According to research from Salesforce, as enterprises deploy AI agents, gaps in orchestration and governance are emerging. among them Connectivity Benchmark ReportAccording to Mulesoft, a Salesforce company, organizations surveyed have an average of 12 agents, and that number is expected to grow by 67% over the next two years.
This is the 11th vendor report, produced in collaboration with Vanson Bourne and with input from Salesforce’s Mulesoft partner Deloitte Digital. The study surveyed 1,050 IT professionals in the US, UK, Ireland, France, Germany, Netherlands, Australia, Singapore, Hong Kong, and Japan between October and November 2025. The UK cohort had 100 respondents.
We found that 89% of organizations in the UK and Ireland have already deployed AI agents, but half of these agents are divided into ‘silos’ and therefore not meaningfully deployed at the enterprise level. Only 54% of organizations have a centralized governance framework to formally monitor agents. Approximately 94% of respondents agreed that successful AI agents will require new IT architectures that are more API-driven.
Businesses use an average of 796 applications, 33% of which are integrated with each other, yet 75% of respondents are concerned that agents create more complexity than business value.
The majority (97%) of organizations say there are barriers to using data for AI use cases, with 35% identifying outdated IT architecture and infrastructure with data silos and disconnected systems as the biggest barrier.
Beena Ammanath, Global Deloitte AI Institute Leader, said in the report’s foreword: “The speed of AI adoption is faster than predicted a year ago, with 84% of enterprise CIOs believing AI will become as important to their business as the internet.
“The next era of AI is now here: the agent-based enterprise. 40% of organizations report deploying autonomous agents to enhance their current processes and teams, and an additional 41% plan to deploy agents within the next year.”
At a press conference regarding the transition from single-agent introduction to multi-agent system, Connectivity Benchmark Report Kurt Anderson, managing director and API transformation leader at Deloitte Consulting, warned of the dangers of shadow AI.
“There is an important leadership principle that puts knowledge in the hands of the people who have the best information to make decisions and build agents, and who can deploy it,” he said. “What’s the difference between this and this? [agentic AI] The revolution is that we accept that knowledge and technology are lowering the bar to building our own capabilities…We can put guardrails and transparency and security in place so we can capture knowledge into our agents.
“This part is really great when compared to past approaches that were more top-down and required solid governance. But we need to learn from the past and ensure good governance. We know what happens when you leave individual contributors alone in terms of building systems that are not secure or reliable, or end up with significant licensing costs.”
Andrew Comstock, senior vice president and general manager at MuleSoft, said in a post pointing to the report: “The true success of an agent company is determined not by the sheer number of agents deployed, but by the overall effectiveness of those agents. We need to think about how agents are discovered, managed, and aligned to work together.”
“As we move into this multi-agent era, IT’s role is evolving from managing silos to building a unified foundation as a central control plane that enables multi-agent systems to be secure, reliable, and scalable.”
Agent AI to improve sales
Meanwhile, other research released by Salesforce this week shows that the use of agent AI software by salespeople with partners has emerged as the top three UK sales techniques for 2026, alongside usage pricing and sales planning.
In its 7th global edition, Sales status A report from Salesforce found that top sales reps are 1.7 times more likely to use an AI agent than those struggling to hit the numbers.
The research surveyed 4,000 sales professionals, 250 of whom were from the UK. The results showed that younger salespeople suffer the most from bottlenecks in the sales process. The average UK seller spends 41% of their time selling, while Gen Z sales staff only spend 35% of their time selling. About 35% of them want to change jobs in 2026, compared to just 23% of Gen Xers and 28% of Millennials, the study found.
Around 90% of those surveyed use some form of AI, and 46% of UK sellers say they have used an agent. UK sales staff surveyed expect agents to spend 38% less time researching leads and 38% less time drafting emails.
Paul O’Sullivan, CTO of Salesforce UK&I, said: “We want to take the busywork out of the way and enable our teams to focus on driving deals forward – building relationships and driving success. AI agents can help make that possible.”
Celonis: AI projects that don’t visualize processes will fail
Meanwhile, Celonis, a supplier of process mining software, recently published research that suggests many companies are moving faster than their operations can handle. the 2026 Process Optimization Reportis based on a survey of 1,600 global executives from companies with revenue of $500 million or more and found that 81% of them said their AI projects would fail without process visibility.
Approximately 85% of respondents to Celonis’ survey said they want to become an agent company within three years, and 90% of organizations say they are already using or considering multi-agent systems. However, 76% said their current processes are holding them back.
