Enterprises struggle to manage rising sprawl of disconnected AI tools
Most enterprises are struggling to integrate a rising number of artificial intelligence tools, with 70% failing to move beyond basic connections between AI applications, according to a new survey from Zapier.
Tool proliferation
The research found that more than one in four enterprises now use more than ten different AI applications. Despite this high adoption rate, companies face significant challenges in managing what Zapier describes as 'tool sprawl'-the growth of AI apps outpacing the ability to connect and oversee them effectively.
Disconnected AI tools have left data siloed and have increased reliance on manual processes. Employees are spending extra time transferring information between different systems, reducing the productivity benefits these technologies are supposed to deliver.
Integration shortfall
The study highlighted that only 35% of enterprise leaders say the AI tools used in their organisation go through formal approval channels. Among the 70% who have not achieved meaningful integration, 43% operate with only basic connections between some applications, while 22% run most tools independently. A further 6% have yet to connect their AI tools at all.
This lack of integration creates operational headaches, with 22% of organisations relying on manual data transfers between siloed AI systems. Thirty-six percent of enterprise leaders say disconnected AI has led to increased security and privacy risks, while 34% cite difficulty in training employees and 30% report wasted spend on redundant software.
Shadow AI emerges
Shadow AI use-where staff deploy unapproved AI applications-was found to be widespread. Only about a third of respondents say their organisation's AI tools go through proper vetting. Each month, approximately 31% of enterprises discover unauthorised AI tools in use.
Operations and IT teams are most likely to use unapproved AI (23%), followed by customer service (21%) and HR (20%). Fourteen percent of companies have no insight into what AI software their employees are using.
Escalating adoption
The complexity is set to grow, with 66% of enterprise leaders indicating they intend to increase their AI tool count over the next year. Three-quarters (76%) report having experienced at least one negative outcome from disconnected AI applications.
Despite the challenges, only 35% of companies have invested in or are considering adopting a dedicated platform to orchestrate and connect their AI tools-a solution that 90% of leaders describe as either critical or important.
Managing sprawl
Some organisations are taking steps to address the issue. Thirty-nine percent have moved, or plan to move, towards standardising with fewer AI tools. Thirty-eight percent conduct regular audits for unauthorised tool usage, and 44% have introduced or are developing specific AI usage policies.
"Tool sprawl is the natural outcome of rapid AI adoption. Every team wants the latest AI capabilities, but without proper integration and governance, you end up with disconnected systems that create more problems than they solve. Our research shows enterprises need to shift from simply adding AI tools to orchestrating them as unified business processes," said Emily Mabie, AI Automation Engineer, Zapier.
Thirty-six percent of executives say they have built or plan to build custom integrations for their AI tools, but the survey suggests this approach becomes increasingly costly and complicated as the number of applications rises.
"We're seeing AI add to workload problems instead of solving them. When tools don't talk to each other, employees spend time copying data between systems. That's the opposite of what AI is supposed to do," said Mabie.
