A closed deal becomes an order, an onboarding plan and a billing setup. A support ticket becomes an engineering handoff, a credit request and a customer reply. The month-end review draws on the ledger, email and a spreadsheet.

Existing software may handle parts of these jobs well. Someone still needs to know how the pieces fit: which information takes precedence, what must be confirmed and who receives the next step. Those operating rules belong in the workflow, wherever the information comes from.

What the vendors announced.

The announcements below describe ways to reach business-system data and actions from AI assistants. The dates and sources identify what was announced; each product's current availability should be checked for the plan and use involved.

  • Atlassian put Jira and Confluence inside Claude in a beta announced on 2025-05-01, inside ChatGPT on 2025-12-03, and declared the connection generally available on 2026-02-04.
  • HubSpot launched a connector for ChatGPT on 2025-06-04 and one for Claude on 2025-07-29, and on 2026-02-26 let the ChatGPT connector create and update CRM records.
  • Salesforce announced its apps for ChatGPT on 2025-10-14, and a sales app in ChatGPT was in open beta by 2026-01-07; with Anthropic it announced a Salesforce plugin for Claude on 2026-08-26, in beta on paid Claude plans from 2026-09-15, for organizations Salesforce approves.
  • Intuit announced its apps for ChatGPT on 2025-11-18 and its integrations for Claude on 2026-02-24, and on 2026-07-28 QuickBooks said invoices could be created and sent from the chat in both.
  • ServiceNow opened its platform's actions to outside AI tools, Claude among them, on 2026-05-05.
  • Microsoft's own documentation, updated 2026-06-05, describes connecting Dynamics data to Claude once an administrator enables it.

Our reading: access makes the workflow possible.

These announcements point toward AI assistants becoming another place to work with business information. Our view is that access and operating knowledge need to develop together. A connection can make a record available; the workflow still needs to know what to do with it.

For a closed deal, that includes which onboarding steps apply to a new customer, what changes for an existing one and how to handle a missing purchase order. An SOP may contain those rules. Often an operations lead supplies the final details from experience.

Start by understanding the job and checking what your current tools already handle. Then decide where an AI workflow can add useful preparation, whether through supplied files or a direct connection. For any connection, confirm access, permissions and approval behavior for the actions you intend to use.

Put the handoff to work from exports.

Illustrative example: a distributor's closed-deal handoff.

The operations lead supplies the CRM export and signed order forms. The workflow prepares order lines for fulfilment, an onboarding checklist and billing information, with missing details collected into questions for the salesperson. The team reviews and enters the records.

The build starts with the distributor's handoff procedure, good completed examples and the lead's explanation of the exceptions. Those define the approach. Each new export supplies the transactions to prepare.

This gives the team a useful first workflow before commissioning a new integration. They can see whether the prepared handoff saves effort and exposes the right gaps. Direct connections can follow as separately scoped work, based on what the team learns.

The opportunity is the complete job: getting the information into a form the next person can act on, with your way of working built in.