AI is increasing IT teams’ workloads, draining budgets and distracting them from strategic work, according to Freshworks research, a pattern that is reflected by wider 2026 research into “botsitting,” governance gaps and enterprise AI readiness.
AI is meant to reduce busywork, but many U.K. organizations are creating a new layer of internal bureaucracy as AI environments become more complex and difficult to govern.
The Freshworks Global Cost of Complexity Report found that 88 percent of IT decision-makers in the U.K. say managing AI complexity has increased their team’s workload, second only to the U.S.
A similar proportion, 87 percent, report that 10-39 percent of their team’s AI-related time is spent managing complexity rather than doing strategic work. On average, U.K. IT teams now spend around a quarter of their AI-related working time managing integrations, troubleshooting and complexity instead of focusing on strategic initiatives.
That burden is also hitting budgets. Organizations estimate that 24 percent of their AI spend is lost to unexpected overheads such as integration, governance, data preparation, rework and ongoing support. Freshworks describes this hidden expenditure as the “complexity tax.”
A third of U.K. IT decision makers describe that complexity as “a very significant issue” for their organization.
As Simon Hayward, Vice President of Sales, EMEA at Freshworks, pointed out, “UK businesses are under pressure to boost productivity and cut through bureaucracy, yet our research shows AI is often adding to the workload rather than easing it… If AI is going to move the needle on UK productivity, organizations need to plan for how it fits into their existing systems and teams, not just for the technology itself.”
AI Is Saving Time But Also Creating New Work
The Freshworks findings come as wider research points to a growing divide between individual productivity gains and organizational value.
Glean’s Work AI Index 2026 found that 75 percent of digital workers say AI makes them more productive. However, only 13 percent say AI has significantly improved their organization’s performance and outcomes.
One reason is the amount of work now required to make AI usable. Glean found that workers spend an average of 6.4 hours per week “botsitting,” feeding AI tools context, checking outputs, debugging errors and cleaning up answers that are confident but wrong answers.
It also found that 77 percent of workers have corrected or redone AI-assisted work in the past month, while 36 percent of AI sessions “fail” outright and require a restart or substantial rework.
This reflects the core issue identified by Freshworks that AI often creates new operational work around the technology. As Hayward told CX Today:
“The extra workload is coming from everything that sits around the AI rather than the AI itself.”
“Every new AI deployment has to be connected into existing systems, monitored for accuracy and kept compliant, and that ongoing management is landing squarely on IT teams. Essentially, the productivity gain organizations expected from AI is being absorbed by the operational effort needed to keep it running.”
The Hidden Cost of AI Overheads
Many organizations are finding that their AI budget is being consumed by the work needed to make systems function inside the enterprise, including connecting AI to existing applications, preparing data, managing governance, reviewing outputs, resolving errors and supporting users.
U.K. organizations estimate that almost a quarter of their AI budgets are being lost to this complexity tax, according to Freshworks, including on integration challenges, oversight, data preparation and the operational details that surround AI.
Hayward said:
“Businesses are budgeting for the AI capability itself and underestimating everything it takes to make that capability work inside their existing environment.”
“Our research puts the cost of this at close to a quarter of total AI budgets, lost to what we call the complexity tax… None of that shows up in a vendor proposal, so it gets treated as a hidden cost rather than something organizations plan for from the outset.”
This is also where governance becomes a practical business problem.
Grant Thornton’s 2026 AI Impact Survey found that 46 percent of executives cite governance and compliance failures as a leading cause of AI underperformance. It also found that 78 percent of business executives lack strong confidence that they could pass an independent AI governance audit within 90 days.
Board-level readiness remains weak too. Grant Thornton found that 48 percent of boards have not set AI governance expectations, while 46 percent have not integrated AI risk into ongoing oversight. Only 20 percent of organizations have a practiced or tested AI incident response plan.
For IT teams, that means AI governance can quickly become another operational burden if ownership, controls and escalation paths are not clear from the start.
Point Solutions Are Compounding the Problem
The Freshworks research suggests that AI complexity is often made worse when organizations deploy tools project by project rather than as part of a coherent enterprise environment.
When different teams buy or build AI independently, IT can be left managing duplicate integrations, inconsistent governance processes and fragmented data flows. Each tool may deliver value in isolation, but the combined environment becomes harder to operate.
“That pattern is a big part of what’s driving the numbers we’re seeing,” Hayward said. “A third of U.K. IT decision-makers now describe AI complexity as a very significant issue, and that tends to happen when AI is bought and deployed tool by tool rather than being considered as part of a wider technology environment. Every team ends up tackling the same integration and compliance problems independently.”
Glean’s research points to a similar issue. Its 2026 Work AI Index found that 77 percent of AI users juggle multiple AI tools every week, and 33 percent use four or more. It also found that 53 percent of workers say critical information they need is not accessible through their AI systems.
If AI tools cannot access the right systems, data or knowledge, employees have to fill the gaps manually, copying information between tools, checking answers against source systems and re-running prompts until they get something usable.
Where It Complexity Becomes a CX Problem
The impact of AI complexity is not confined to IT. It can also affect the quality of enterprise customer service.
Customer service AI depends on access to accurate, current and connected information, from CRM data to case histories, knowledge articles, product information, customer entitlements, policies and escalation rules. If those systems are disconnected, AI can give incomplete answers, miss important context or create additional work for agents.
Enterprises need the right data, workflows, guardrails and platform architecture to turn AI into better service outcomes.
The Freshworks findings reinforce that point, because if AI creates more work for IT, that burden can eventually surface in the contact center through slower deployments, unreliable integrations, poor knowledge access and more manual oversight for agents.
The Real Test: Does AI Remove Work?
For leaders assessing AI investments, the Freshworks research suggests that the most important question is whether a tool actually removes work from the organization.
If AI automates a process end to end and reduces the number of steps employees need to take, it can create capacity. If it adds another system to monitor, another output to validate or another workflow to govern, it may simply move the work elsewhere.
Hayward said:
“The clearest differentiator is what happens to a team’s time, not what a tool claims to do… AI that removes work reduces the number of steps a person has to take and takes something off their plate entirely. That’s opposed to AI that adds friction by introducing a new system to monitor, a new output to check, or a new process to oversee, so the human ends up coordinating around it rather than being freed up by it.”
In customer service, agent assist, self-service automation and AI-powered routing can all improve efficiency when they are embedded into existing workflows and supported by reliable data. But they risk increasing the burden on agents rather than reducing it if they require constant checking, manual correction or repeated context gathering,
What Leaders Should Do Next
Freshworks argues that organizations need to measure the hidden work around AI such as integration effort, governance reviews, output checking, data preparation, rework and support tickets.
From there, leaders need to consolidate overlapping tools and ensure AI deployments are connected to a wider technology strategy. Hayward advised:
“The first step is to properly understand where time and budget are actually going. From there, the best move is combining consolidation with accountability. Put one team in charge of moving overlapping tools onto a single, connected platform, rather than leaving each function to bolt-on point solutions and solve the same integration problems from scratch. That single point of ownership is what stops AI complexity compounding project by project.”
Before adding another AI tool, organizations should ask:
- Does it connect to the systems employees already use?
- Does it reduce steps in the workflow, or add another layer to manage?
- Who owns governance, monitoring and incident response?
- What data does the AI need, and is that data accessible and reliable?
- How much human review will be required?
- What work will be retired once the AI is deployed?
Without those answers, AI risks adding to the bureaucracy it was supposed to eliminate.
