Companies are embracing the artificial intelligence revolution, turning to AI agents to help them scale for the future. That’s good news for enterprise agentic platform providers like Salesforce, Zendesk, and ServiceNow and startups like Cognition Labs, Adept AI, and LangChain. However, while there’s tech adoption, it’s not being smartly executed.
New research from MuleSoft, Venson Bourne, and Deloitte Digital shows that IT leaders are eager to adopt AI agents. In fact, 93 percent plan to implement them within the next two years. Yet, they still face two main barriers: An inability to keep up with the growing demand for AI and difficulty integrating the tech with key data systems.
IT Leaders Want AI Agents
“Integration challenges hinder companies from fully realizing the technology’s potential to create a limitless digital workforce, which can significantly alleviate IT workloads,” Andrew Comstock, Salesforce’s General Manager for MuleSoft, remarked.
He continued, “80 percent of responders say that data integration is one of the biggest challenges organizations face when using AI. Therefore, integration is incredibly foundational to making AI agents work. This is because AI agent outputs depend on connected data that enables a comprehensive understanding of the context and nuances within user queries. These agents gather structured and unstructured data from diverse sources…and use it to make decisions and take actions for any business processes across systems.”
Interest in AI remains high, with companies planning to use the tech more. To meet this demand, investment in IT staffing is growing, rising 61 percent from $10.5 million in 2023 to $16.9 million last year. “Ninety-three percent of IT leaders say that AI will increase the productivity of their developers in the next three years—so it’s not too far away,” Comstock stated, emphasizing that 98 percent of respondents support agentic automation within their organization.
That being said, although more resources are pouring in, the percentage of IT projects failing to be completed on time is growing. MuleSoft reports that the percentage of efforts not being delivered on time grew by three percentage points to 29 percent.
What’s Holding Up AI Projects?
Rolling out AI agents within an organization is one thing. Ensuring they have access to the right data to deliver high-quality responses is another. A company’s complex infrastructure is cited as the reason behind the sluggish execution. The existence of data silos—repositories of information that are useful for a business unit but are inaccessible to other departments that might also benefit from the data—impede AI agents’ ability to operate effectively.
“Organizations today continue to use 897 apps on average, which is an incredible number,” Comstock shared. “Only 29 percent of applications are typically connected within the organization, which is impacting the accuracy and usefulness of AI agents.”
Underscoring the importance of proper integration: “Integration and APIs directly enhance an AI agent’s performance by enabling to access critical business specific data and interact directly with their existing systems, automation, and other agents across the enterprise, so that they don’t have to refit everything for the AI world. Among organizations with agents, those using APIs are taking advantage of their capabilities to improve their IT infrastructure, enable data sharing across teams, and integrate disparate systems. Also, on average, 50 percent of an organization’s internal software assets and components are available for internal reuse, and this is incredibly significant because companies can leverage their reliable data and existing investments with APIs to unlock the full potential of their data to efficiently power…AI agents.”
MuleSoft’s survey of 1,050 enterprise IT leaders highlighted a widespread need for better systemic integration across teams. From engineering and customer service to business analysts, marketing, human resources, finance, and product, all departments within an organization want more data sharing.
This trend’s percentage was already high in 2024, with nearly three-fourths of respondents calling for improved integration. A year later, companies don’t seem to be doing enough, as the percentages have risen to around 80 percent. And with AI firmly taking hold of the business world, organizations risk further falling behind if they don’t break down their data silos.
“In the agentic era, IT leaders are seeking ways to meet the increased demand for efficiency and productivity,” Beena Ammanath, the head of Deloitte’s Global AI Institute, said. “This strain on their teams is growing as they navigate balancing current capabilities with business aspirations: steadily incorporating AI and autonomous agents across an ecosystem of hundreds of distinct applications while also maintaining those same systems. Leading organizations must establish a proactive integration strategy for unifying the entire IT estate, encompassing apps and systems, autonomations, and APIS, all of which are vital for driving revenue growth and reducing operational costs.”
Subscribe to The AI Economy
What Salesforce Has Been Saying For A While
This research is based on MuleSoft’s Connectivity Benchmark Report. For more than a decade, the Salesforce-owned company has been tracking the state of enterprise tech integration. It’s fitting that it comes from MuleSoft, which specializes in helping companies connect applications, data sources, and devices within their IT infrastructure—a middleware provider.
This year’s study coincides with the AI narrative Salesforce has been pitching to customers and investors since it launched its Agentforce platform. Salesforce Chief Executive Marc Benioff and others on his team have pontificated about this inflection point where artificial intelligence will help organizations better connect with their customers in a “whole new way.”
“This is a moment where technology can really step in. This is the moment where we could ask really meaningful, powerful questions like, ‘What if these workforces had no limits at all?’” Benioff boasted in September.
MuleSoft is an integral part of the Agentforce platform. Although Salesforce will likely house a significant amount of a company’s data, there are third-party and legacy platforms that AI agents need to be truly efficient and effective. MuleSoft’s role as a connector of these services to Salesforce and by extension, Agentforce, eliminates the data silo obstacles that plague IT teams from properly executing AI projects.
It’s in Salesforce’s best interest to raise the integration red flag now. After all, if companies fail to take action, Benioff’s vision of the digital labor force created by autonomous AI agents won’t be realized. And although MuleSoft’s report doesn’t mention it, it’s hard not to see the subtle reference to Agentforce baked into it. When it comes down to it, Salesforce is telling readers that IT leaders’ plight can be remedied by switching over to its agentic development platform.
Alternatively, if Salesforce wants to lead the enterprise’s AI adoption, it must proactively educate companies about adequately implementing the technology. Absent this leadership, there will be a power vacuum, and organizations will be left with a hodgepodge of best practices and left to their own devices to figure things out. Ultimately, this will result in an ecosystem without standards and success, creating more doubters than proponents.
When asked what enterprise organizations should do to become so-called agent-first companies, Comstock provided this advice:
“You want to make sure your processes, your actions that you want the agents to be taking with the data, are as a great of place as possible. And this is something that many enterprises have been doing as part of their digital transformation that we mentioned with the cloud, where they’re looking at trying to bring those processes in-house and instantiate them in technology. And if you have, I think that’s an incredible starting place for getting ready for agents. And if you haven’t, you still have plenty of time to be investing in products to enable that. so you can build on top of that sophistication. Agents, with their independent actions and reasoning, are going to accelerate bad decisions if they’re built on bad data and bad actions. And so, the stronger that you have as a starting point, the stronger that foundation is, the stronger your agents will be down the road.”
To read MuleSoft’s 2025 Connectivity Benchmark Report, click here.
Featured Image: An AI-generated image of a robot sitting at a cluttered office desk working on a computer. Credit: Adobe Firefly
Leave a Reply
You must be logged in to post a comment.