The adoption of AI agents has reached critical mass, according to new data from Salesforce, with fleets of autonomous “digital workers” being deployed across enterprise firms, but poor integration and weak interoperability are blocking progress.

Polling 1,050 enterprise IT leaders, the CRM giant’s latest Connectivity Benchmark Report found that organisations are now using an average of twelve agents, with the number projected to climb 67% within two years.

Already, the vast majority of firms (83%) report that most or all of their teams have incorporated agents into their workflows, a pace which has left IT leaders grappling with mounting challenges in their coordination and oversight.

The study found, for example, that half of all AI agents operate in isolated silos rather than as part of multi-agent systems, despite almost all IT leaders (96%) agreeing that the success of an agent depends on seamless integration across all systems.

App sprawl is adding to the complexity, with Salesforce discovering that firms use an average of 957 apps, only 27% of which are integrated together, leading to concerns that the infrastructure supporting AI agents is fundamentally flawed. 

Disconnected data is also presenting problems, with 96% of businesses experiencing barriers to using data for AI use cases, and 40% identifying outdated IT architecture and infrastructure due to data silos as a roadblock.

The majority of IT leaders (86%) believe this fragmented infrastructure will result in agents introducing more complexity and little real value, with 94% agreeing that success will require building in more APIs – the fundamental building blocks for connecting applications, data, and AI.

While a third of teams are leveraging APIs to speed up integration across systems, and 50% are using APIs to connect and govern AI, around a quarter of APIs (27%) are currently ungoverned, a number which should shrink as companies build out the infrastructure to support multi-agent systems.

Recommended reading

• UK Leaders Overestimate AI Readiness, Study Finds
• AI Agents Are Transforming Enterprise AI Adoption
• Report: 1.5 Million Enterprise AI Agents At Risk of ‘Going Rogue’

Aside from more layers of APIs, Salesforce also found growing interest in a range of standards or protocols designed to connect and manage agents, including Agent Network Protocol (43%), Agent-to-Agent Protocol (40%), and Universal Tool Calling Protocol (34%).

“The true success of an agentic enterprise isn’t found in the sheer number of agents deployed but the overall effectiveness of those agents,” said Andrew Comstock, SVP and GM of MuleSoft at Salesforce.

“We need to think about how they are discovered, governed, and orchestrated to work together. As we move into this multi-agent era, the role of IT is evolving from managing silos to building a unified foundation as the central control plane that allows multi-agent systems to be safe, reliable, and scalable.”