Salesforce Inc.’s MuleSoft integration platform today announced support for a new event-driven application programming interface tool that will empower enterprise organizations to build artificial intelligence agents with real-time data.

Now in general availability, the MuleSoft support is for AsyncAPI, an industry-standard open-source initiative designed to provide connections for delivering real-time data through event-driven architectures. Using this standard, AI systems can react to changes in internal and external systems such as modifications to data, business rules, prebuilt automations or other signals.

MuleSoft said that real-time events are common for driving data to customers and users. However, the company cited a recent report revealing that 43% of information technology leaders have not managed to integrate traditional systems with their EDAs.

Salesforce recently released Agentforce, a suite of tools for building and customizing AI agents. Unlike traditional chatbots, AI agents can take action on their own, without human intervention, and can be triggered by events. The integration of simpler methods for consuming events provides enterprise developers with a key tool for building better agents.

Using a tool such as AsyncAPI allows AI agents to work asynchronously, meaning that they do not need to ask for an update and wait. Instead, an agent can link up with an event-driven system during its process, ask to be updated if something changes and go on with its work.

Andrew Comstock, vice president of product at Salesforce, told SiliconANGLE in an interview that makes this very powerful for AI agent use cases.

“If you think about the way development worked in the past, most APIs are measured in microseconds and milliseconds,” Comstock said. “Well, an agent might be doing a very complex task that’s a lot longer.”

Comstock explained that by using the new standard, businesses would be able to build onto their event-driven architectures reliably to access real-time data streams. Agents would then be able to take action based on changing information and adjust their responses in real time as soon as an event happens. An example is if some important data collected earlier in the process was updated that would affect the agent’s reasoning — similar to how people adjust their thinking when new information comes in.

The other automation use case enables AI agents to be activated by real-time events with standardized information provided by the API and a data source so that they can take immediate action. “This is where the rubber meets the road in the desire for agents,” Comstock said.

Numerous industries would benefit from this standardization and incorporation of real-time adjustments. For example, a customer support agent might be working on a call to resolve an inventory issue when new stock comes in while the agent is processing information in the background. If the warehouse database updates while the agent is working that resolves the problem, it can quickly change course. Another agent might be triggered to take action when certain inventory thresholds fall below certain levels and inform team members or call for new orders.

Real-time event data allows AI agents to take automated actions to reduce risks and prevent service interruptions in information technology systems. For example, an agent monitoring compute with historical data on usage spikes could receive events for sudden increases in resource use and spin up backup servers without human intervention.

Comstock explained that many businesses that didn’t have access to or the skills in this technology before have begun building out now that the tooling is more available.

“By enabling products like MuleSoft we allow them to get into technology spaces that they might not have otherwise been able to,” Comstock said. “I see that as being a potentially big unlock where they might not have felt comfortable or felt they had the skills. Using a tool like MuleSoft to supplement the skills that they have is an empowering capability for allowing these types of use cases and agentic AI integrations.”

Image: SiliconANGLE/Microsoft Designer