Salesforce Launches Suite of AI Agents for Sales and Support Workflows

Salesforce Inc. has introduced a series of artificial intelligence agents designed to automate specific tasks for sales and technical support teams. The company states that these tools aim to increase productivity by handling repetitive administrative work, allowing human employees to focus on more complex interactions. Most of the new features are generally available immediately, with the remaining capabilities scheduled to launch by the end of the year.

The rollout includes a flagship agent named Hunter, which assists business-to-business salespeople in identifying leads and drafting outreach emails. According to reports from GN technics/ai (en-US), the agent is also capable of preparing presentations and generating proposals. This functionality is particularly relevant for long-cycle deals that can take weeks to close, as the system is built to maintain context over extended periods.

Long-term planning capabilities define new agent

A key component enabling Hunter's performance is a module referred to as the long-horizon runtime. This feature allows the agent to develop multi-step work plans and retain data across different chat sessions. This addresses a common limitation in conversational AI, where systems often lose track of previous interactions or fail to execute complex, multi-day workflows without human intervention.

However, this long-term memory capability is currently restricted to the Hunter agent. Salesforce has indicated plans to integrate this runtime into other agents in the future, but for now, it remains a singular feature. This creates a disparity in functionality between the different tools in the suite, meaning not all agents can currently handle extended, stateful tasks with the same degree of autonomy.

Broader suite covers support and internal operations

Beyond sales, Salesforce is deploying agents for customer support and internal operations. Casey and Fin are designed to handle support tasks such as processing product returns and answering consumer queries. Fin, in particular, leverages technology acquired through a recent multi-billion dollar acquisition, suggesting a strategic move to consolidate capabilities. Additionally, agents named Paige and Marshall target internal help desks and supply chain teams, respectively, automating routine HR and logistical requests.

The company also launched Piper and Carter, which are intended for website integration. Piper directs business prospects to the appropriate sales representative, while Carter assists online retailers by answering product questions and facilitating checkout. This expansion indicates a push to embed AI directly into the customer-facing digital experience, rather than limiting it to back-office processes.

Customization tools aim to reduce AI errors

To address concerns about AI reliability, Salesforce is releasing a programming syntax called Agent Script. This tool allows developers to define strict rules for how agents must execute specific tasks. By enforcing these constraints, the company aims to reduce the likelihood of hallucinations, where AI systems generate incorrect or fabricated information. This is a critical trade-off, as overly strict rules can limit the flexibility and natural conversational flow that users might expect from an AI assistant.

These agents operate within an updated version of Agentforce Coworker, which includes new features for orchestrating multiple sub-agents. This multi-agent orchestration allows complex tasks to be split among specialized components, with a continuous optimization process to monitor output quality. While this increases efficiency, it also adds a layer of complexity that requires careful management to ensure that the various agents work together without conflicting instructions or data inconsistencies.