The Playbook looks at stories from various industries to see what problems AI is solving, who’s driving these initiatives, and how it’s reshaping strategies.
Salesforce, well known for its CRM software used by over 150,000 companies like Amazon and Walmart, is no stranger to innovation. It also owns Slack, the popular workplace communication app. Salesforce is now stepping up its AI game with Agentforce, a platform that lets businesses to build and deploy digital agents to automate tasks such as creating sales reports and summarising Slack conversations.
What problem is it solving?
Salesforce has been working with AI for years. In 2016, it launched Einstein, an AI feature baked into its CRM platform. Einstein handled basic scriptable tasks, but the rise of generative AI brought a chance to do more. Smarter tools could now make better decisions and understand natural language.
This sparked a transformation. First came Einstein GPT, then Einstein Copilot, and now Agentforce—a platform designed for flexibility with prebuilt and customisable agents to handle diverse business needs.
“Our customers wanted more. Some wanted to tweak the agents we offer, while others wanted to create their own,” said Tyler Carlson, Salesforce’s VP of Business Development.
The tech behind it
Agentforce is powered by Salesforce’s Atlas Reasoning Engine, developed in-house. The platform connects with AI models from major players like OpenAI, Anthropic, Amazon, and Google, giving businesses access to a variety of tools.
Slack has become a testing ground for these AI agents. Currently in beta, Agentforce’s Slack integration puts automations where employees already spend their time. “Slack makes these tools easy to use and accessible,” Carlson added.
Smarter, more flexible AI
Agentforce uses ReAct prompting, a technique that helps agents break down problems into smaller steps and adjust their approach as they go. This leads to more accurate responses and hands-off task management, from answering questions to scheduling meetings.
Agentforce works with Salesforce’s proprietary LLMs and third-party models, giving clients plenty of options. To ensure security, Salesforce enforces strict data privacy policies, including limits on data retention.
Making it work for businesses
With tools like Agentbuilder, companies can design AI agents tailored to their needs. For example, an agent could sort emails or answer specific HR questions using internal data. One example is Salesforce’s collaboration with Workday to create an AI service agent for employee queries.
Salesforce is already seeing results, with Agentforce resolving 90% of customer inquiries in early trials. The goal? Broader adoption, more capabilities, and higher workloads handled by these agents.
“We’re building a bigger ecosystem of partners and skills,” Carlson said. “By next year, we want Agentforce to be a must-have for businesses.”
