Handoffs

A handoff is when a user moves from one part of their workflow to another. Handoffs keep important content and provide the right tools and support users need to pick up where they left off and finish their work confidently.

Depending on the task, handoffs can be within an app or between multiple apps.

Handoff principles

Handoffs should always enable three foundational principles.

Guide seamlessly

Maintain context

Unify experiences

Guide seamlessly

Copilot sets the stage for what comes next and clearly communicates when a task can or should be completed elsewhere. Handoffs guide users to their next logical step and creates a seamless transition within workflows. They take the onus off the user to figure out how to complete their task.

Maintain context

Successful handoffs maintain all important context between one step and the next and don’t carry over things the user doesn’t need anymore. That allows the user to focus on what’s important. Bookending messages clarify wayfinding so that users always have the important information and breadcrumbs to find their way back if needed.

Unify experiences

When handoffs work across endpoints, it makes our experiences feel intentional and unified. Clear CTAs let the user know that they remain in control and can initiate handoffs.

The role of AI in handoffs

Copilot can facilitate handoffs in a number of ways.

Understanding content

Copilot reads and parses user’s content for structure, purpose, and key points. It builds context so that it can logically determine a next step.

Transforming content

Copilot can help reshape and reformat content to fit the needs of the new app. For example, taking an outline and creating slides or taking a data visualization and charting it into a table.

Enhancing content

Copilot helps improve content primarily by adjusting formatting as needed, but also adjusting tone, summarizing, expanding, or cleaning up format for new creation. It can sometimes also enhance content with images and additional materials.

Guide users

Copilot helps decide the best app for content based on user intent. It then walks the user through handoff with suggestions, previews, and optional adjustments.

Learn user preferences

AI helps the user through handoffs with suggestions, optional adjustments, previews.

Handoffs & user intent

In order to make handoffs work, Copilot infers a user’s intention and suggests a plausible next step. The more information Copilot has, the better its suggestions will be.

Strong intent

User is goal oriented and prompt clearly expresses a specific goal or action they want to take. Copilot can confidently suggest or trigger a relevant next step without requiring further clarification.

Semi-formed intent

The user is exploring a concept with some structure in mind. Users prompt hints possible goal or action but lacks clarity or specificity on the end state.

Loose intent

The user is just brainstorming and has no format or end point in mind. In these cases, Copilot should let the user know of handy possibilities and use these situations to communicate its feature capabilities.

Visual patterns for handoffs

System messages are always written in third person, describing what Copilot did and where. The message is a full sentence, and ends with a period. Use a conversational voice and tone, but keep it as brief as possible. One line is ideal, but the message may wrap to another line if needed.

Chat responses

A chat response is any output presented in the chat interface generated to address the user’s prompt or input. These may include include text, visual artifacts, or entities.

Artifacts

Artifacts are Chat-created rich responses and can be both edited and interactive. Artifacts are only stored in Chat. They are visual representations of chat responses in mainline chat.

Entities

Entities are a representation of M365 files in the chat. In side-by-side view, they may appear with full app functionality, in view-only mode or limited editing capabilities.

Side-by-side

Entities are a representation of M365 files in the chat. In side-by-side view, they may appear with full app functionality, in view-only mode or limited editing capabilities.

Module inline

Inline handoffs are CTA elements embedded directly within AI hub pages. They appear within page’s primary layout to guide users toward the next step, for example, starting a new workflow, opening a related experience, learning more about a feature, or completing a key task.

Content patterns for handoff

Common handoff CTAs

There are a number of common CTAs used for handoffs that allow users to infer the action that will come next. Stick to these CTAs for these scenarios to create cohesive and intuitive experiences.

Create …

Use “Create,” “Create in [app],” or “Create [file type]” when users want to build something new out of existing content. This normally happens in a new app.

Open in …

Use “Open in …” when users want to access already existing content, not create anything new or transform anything that already exists. Little or no AI involvement is required at this stage/No major transformation needed.

Continue in …

Use “Continue in [app]” when users want to continue working somewhere else that may provide deeper functionality. For example, a user might take an outline they drafted in Pages and choose to continue in Word to create a full white paper.

Try in …

Use “Try in [app]” when users want to experiment and explore what an output may look like or how it may work. Try in is best for loose intent, playful or experimental flows and works well with test data to teach about Copilot’s capabilities.

System messages & bookends

When handing off from chat to other apps, use system messages to facilitate wayfinding and tracking handoff. The bookends are system message pairs displayed in the initiating and the destination apps. They announce to the user what Copilot did, and where. In general all of these system messages follow the pattern: “Copilot [past-tense verb] [an object] [in/from] [AppName].”