Salesforce did not launch a single new AI agent on 11 September. It launched seven, each with a name, a specific job description, and a defined function inside sales, service, commerce, IT, HR, supply chain and customer experience. Casey handles customer service across voice, SMS, WhatsApp and web chat. Paige covers IT and HR requests. Carter works the shopper path through to checkout. Marshall orchestrates back-office processes with what Salesforce describes as deterministic execution and a full audit record. Six of the seven are generally available now. This is not a product update. It is a statement about where the agent market is heading, and it changes the build versus buy calculation every enterprise evaluating agentic AI is currently making.
For the past two years, most enterprises building agentic AI capability have been doing exactly that, building. Custom agents, custom orchestration, custom integration into existing systems. Salesforce's move signals that at least one major vendor now believes the market is ready for prebuilt, named, job-specific agents rather than a generic agent-building platform alone. Whether that bet is right matters less than what it reveals: the packaged agent category is now real enough that every enterprise strategy conversation about agentic AI needs to explicitly weigh buying a named, purpose-built agent against building a custom one, rather than defaulting to custom by habit.
What "Job-Ready" Actually Means Here
The specificity is the point. Hunter is not a generic sales assistant, it is described as working a pipeline from research through to outreach, collaborating with human sellers over weeks and months using what Salesforce calls a long-horizon runtime, a departure from single-session chat interactions toward an agent that pursues a goal across an extended timeframe. Marshall is not a generic automation tool, it is positioned specifically around back-office process orchestration with deterministic execution and an audit trail of every action taken. Naming and scoping agents this tightly is a deliberate move away from the "general purpose agent" framing that dominated the market through 2025.
Why This Changes the Build Versus Buy Conversation
A named, prebuilt agent with a defined scope is a fundamentally different evaluation than a platform an organisation configures from scratch. The question shifts from "can we build this" to "does this prebuilt agent's defined scope actually match our process, and is the gap between the two small enough to configure rather than requiring a custom build." That is a more concrete, answerable question, and it means organisations evaluating agentic AI now have a genuine third option beyond the traditional build versus buy binary: buy a scoped agent and integrate it, rather than build a custom one or buy a generic platform and build on top of it.
The Audit Trail Detail Worth Noticing
Marshall's positioning around deterministic execution and a full audit record of every action is a governance detail as much as a product feature, and it is worth reading as a signal of where enterprise buyers are pushing vendors. An agent that cannot produce a clear record of what it did and why is becoming a harder sell to enterprise procurement and risk functions, regardless of vendor. Organisations building custom agents in house should treat this as a benchmark for what their own agents need to demonstrate, not just a competitive feature to note.
What to Actually Evaluate Before Choosing Either Path
The practical question for an organisation currently deciding between a prebuilt agent and a custom build is not which approach is generally better. It is whether a specific, named agent's defined scope maps closely enough to an actual internal process to make integration cheaper and faster than a custom build, and whether the organisation is comfortable with the platform dependency that choosing a prebuilt agent inevitably creates. Neither answer is universal, and the organisations getting this right are evaluating it process by process rather than committing to one approach across the board.
What This Means for Your Organisation
What we see across implementation engagements is that organisations often default to building custom agents because it feels like the more defensible, controlled choice, without actually costing out how much of that build effort is being spent solving a problem a prebuilt, named agent already solves adequately. The emergence of job-ready agents from a major vendor is a useful forcing function, it makes the build versus buy question concrete and specific rather than an abstract strategic debate.
Key Takeaways
- Salesforce launched seven named, job-specific AI agents on 11 September, with six generally available now and one, Hunter, in pilot using a new long-horizon runtime that pursues goals over weeks rather than single sessions.
- The shift toward tightly scoped, named agents rather than generic agent platforms adds a genuine third option to the traditional build versus buy decision: buying a scoped agent and integrating it.
- One agent, Marshall, is positioned specifically around deterministic execution and a full audit record, a governance detail that is becoming a benchmark buyers expect regardless of vendor.
- Organisations should evaluate build versus buy process by process, based on how closely a specific named agent's scope matches an actual internal workflow, rather than committing to one approach across the board.
How Trusenta Can Help
AI Strategy Enterprise evaluates build versus buy decisions process by process, so an organisation is not defaulting to a custom build where a scoped, prebuilt agent would serve the same need faster and cheaper.
Custom AI Development builds the agents genuinely worth building in house, with the same deterministic execution and audit record standard the market is now demonstrating buyers expect.
AI Integration Services connects prebuilt agents, where they are the right choice, into existing systems without creating unnecessary platform lock-in.
Conclusion
Whether Salesforce's specific bet on seven named agents succeeds commercially is almost beside the point. What matters is that a major vendor now believes the market is ready to buy job-specific agents rather than build generic ones, and that belief alone changes the conversation every enterprise strategy team is having about agentic AI. Organisations still treating build versus buy as an abstract, one-time strategic choice are behind the organisations already asking the more useful question, process by process.
