Use of AI agents will increase by 67% by 2027

Businesses are rapidly expanding their use of AI agents, but unless they modernize how they connect systems, data, and applications across their organizations, they risk undermining their efforts, according to a new report from Salesforce.

Salesforce’s 11th Annual Connectivity Benchmark Report shows that enterprises are now using an average of 12 AI agents. We predict this number will increase by 67% within two years as companies move toward what Salesforce calls the “Agent Enterprise,” where humans and AI agents work together across workflows. But half of those agents are now operating alone rather than as part of a coordinated multi-agent system, creating fragmented automation, governance risks, and what IT leaders describe as the rise of “shadow AI.”

The findings are based on a survey of 1,050 IT leaders conducted between October and November 2025 by research firm Vanson Bourne in partnership with Deloitte.

According to Digital Commerce 360 data, 76 of the top 2000 online retailers in North America use Salesforce as their e-commerce platform. Top 2000 is a Digital Commerce 360 database that ranks North America’s largest online retailers by annual e-commerce sales.

Nearly all respondents (96%) said the success of their AI agents relies on seamless data integration between systems. At the same time, of the average 957 enterprise applications, only 27% are currently integrated. This highlighted what respondents claim is a widening gap between orchestration and governance.

“Agents are no longer experimental,” Andrew Comstock, Salesforce senior vice president and MuleSoft general manager, said in the report. “The real question for IT is how to discover, manage, and coordinate IT so that it works together as a system rather than as a disconnected tool.”

This report shows that companies are developing AI agents through multiple channels.

36% pre-built SaaS agents.
34% are embedded in enterprise platforms.
30% custom built agents in-house.

This diversity creates new management challenges for IT teams.

Concerns about introducing AI agents into workflows

86% of IT leaders say they are concerned that without a stronger integration framework, agents could bring more complexity than value. 42% cite risk management, compliance, and legal concerns as the biggest barrier to agent transformation. Lack of in-house AI expertise (41%) was followed by legacy infrastructure (37%) and difficulty integrating siled data (35%).

Data governance has emerged as a central issue. On average, 27% of enterprise APIs are considered unmanaged. And only 54% of organizations report having a centralized governance framework that formally oversees AI and agent functionality. Half of respondents (49%) cited data governance between applications as their biggest integration challenge.

To fill these gaps, IT leaders are turning to API-driven architectures as the foundation for multi-agent environments. 94% said that the success of future AI agents will require IT architectures to become more API-centric, with APIs acting as the connection layer between applications, data, and AI systems.

One-third of teams say they are already using APIs to accelerate system integration. Additionally, 50% report that they currently use APIs to connect and manage their AI capabilities.

Set standards for agent AI

Interest in new agent communication standards is also growing. Respondents said they are evaluating or planning support for protocols that allow agents to share context and collaborate securely. They include:

agent network protocol
Agent communication protocol
Agent-to-agent protocol
model context protocol
Universal tool calling protocol

“This is a tipping point,” Kurt Anderson, managing director and API transformation leader at Deloitte Consulting LLP, said in the report. “Enterprises need to go beyond simply deploying agents and enable them to operate at scale through sustainable and secure integration strategies.”

The report cites early adopters who have moved beyond experimental use using the unified agent framework.

AstraZeneca leverages Salesforce’s Agentforce Life Sciences platform and MuleSoft’s integration capabilities to coordinate AI agents across field efforts, commercial operations, and regional brands to improve interactions with healthcare professionals.

r.Potential, an enterprise intelligence company, combines the Salesforce platform with multiple expert agents and model context tools to generate executive-level workforce insights supported by a managed API foundation.

Despite rapid adoption, 96% of organizations reported barriers to using data for AI use cases. 40% cite outdated IT architectures caused by data silos and disconnected systems as a key deterrent.

The number of applications in the enterprise also increased year over year from 897 to 957, further increasing integration complexity.

The report concludes that without unified integration and governance, businesses risk building a chaotic network of intelligent tools that cannot work together effectively, limiting the productivity gains that AI agents aim to provide.

sign up

Please sign up for Free subscription to Digital Commerce 360 B2B News. Covering technology and business trends in the growing B2B e-commerce industry. Contact Mark Brohan, Senior Vice President of B2B and Market Research. mark@digitalcommerce360.com. Follow him on Twitter @markbrohan. Follow us on LinkedIn. X (old Twitter)Facebook, YouTube.