Agentforce Coworker: What It Is and How to Roll It Out
Agentforce Coworker brings an AI teammate into every Salesforce search bar. Here is how it works, how it gets enabled, what admins should govern and how to roll it out.
.png)
Agentforce Coworker is Salesforce's AI teammate for employees. It lives in the Salesforce search bar, answers questions using your CRM and Data 360 data, and can plan and run multi-step work through Agentforce agents, flows and actions, always within the permissions each user already has. For orgs with Unmetered User-Based AI seats, Salesforce began enabling it automatically on a rolling basis, so admins need to decide now how to configure, govern and roll it out rather than wait for users to discover it. This guide explains what Agentforce Coworker is, how automatic enablement works, what admins should configure first, and a practical rollout plan for mid-market teams.
What is Agentforce Coworker?
Salesforce describes Agentforce Coworker as an autonomous AI teammate built directly into Salesforce and powered by Data 360. Instead of opening a separate chat window, users click Ask next to the global search bar and describe what they need in plain language. Coworker figures out which data, agents and actions are involved, and either answers or carries out the work.
In the Salesforce Help FAQ on automatic enablement, the experience is split into two modes:
- Ask Mode: AI-generated answers grounded in records like Accounts, Contacts and Knowledge.
- Search Mode: Agentforce responses shown alongside standard search results, so users still get the familiar list of records.
The important shift is from "search and click" to "describe the outcome." A sales rep can ask for a catch-up on an account before a call, a service lead can ask which open cases are at risk, and a manager can ask Coworker to plan follow-ups on overdue deals.
How Agentforce Coworker works
Under the hood, Coworker combines three things your org may already have: your data, your automation and your security model.
1. It grounds answers in trusted data
Coworker searches CRM records (Accounts, Opportunities, Contacts, Cases and custom objects) and data connected through Data 360. According to Salesforce, Slack can optionally be connected as a source, limited to messages each user can already access. Salesforce's announcements also reference connectors to Google Workspace, Microsoft 365 and more than 270 other sources through Data 360.
2. It plans and acts through your existing automation
Coworker does not replace your agents and flows. It orchestrates them. When a request needs action, it plans the steps and invokes the Agentforce agents, flows and actions you have exposed. That means the quality of Coworker's actions depends heavily on the quality of the automation behind it.
3. It inherits your permissions and governance
Salesforce states that Coworker inherits the permissions, governance and security controls already in your org. Search results and actions respect the same access rules as the source systems, and Data 360 adds controls like PII masking, toxicity detection and result grounding. In practice, if your sharing model is too open, Coworker will surface that problem quickly.
Where it shows up
The core experience is the Salesforce search bar, and it is also available to authenticated Experience Cloud users. In its November 5, 2026 press release, Salesforce announced Microsoft Teams, an iOS app and desktop apps for Mac and Windows, plus planned support in industry mobile apps such as Field Service. The same announcement reported 21x usage growth in three months and more than 185,000 users. Check current availability for each surface before you promise it to users, since Salesforce rolls these out in stages.
How to enable Agentforce Coworker (including automatic enablement)
There are two paths: Salesforce turns it on for you, or you turn it on yourself.
Automatic enablement
Salesforce's help article says that starting August 10, 2026, Agentforce Coworker is being enabled automatically, on a rolling basis, for customers with Unmetered User-Based AI seats, with no additional setup or cost. Admins receive an email before it goes live and see an in-product banner, and users see an onboarding banner when they log in. Einstein must be active, since it powers the AI features.
Some orgs are excluded from automatic enablement, according to the same FAQ:
- Orgs that have opted out of generative AI.
- Multi-data-space or multi-org topologies.
- HIPAA, healthcare and life sciences, and government orgs.
- Orgs on Flex usage-based seats without Unmetered User-Based AI entitlements.
If you are in a regulated industry, note that Salesforce states Coworker is not currently covered by its Business Associate Addendum.
Opting out or turning it off
- Before enablement: use the opt-out link in the notification email, or go to Setup, open Agentforce Coworker and choose to opt out of auto activation.
- After enablement: turn the Agentforce Coworker toggle on or off from the Agentforce Coworker setup page at any time.
Manual enablement
If your org is not auto-enabled, an admin can turn Coworker on from Setup by searching for Agentforce Coworker, clicking Get Started and switching it on. Salesforce's general FAQ lists the main prerequisites: an eligible Agentforce plan or allocated Flex Credits, the AI Search Setup permission set license for admins, and the right permission sets for users. Multi-data-space orgs also bind Coworker to a single data space.
Does it cost extra?
For users with Unmetered User-Based AI seats, Salesforce says auto-enabled Coworker comes at no additional cost, and CRM search plus optional Slack search do not consume Flex Credits or Data Services Credits for Agentforce 1 Edition or Agentforce for Sales, Service or Industries seats. Data 360 ingestion, processing and indexing still consume credits regardless of plan. Salesforce's November announcement also lists a per-user, per-month SKU at $50 per user per month, noting that pricing and packaging are subject to change. Confirm your specific entitlements with your Salesforce account executive before rolling out widely.
What admins should configure and govern
Turning Coworker on takes minutes. Making it useful and safe takes deliberate setup. These are the areas we would review first.
User access
- Decide who gets Coworker first rather than defaulting to everyone. A pilot group gives you feedback before the whole company forms an opinion.
- Remember that users who gain entitlements after enablement see standard search until an admin grants access, so build access into your onboarding process.
- Assign the user-level permission sets Salesforce documents for Coworker and make sure the right AI licenses are applied, so usage is billed the way you expect.
Sharing model and data visibility
- Audit org-wide defaults, sharing rules and profile permissions on sensitive objects. Coworker respects them, which means it also exposes anything that was shared too broadly.
- Review field-level security on compensation, health, financial or other confidential fields.
- Check which Slack channels and external sources you connect, and confirm each one is permission-aware or governed manually.
Data sources and quality
- Start with the sources that answer your users' most common questions, not every connector available.
- Clean up duplicate accounts, stale opportunities and outdated Knowledge articles. AI answers are only as good as the records behind them.
- Track Data 360 credit consumption from ingestion and indexing in Digital Wallet.
Actions and automation
- Inventory the agents, flows and actions Coworker can invoke, and remove or restrict anything that should not run on a casual request.
- Require confirmation for actions that update records in bulk, send external communications or change ownership.
- Use the approval and audit capabilities Salesforce provides for supported operations, and review logs during the pilot.
Known limits to plan around
Salesforce's FAQ lists several beta limits, including English-only support, files over 75MB not being indexed and no out-of-the-box calendar connectivity. Seat-based licenses also have a usage guideline of 30 LLM calls per minute per user. Share these with your pilot group so expectations match reality.
A practical Agentforce Coworker rollout plan for mid-market teams
Mid-market teams rarely have a dedicated AI team, so the rollout needs to be light, staged and tied to real work. This is the sequence we recommend.
- Confirm your status (week 1). Check whether your org is eligible for automatic enablement, look for the notification email or Setup banner, and decide whether to let it activate, opt out temporarily or enable it manually on your own timeline.
- Run a permissions and data audit (weeks 1 to 2). Review sharing, field-level security and the sources Coworker will search. Fix the obvious gaps before anyone types a question.
- Pick three to five high-value use cases (week 2). Good starting points are account catch-ups before meetings, pipeline risk summaries, case triage for service leads and finding the right Knowledge article. Write down what "good" looks like for each.
- Launch a pilot with 10 to 25 users (weeks 3 to 6). Mix power users and skeptics across sales and service. Give them example prompts, a feedback channel and a short weekly check-in.
- Tune agents, actions and sources (weeks 4 to 7). Use pilot feedback to adjust which actions are exposed, improve flows that fail, and add or remove data sources. This is where most of the value is created.
- Train by role and expand in waves (weeks 7 to 10). Roll out team by team with role-specific prompt libraries. Short live sessions beat long documentation.
- Measure and govern continuously. Track adoption, time saved on target tasks, answer accuracy and credit consumption. Review new Salesforce releases each season, since Coworker's capabilities and surfaces are changing quickly.
If you want a framework for proving value, our guide on how to measure the ROI of your Salesforce investment applies directly to AI features like Coworker.
Common mistakes to avoid
- Letting auto-enablement be the rollout plan. If users find Coworker on their own with no guidance, first impressions are random and hard to change.
- Skipping the data cleanup. Coworker will confidently summarize bad data. Our article on why your Salesforce org is ready for AI agents but your data isn't covers what to fix first.
- Exposing every action on day one. Start read-heavy, then add actions that write data once you trust the results.
- Ignoring licensing details. Users without the right AI seats or permission sets may consume credits or not see Coworker at all.
- Treating it as a one-time project. Salesforce is adding surfaces, skills and connectors frequently. Assign an owner to keep the configuration current.
How Agentforce Coworker fits into your AI strategy
Coworker is the employee-facing front door to Agentforce. Your customer-facing agents, internal agents, flows and Data 360 investments all become more visible once employees can reach them from the search bar. That makes Coworker a good forcing function: it rewards orgs with clean data, clear permissions and well-built automation, and it exposes the gaps in orgs without them. For more context on how Salesforce's AI layers fit together, see our explainers on what AIforce is and what was worth paying attention to at Dreamforce.
Hikko is also a Claude partner, so if your team is evaluating how Coworker sits alongside other AI tools your employees use, our post on the Salesforce and Claude dual-partner advantage is a useful next read.
Want help rolling out Agentforce Coworker the right way? Hikko's Salesforce team can audit your permissions and data, design your pilot and tune the agents behind it. Talk to our team.
Frequently asked questions
What is Agentforce Coworker in Salesforce?
Agentforce Coworker is Salesforce's AI teammate for employees. It appears next to the global search bar and lets users ask questions or request work in plain language. It answers using CRM and Data 360 data and can run Agentforce agents, flows and actions, while respecting each user's existing permissions.
How do I enable Agentforce Coworker?
If your org has Unmetered User-Based AI seats, Salesforce may enable it automatically and notify you by email first. Otherwise, an admin can go to Setup, search for Agentforce Coworker, click Get Started and turn it on, then assign the required permission sets to users. Einstein must be active.
Can I opt out of Agentforce Coworker automatic enablement?
Yes. Before it activates, use the opt-out link in Salesforce's notification email or choose to opt out of auto activation in Agentforce Coworker setup. After activation, admins can turn Coworker on or off from the setup page at any time.
Does Agentforce Coworker cost extra?
For auto-enabled orgs with Unmetered User-Based AI seats, Salesforce says there is no additional cost for Coworker itself, although Data 360 ingestion and indexing still consume credits. Salesforce has also announced a per-user, per-month SKU. Pricing and packaging can change, so confirm with your account executive.
Is Agentforce Coworker secure?
Coworker inherits your org's permissions and sharing rules, so users only see data they already have access to. Data 360 adds PII masking, toxicity detection and grounding. The main risk is an overly open sharing model, which is why a permissions audit should come before rollout.
How long does an Agentforce Coworker rollout take?
Technically, Coworker can be turned on in minutes. The staged plan above, including a permissions audit, a pilot and role-based training, fits in about eight to ten weeks for a typical mid-market team, depending on data quality and how much automation needs tuning.
Ready to stop fixing and start scaling?
Let’s discuss a post-implementation health check and a roadmap for maximizing your Salesforce ROI.
The views expressed in this article are those of the author and do not necessarily reflect the official policy or position of Hikko.
.png)
.png)
