Dreamforce 2026: What the Agentforce Announcements Mean for Your Salesforce Org
Dreamforce 2026 moved Salesforce's AI story from single chat assistants to agents that do real work: an AI coworker inside Lightning, ready-made agents for sales and service, a CRM-specific reasoning model called Koa, and new ways to use Salesforce data inside Slack and Claude. For most companies the practical message is simple: agents are now a standard part of Salesforce, and the orgs that get value first are the ones with clean data, clear permissions and one well-chosen use case. This guide explains the main announcements in plain language and what to do about them.
The headline announcements in plain language
Salesforce grouped many of its updates under the name AIforce, which brings Salesforce data, apps and workflows into the tools people already use while keeping existing permissions. The pieces most businesses will hear about are below. Availability differs by product and edition, so always confirm the current status in Salesforce's release notes before you plan a rollout.
- Agentforce Coworker: an AI teammate inside Salesforce that answers questions across CRM, Slack and other connected sources, and can plan and run tasks by coordinating other agents. Salesforce announced it as generally available.
- Ready-made job agents: Piper, an AI sales development agent that nurtures inbound leads and books meetings; Fin, a customer service agent across chat, email, WhatsApp, SMS, voice and Slack; and Hunter for outbound prospecting, announced for November 2026.
- Koa: Salesforce's first CRM reasoning model for Agentforce, trained on synthetic CRM scenarios and run inside Salesforce's own infrastructure. Announced as a pilot, with general availability expected later.
- Slackforce: Slack as a working interface for Salesforce records and actions.
- Claudeforce: Salesforce data and actions available inside Claude through MCP, limited to the data each user can already see. Announced as a beta.
- Long-horizon agents: agents that pursue goals over days or months, with approval checkpoints and memory across sessions. Announced as a pilot.
The tools for building and running agents got stronger
The less flashy announcements may matter more to admins and developers, because they make agents easier to test, monitor and improve once they are live.
- Agent Optimizer finds recurring failure patterns in real sessions and helps refine agents in Agentforce Builder.
- AI Skills are reusable, governed instructions that any agent can run, managed in Agentforce Studio.
- Prompt Builder adds conditional logic, batch prompt testing and visibility into the flows and actions a prompt depends on.
- Agent Health Monitoring, Custom Scorers, A/B Experimentation and Voice and Conversation Testing give teams proper quality control before and after go-live.
- An Upgrade Agent in Agentforce Labs converts an existing Einstein Bot into an Agentforce Service Agent while the original bot keeps running.
What this means for a typical Salesforce customer
You do not need to adopt every new product. The real shift is that agents now read and act on your CRM data at scale. That raises the stakes for three things most orgs have postponed: data quality, permission design and process documentation. An agent that books meetings from messy lead data, or that can see records a user should not see, creates problems faster than a human ever could.
- Sales teams: Piper and Hunter target the top of the funnel, so lead routing, duplicate rules and qualification criteria need to be agreed and written down first.
- Service teams: an AI service agent is only as good as your knowledge articles, case categories and escalation rules.
- Admins: permission sets, sharing rules and field-level security now control what agents can see and do, not only what people can see.
- Leadership: you need a short list of measurable outcomes, such as response time, meetings booked or cases resolved without escalation, before you switch anything on.
A 90-day plan after Dreamforce
- Weeks 1 to 2: list three candidate use cases and score them on volume, data readiness and risk. Pick one.
- Weeks 2 to 4: clean the data the agent will use. Merge duplicates, fill required fields and retire unused picklist values.
- Weeks 3 to 5: review permission sets and sharing for the agent user. Give it the minimum access it needs.
- Weeks 4 to 8: build the agent in Agentforce Builder, write clear topics and instructions, and test with real conversations, including edge cases.
- Weeks 8 to 10: launch to a small group or a single channel with Agent Health Monitoring switched on.
- Weeks 10 to 13: review failed sessions, refine instructions and decide whether to expand or try the next use case.
Licensing and cost: ask before you build
Agentforce pricing and packaging have changed several times, and Dreamforce added new products on top. Do not assume a feature is included in your current licence. Ask your Salesforce account executive which edition, add-on or consumption model each feature needs, and check the vendor's current pricing before you commit to a design that depends on it.
Frequently asked questions
What is Agentforce Coworker?
Agentforce Coworker is an AI teammate inside Salesforce that answers questions across your CRM, Slack and connected sources, and can plan and run tasks by coordinating other agents, while respecting existing Salesforce permissions.
What is Koa in Salesforce?
Koa is Salesforce's first CRM reasoning model for Agentforce. It is trained on synthetic CRM scenarios and runs inside Salesforce's infrastructure. It was announced at Dreamforce 2026 as a pilot.
Do I need Data Cloud (Data 360) to use Agentforce?
Many Agentforce features rely on Salesforce's data platform to ground answers in your data. Requirements vary by feature and edition, so confirm with Salesforce before you plan.
How long does a first Agentforce implementation take?
A focused first agent with clean data can typically be designed, tested and launched to a small group in about 8 to 12 weeks. Larger multi-agent projects take longer.
How Quant Tech can help
Quant Tech helps companies in North America, Australia and India turn Salesforce announcements into working results: use-case selection, data clean-up, permission design, Agentforce build and testing, and support after launch. If you want a practical view of which Dreamforce features fit your org, book a free consultation with our Salesforce team.
Talk to a Salesforce, HubSpot or Zoho expert—free 30-minute consultation.
