If you’ve handed your team a suite of AI tools and expected productivity to just… happen, a new study from HERE Technologies has a message you need to hear: you may have created more work, not less.
HERE’s “AI Toggle Tax Report,” released August 12 and fielded across 1,000 full-time employees in finance, healthcare, and the public sector, found that enterprise AI deployments are generating a hidden cost. Workers are spending significant time and mental energy managing AI systems, correcting outputs, and bridging the gap between AI tools and the rest of their workflow.
The Numbers Are Hard to Ignore
The research, conducted by Atomik Research across finance (400 respondents), healthcare (300), and public sector (300) workers, paints a complicated picture of the enterprise AI rollout so far:
- 3 in 10 professionals spend at least half their workday copying information between AI tools and other systems
- 40% of users now handle a higher volume of tasks than before because they’re reviewing and overseeing AI output
- 6 in 10 workers say their organization expects more from them since AI was introduced
- 72% of employees admit to circumventing their organization’s official AI restrictions
That last stat is striking. Nearly three quarters of workers are going around the guardrails their employers put in place, likely because the approved tools don’t actually meet the needs of the job. This is a shadow AI problem that most leadership teams haven’t properly accounted for.
What the “Toggle Tax” Actually Means
HERE coined the term “toggle tax” to describe the hidden friction cost of switching between AI systems, copying outputs into other tools, and manually verifying work that AI produces. It’s the gap between what AI promises and what it actually delivers in practice.
For a knowledge worker who spends 30 minutes copying ChatGPT summaries into a CRM, writing a prompt to re-format a report for a client, and then double-checking the numbers because the AI hallucinated last week, the toggle tax eats up a meaningful chunk of the day. Multiply that across a team of 50 and you’re looking at hundreds of hours of unproductive overhead per week.
This is the productivity paradox of 2026. AI tools are genuinely capable. But capability alone doesn’t translate to ROI. How you deploy AI, what it connects to, and how your team works around it versus with it determines whether the investment pays off.
The Real Problem Is Integration, Not Intelligence
Most enterprise AI rollouts treat AI as a standalone capability. A team gets access to a chatbot, a code assistant, or a document summariser, and then figures out how to stitch it into existing workflows on their own. That stitching process is exactly where the toggle tax accrues.
Workers become unofficial integration engineers. They build manual workarounds. They duplicate effort. They spend time on AI management that used to be spent on actual work.
The study’s finding that 72% bypass official AI policies is a symptom of this. When approved tools create friction rather than remove it, people route around them. That’s human nature. But it also creates risk, security gaps, and inconsistency in how AI is actually being used across the organisation.
What This Means for Business
If your AI deployment feels like it’s stalled, or you’re not seeing the productivity gains you expected, this research offers a diagnosis. The issue probably isn’t the AI itself. It’s the integration, the governance, and the workflow design around it.
A few things worth examining:
Are your tools connected or isolated? AI that operates in a silo forces workers to act as the connective tissue between systems. Purpose-built AI agents that operate inside your existing workflows, connected to your data sources, eliminate most of the toggle tax by design.
What are your workers actually circumventing? The 72% who bypass restrictions are telling you something. They found a better tool for the job. Rather than cracking down, it’s worth understanding what they’re using and why, then deciding whether to formally adopt it or improve what’s officially sanctioned.
Have you measured the overhead? Most productivity assessments track output but miss the management overhead AI creates. Before you expand your AI footprint, quantify the toggle tax in your current environment. You may need to consolidate, integrate, or reconfigure before adding more tools.
Is your team trained on workflow design, not just tool use? Knowing how to use an AI tool is different from knowing how to redesign a workflow around it. The organisations seeing real productivity gains from AI are investing in both.
The promise of AI in the enterprise is real. But the HERE research is a useful corrective to the assumption that more tools automatically means more output. In many cases, the opposite is happening. The toggle tax is real, it’s measurable, and it’s avoidable with the right approach to deployment.
Enterprise DNA works with organisations at exactly this inflection point. Whether that means designing AI agent workflows that eliminate manual handoffs, upskilling teams on how to think about AI integration, or building bespoke AI apps that connect your tools into a coherent system, the starting point is the same: understand where the friction actually lives before adding more capability on top of it.
Research from HERE Technologies, published August 12, 2026. Survey of 1,000 full-time US employees in finance, healthcare, and public sector, conducted by Atomik Research.