most of the time gets lost in handoffs—rephrasing, information that doesn’t cross the boundary, follow-ups. ChatGPT can become the thread that smooths these relays.

The shared project as a cross-functional space

A shared project brings together people or groups from multiple teams—up to one hundred collaborators in a workspace project—with each person granted either conversation access or edit access. It’s the natural home for cross-team processes. Project instructions make context available to everyone: a lawyer who joins mid-stream doesn’t need the stakes re-explained. Common reference files—templates, guidelines, data—eliminate the divergent versions that used to circulate by email. Project conversations, which ChatGPT draws on, let everyone pick up where others left off.

This setup pays off whenever a deliverable depends on multiple teams: a product launch (Marketing, Product, Sales), an internal audit (Finance, Legal, IT), a client project (Sales, Delivery, Support), or a recruitment campaign (HR, managers, Communications).

Five workflows that work

For a product launch, the Product team drops its brief into the project. A “Launch” assistant pulls from it to generate the press release, sales arguments, product FAQ, and social media messages. Marketing adjusts tone and approves, Sales takes the arguments and FAQ, and Support prepares its customer responses. Content stays consistent from the first draft because it all comes from the same brief.

A response to an RFP follows the same logic. The sales rep imports the statement of work; the assistant extracts key requirements; the technical team fills in what falls under its remit; Legal validates contractual clauses; and the assistant consolidates everything into a structured document, with no requirement lost between reviews.

Client onboarding addresses the classic break between sales and delivery. At signature, the sales rep drops their brief; the assistant produces the handover file—context, commitments, watch points; Delivery turns that into a deployment plan; Support prepares its client sheet. The benefit shows up fast: the client never has to repeat their story.

Monthly reporting brings together Finance, Operations, and Leadership. Finance drops in its data; each department adds its KPIs in a standardized format; the assistant consolidates and drafts the report with analysis, trends, and alerts. Leadership asks its questions in the same project, with no parallel email thread.

Incident management is where time matters most. IT opens an “Incident” project with technical details; the assistant prepares messages for each audience—customers, employees, press; Communications adjusts and approves; and Management tracks progress in the same place. Communication stays consistent and fast, without the back-and-forth that multiplies at the worst moment.

Designing yours

Five principles emerge. Create one project per workflow: two spaces for the same process guarantee that half the information will end up in the wrong place. Write instructions that describe the process step by step; otherwise, ChatGPT won’t know it’s part of a chain and will treat each request in isolation. Define roles by name: who initiates, who contributes, who approves, who closes; grant edit access to contributors and conversation access to everyone else. Enforce standardized templates so every contribution arrives in the same format—a prerequisite for reliable consolidation. And name a workflow owner, responsible for checking that the process still runs three months later.

Mistakes repeat with regularity. Too many members turn the space into an observation room: limit access to those who contribute. Missing instructions strip the project of what makes it a workflow. Skipping human review is the costliest: every output must be reviewed before feeding the next step, or an approximation propagates to the final deliverable with no way to trace where it started. Finally, don’t mix workflow and informal discussions; keep those in your Slack or Teams channels.

Integrating with your existing stack

Projects deliver their full value when connected to the rest of your information system. Added as sources, Google Drive files or folders avoid uploading stale copies; on plans that offer them, skills and plugins provide reusable methods and connections from one project to another.

The goal isn’t to replace your project management tools: Jira, Asana, or Monday continue to own tasks, milestones, and responsibilities. The aim is to smooth content production and analysis steps—the ones that slow your cross-functional processes without ever appearing on a timeline.

Your turn

Map a cross-team workflow. Pick a deliverable that depends on at least three teams in your organization. Write the project instructions describing each step, assign edit and conversation access, prepare the contribution template expected from each team, and designate the workflow owner.

Official sources: Projects in ChatGPT · Skills in ChatGPT · Plugins in ChatGPT (OpenAI Help Center)

Key takeaways

• The biggest gains happen between teams, where handoffs get lost.
• A shared project brings together common instructions, files, and conversations, with conversation or edit access.
• Product launches, RFP responses, client onboarding, reporting, and incident management fit this model well.
• One project per workflow, detailed instructions, named roles, standardized templates, and an owner.
• Review every output before feeding the next step.