Agentforce is Salesforce's AI agent layer: agents that handle customer service, qualify pipeline and run processes on top of CRM data, escalating to humans when judgement is needed. It became generally available with the Winter '27 release on 12 October 2026, for organisations with eligible SKUs. The definition is one line. The decision it opens is not. UK Salesforce customers now have agent infrastructure in their org, and the market moved first: 313 live UK tech postings mention Agentforce, for a platform that only reached general availability this week.
The demand came first. Before the general availability date, UK employers were already hiring against the platform: Accenture UK is advertising a Salesforce Solution Architect for financial services and a Service Cloud lead, and Salesforce itself is hiring Forward Deployed Engineers for its public sector practice. The platform landed on Monday. The teams to make it earn its keep have been scarce for months.
TL;DR. Agentforce is Salesforce's platform for AI agents: ready-made agents for service and sales, an Agent Builder to create your own, and the data and governance layer underneath, with the Atlas Reasoning Engine doing the thinking. It is generally available since 12 October 2026 in the Winter '27 release for eligible SKUs. UK demand is already real: 313 live tech postings mention Agentforce. What it is, what changed on 12 October, and the team it takes to use it.
What Agentforce actually is
Three pieces, described on the official Salesforce page. First, agents built for specific jobs: customer service with autonomous resolution and hand-off to a person, plus sales assistants that prepare meetings and chase opportunities; second, a builder, Agent Builder, where the organisation defines what the agent may do, which knowledge it draws on and when it escalates, on top of orchestration tools such as Flow. Third, the substrate: CRM data combined with Data Cloud and the governance of Salesforce's Trust Layer, which decides what each agent can see.
The engine underneath is the Atlas Reasoning Engine. A chatbot walks a decision tree. An Agentforce agent reasons about the case, picks the action and executes it against your systems. Answering a question, or closing the ticket. That is the difference.
One vocabulary note, because name confusion costs money in board meetings. Agentforce is the agent platform. It is not the consumer chat assistant, it is not the old Einstein support bot and it is not another CRM licence. When a posting asks for an Agentforce Architect, it wants someone who designs these agents, their data plumbing and their limits.
What changed on 12 October
Being in the catalogue and being in your organisation are different things. With the Winter '27 release, generally available from 12 October 2026, the platform reaches organisations with eligible SKUs, per Salesforce's own release information. For UK teams already on Salesforce, the practical question is not whether Agentforce will touch them. It is which processes get handed over first, under what criteria, and who answers for what the agent does.
That question has an uncomfortable answer. The platform ships controls; the real governance does not ship at all. Which data the agent sees, when it escalates, how its work is measured: all of that is decided inside each organisation, and it is team work, with product, service, data and platform engineering at the same table. The launch is the starting gun. It is not the race.
The week all three giants shipped agents
Agentforce did not land alone. Within days, the three dominant enterprise stacks have all made their move: Microsoft with its agents across the workplace suite, Salesforce with general availability on 12 October, and Google with the Gemini Agent presented at Gemini at Work. Three launches. One message. Agents have stopped being a pilot and become standard infrastructure, whether the team that has to run them is ready or not.
For a technology leader, the useful consequence is this: the stack decision now includes the agent layer, and every stack shares the same bottleneck. Profiles that can build, govern and measure agents on top of company data. If your organisation is growing that capability, the piece on what an AI builder actually does covers the adjacent ground.
What the UK market is already asking for
Here is the number no vendor announcement contains: measured through a full-text search of live UK tech postings, with the sample verified one by one, Shakers' market analysis counts 313 positions that mention Agentforce (October 2026). The titles draw the profile: Accenture UK wants a Salesforce Solution Architect in financial services and a Service Cloud lead, Salesforce itself is hiring Forward Deployed Engineers for public sector, and the surrounding consultancy postings list agents as project material. The skills that repeat are Apex, Flow and CRM depth, plus the ability to design a conversation and its boundaries. Classic platform programming, yes. Agent architecture, too.
The reference team has four slots, and none of them is optional.
- A Salesforce platform developer fluent in Apex and Flow, who builds the actions the agent executes.
- An agent architect, deciding which processes to automate, with which data and where the human edge sits.
- A data and governance lead, because the agent inherits whatever mess the CRM is in.
- A business owner per process. Service or sales. Defining what is worth automating, and whether the agent is doing it well.
The first two slots are the hardest to fill internally. The profile that joins the Salesforce platform to the craft of building agents is new, training does not produce it at the pace of demand, and every week adds postings to the queue. It is the implementation gap of this wave: the platform activates in a day, the last mile depends on accredited talent.
Frequently asked questions about Agentforce
- What is Agentforce?
-
Salesforce's platform for AI agents: ready-made agents for service and sales, the Agent Builder for creating your own, and the underlying data and governance layer of Data Cloud and the Trust Layer, with the Atlas Reasoning Engine as the reasoning core.
- Is Agentforce available in the UK?
-
Yes, as part of the global rollout. Agentforce became generally available with the Salesforce Winter '27 release on 12 October 2026 for organisations with eligible SKUs, and early access programmes existed before that date.
- How much does Salesforce Agentforce cost?
-
Salesforce prices Agentforce by consumption: from £1.60 per resolved conversation, or Flex Credits at £400 per 100,000 credits with each agent action costing 20 credits (Agentforce pricing page, October 2026). There is also a metered Agentforce User Licence at 5 US dollars per user per month, add-ons from 125 dollars per user per month for unmetered employee use, and a free tier with Agentforce Builder inside Salesforce Foundations.
- What is the difference between an agent and a chatbot?
-
A chatbot walks a closed tree of answers. An Agentforce agent reasons about the case, chooses an action and executes it against company systems, with governed access to CRM data and escalation to humans when judgement is required.
- What roles does the UK market ask for around Agentforce?
-
Platform developers with Apex and Flow, agent architects, data and governance leads, and business owners per process. Shakers' market analysis counts 313 live UK tech postings mentioning Agentforce in October 2026, with Accenture UK and Salesforce itself among the employers.
The platform is here and the clock is running for anyone competing on Salesforce: every process an agent absorbs is capacity for work that does need judgement. Building that last mile in-house takes months. Doing it by project, with accredited Salesforce platform talent, takes weeks.
Two pieces complete the picture: what an AI builder does in organisations beyond one stack, and what JEV means for software work when agents start writing code.
