A client handoff doc usually starts the same way. Someone opens a blank page, copies deal notes out of the CRM, pastes in contact names, and hopes nothing changed since the last export. HubSpot Work starts that doc somewhere else. Documents in HubSpot Work are created inside the same platform as your contacts, deals, companies, and project tables, so the AI assistant drafts them from records it can already see.
In practice, you can describe the document you need in a prompt, reference a project table with an @ mention, and the assistant can ask clarifying questions before it writes. The draft then lives inside a project next to the tracker, form, and agent that feed it. Your team co-edits it, leaves comments, checks version history, and keeps it current as the work moves.
All of that runs on the quality of the CRM data underneath it. Campaign Creators helps teams get that foundation right through HubSpot onboarding and setup, CRM architecture, data migration, and HubSpot AI agent implementation, so tools like HubSpot Work have clean, connected context to draw from on day one.
Documents in HubSpot Work are rich-text assets that live inside a project, next to the tables, forms, and agents serving the same process. The structure has three layers.
HubSpot positions documents for information that doesn't suit a table, such as briefs, operational playbooks, and meeting recaps. A document can also serve as a summary that uses one or more project tables as its source, so a status recap is written straight from the tracker it describes.
Tables can point back to documents too. One column type links each row to a HubSpot Work document, which gives every client in an onboarding tracker a direct path to its own kickoff notes or plan.
One naming detail saves confusion later. Projects in HubSpot Work are a separate feature from the CRM projects object and from Breeze Assistant projects, even though the names overlap.
HubSpot Work is an AI-first work management product built into HubSpot, where teams and AI agents plan, run, and automate the operational work around your CRM records. HubSpot introduced it at UNBOUND 2026 in September, and partners at the product briefing reported it as the first product HubSpot designed AI-first. It is currently in public beta.
The work it targets usually sits in spreadsheets and project tools beside HubSpot, such as client onboarding, webinar coordination, deal desk requests, and content pipelines. That split costs teams real hours. Research found executives and teams spend a quarter of their workweek searching for information, and it estimates that search costs the Fortune 500 alone 2.4 billion hours a year.
Each HubSpot Work project is built from four kinds of assets. They are tables, documents, forms, and AI agents. Documents cover the written side of a process. Because they sit in the same platform as your customer data, the assistant writes them with that data already in view.
HubSpot Work begins with what your portal already knows. That covers your contacts, companies, deals, and their interaction history. HubSpot calls this wider picture Growth Context, the understanding of your customers, team, and business that its AI needs to drive outcomes across marketing, sales, and customer success.
For a document, that context arrives from four places.
|
Context source |
What it adds to the document |
|
CRM records |
Client names, key contacts, owners, deal stages, and products or services purchased |
|
Interaction history |
What has already been discussed with the customer across logged activity |
|
Project tables |
Stages, tasks, owners, dates, risks, and blockers tracked for the process, including CRM lookup columns |
|
Imported files |
Existing playbooks, SOPs, or briefs you upload when you set up a workspace |
The quality of that context shows up in results. HubSpot reports that customers using AI with high-quality context generated 5x more MQLs, closed 4.7x more deals, and resolved 6.8x more tickets compared with customers working from low-quality context. At UNBOUND, HubSpot also shared data showing that AI running on poor or missing context underperforms having no AI at all. A document drafted from complete, current records needs fewer corrections before anyone can use it.
Related reading: Managing Business Work With CRM Context
The AI assistant creates a document from a plain-language prompt typed into the prompt bar at the bottom of the Work Management screen.
A strong document prompt names five things, which are the audience, the source material, the format, the tone, and who gets access. HubSpot's documentation frames these as the questions to settle before you press Enter.
|
Prompt element |
What to specify |
|
Audience |
The team or role the document serves, such as Customer Success or Implementation |
|
Source material |
The tables you @ mention, plus any files the assistant should draw from |
|
Format |
Checklist, brief, playbook, proposal, or recap, along with the sections you expect |
|
Tone |
Plain, concise, formal, or conversational |
|
Access |
Which teams can edit, which can view, and who holds full admin control |
Here is a prompt that covers all five for a sales-to-service handoff.
Create a Sales-to-CS Handoff Brief for each deal that closed this week. Use the @Closed Won Handoffs table as the source, including deal amount, products purchased, key contacts, promised timelines, and open risks. Format it as a one-page brief with a summary, a stakeholder list, commitments made during the sale, and first-30-day priorities. Keep the tone direct and plain. Give Customer Success edit access and Sales view access. Ask me clarifying questions before you build it.
The more of your real process a prompt names, the less editing the draft needs once it lands.
Teams build the written assets that hold a recurring process in place, from playbooks and SOPs to briefs, handoff docs, and meeting recaps. Each one draws on a different slice of CRM and project data.
|
Team |
Document |
Context it draws on |
|
Customer Success |
Client onboarding playbook organized by stage |
Onboarding tracker stages, required tasks, owners, timelines, and escalation criteria |
|
Sales and Customer Success |
Handoff doc for a newly closed deal |
Deal history, products purchased, key contacts, and promised dates |
|
Sales |
Account plan for each named account |
Company records, deals, and meeting transcripts |
|
Marketing |
Webinar brief and run-of-show |
Webinar timeline table, task owners, and deadlines |
|
RevOps and deal desk |
Request summary for a pricing or contract exception |
Deal properties and the routing queue the request came through |
|
Operations |
SOP for an internal request process |
Intake form submissions and the request table they feed |
|
Any project team |
Weekly status recap |
Open items, risks, and blockers from the project tables |
Documents also scale per record. A per-client onboarding folder, for example, gives every new customer its own set of working docs inside the same project, each tied to that client's row in the tracker.
Anyone with Project Contributor access can co-edit a document with the rest of the project. The editor covers the full review cycle.
A practical pattern is to let the assistant write the first draft, then have the process owner review it through comments before the doc reaches the wider team. Version history keeps a clear record of what the AI wrote and what people changed after.
Documents sit at the end of a chain that starts with forms and tables and runs through agents. Forms collect information and send each submission to a single table, which keeps rows consistent and can trigger automations. Agents then act on that table, and documents turn the results into something people can read and share.
A client onboarding process shows how the pieces line up.
HubSpot describes this always-on side of the product as drafting handoff docs from deal history, routing requests based on real CRM properties, and nudging owners before deadlines slip. As the work unfolds, the assistant also recommends ways to improve the process itself.
A document inherits its access from the project it sits in, and the project inherits from its workspace, unless someone sets a tighter rule further down. Roles exist at both the workspace and project level.
|
Level |
Role |
What the role can do |
|
Workspace |
Admin |
Change anything inside the workspace |
|
Workspace |
Member |
View and contribute to projects and assets, without changing workspace settings or sharing the workspace |
|
Project |
Owner |
Change the project configuration and every asset in it |
|
Project |
Contributor |
Create and edit assets, including documents and agents |
|
Project |
Viewer |
View the project and its assets with no editing rights |
Workspaces themselves can be private, viewable by the whole organization, or editable by the whole organization. Organization-wide options only reach users who have the Work Management permission switched on in their user settings.
Sensitive documents get their own lock. Setting a single document to invite-only overrides the project's inherited access, so only the people added directly can open it. That fits pricing exceptions, contract terms, or account plans that only a few people should see.
Admins also control the AI layer itself. AI feature access and the data shared with HubSpot's AI tools are both managed in the account's AI settings.
You need an eligible subscription, an assigned seat, and a Super Admin to switch on two settings.
Once enrollment is complete, HubSpot Work appears under More in the main navigation as Work Management.
Prepare your CRM by cleaning the records, properties, and associations your documents will pull from, then pilot a single process.
For more information about HubSpot Workspace, read this article.
Find out before you roll HubSpot Work out to your teams. Campaign Creators' free HubSpot AI Readiness Index is a 12-question diagnostic built from the portal audits we run as a HubSpot Elite Solutions Partner. It scores your portal across five dimensions, including Data Foundation and Architecture, then names the three moves that would raise your grade fastest. It takes about three minutes and needs no portal login.
When you're ready to act on the results, talk with our team about cleaning up your data, structuring your first projects, and building documents your teams will use every week.