GoTo’s Pulse of Work 2026: AI Adoption Has Outrun Work Design

GoTo’s Pulse of Work 2026: AI Adoption Has Outrun Work Design

GoTo’s The Pulse of Work in 2026 reinforces our 2026 State of AI findings: the first phase of enterprise AI adoption was tool diffusion; the next phase must be work experience design. GoTo’s central finding is not that AI is useful. That question has largely been answered. The survey finds that 98% of IT leaders say their company uses AI, 82% of employees report using AI at work, and employees estimate that AI saves them 2.3 hours per day. The report also clearly states that value, risk, confidence, and governance are now moving at different speeds. GoTo’s data makes that misalignment visible.

GoTo’s report provides new evidence for a core Serious Insights position: AI should not be treated as a software rollout. AI requires changes in work design, knowledge practices, decision rights, accountability, measurement, and managerial assumptions. For IT leaders, this means AI roadmaps must extend beyond platforms and licenses into how work is actually designed and managed. As I say in the report’s press release, “The next phase of AI value will not come from simply putting more tools in people’s hands. It will come from designing the management system around AI: practical policies, role-based enablement, human judgment, knowledge-sharing practices, and measurement that connect AI use to meaningful outcomes.”

The productivity story is real, but incomplete

The 2.3‑hour daily savings number will attract attention, as will the report’s estimate that employees spend another 2.6 hours per day on tasks AI could handle, representing more than $2.9 trillion in potential annual U.S. efficiency gains. Those numbers fit the optimistic AI narrative, but they should not be read in isolation from the outcomes of that work. Time saved is not the same as value created. Time shifted from one task to another may improve output, create rework, increase review burden, or simply mask quality problems until later in the workflow.

That distinction proves important because the report states that 43% of IT leaders say their companies are not measuring AI ROI very well. That is the signal that adoption is being counted more consistently than outcomes. This aligns with the State of AI 2026 view that AI programs remain too often trapped in activity metrics: users provisioned, pilots launched, prompts written, licenses consumed. The more useful questions ask what improved, what degraded, what risks were introduced, what knowledge was captured, and what assumptions were invalidated.

Overreliance is becoming a knowledge management issue

The most provocative data point is not just that half of employees say they rely too much on AI. It is that 39% say relying on AI is eroding their skills and making them less intelligent, with Gen Z even higher at 46%. Another 30% say they cannot function without AI, and 28% say they trust AI more than their own judgment.

The most provocative data point is not just that half of employees say they rely too much on AI. It is that 39% say relying on AI is eroding their skills and making them less intelligent, with Gen Z even higher at 46%. Another 30% say they cannot function without AI, and 28% say they trust AI more than their own judgment.

That is not simply an AI literacy issue. It is a knowledge management issue. Organizations are at risk of moving from tacit knowledge trapped in people’s heads to synthetic confidence trapped in systems. AI can accelerate learning when it is used to challenge assumptions, summarize alternatives, expose gaps, and support reflection. It can also short‑circuit learning when it becomes the first, final, and unexamined answer. The result is not augmentation. It is cognitive offloading without a learning loop.

This is where the GoTo report aligns closely with the State of AI 2026 argument about capability. Long-term value from AI will come from substituting machine output for human thought. It comes from redesigning work so people learn with AI, not merely through AI. That requires review practices, peer learning, examples of good work, and explicit expectations for when human judgment must intervene. Ignoring this reality will likely lock organizations into a suboptimal past, having given up the means to invent, adapt and evolve as they shed human insight and ingenuity for artificial productivity.

Ignoring this reality will likely lock organizations into a suboptimal past, having given up the means to invent, adapt and evolve as they shed human insight and ingenuity for artificial productivity.

Workslop is the operational tax of unmanaged AI

The “workslop” findings offer the GoTo report’s sharpest operational insight: 43% of employees have used AI outputs even though they suspected the outputs were low quality or might contain errors or fabricated information. Among employees who review others’ AI outputs, 79% say they regularly receive workslop; 66% say reviewing it creates more work, 64% say it makes them question AI’s value, and 48% say it makes them doubt colleagues’ competence and feel resentful.

This is the AI tax made visible. Organizations may believe they are saving time at the point of generation while quietly shifting cost downstream to reviewers, managers, compliance teams, customer‑facing employees, and colleagues who must repair weak work. The results appear positive only if the measurement stops at the person who invoked the AI. The most serious issue becomes trust erosion. Workslop does not just degrade productivity; it also degrades confidence in coworkers, processes, and AI. Poor AI output becomes a social pollutant. It forces teams to ask whether a colleague thought through the work, whether the AI fabricated something, or whether the organization has silently lowered its standards. That ambiguity is corrosive.

A simple example makes this concrete. A sales team uses generative tools to draft proposals and outreach. Time‑to‑first‑draft drops, but legal, security, and sales operations teams spend increasing time correcting claims, fixing misaligned pricing, and chasing down sources. The organization thinks it has accelerated sales, but from a systems perspective, it has simply moved work and risk downstream, which, if not caught before documents reach the customer, will likely result in reduced sales as mistrust spreads beyond the boundaries of internal systems. Productivity that erodes trust is not a valid investment.

Misuse shows the failure of policy without practice

The GoTo report finds that 70% of employees admit they have improperly used AI for sensitive or high‑stakes tasks, up from 54% the previous year. The categories include legal or compliance work, emotional intelligence work, safety‑impacting tasks, high‑stakes strategy, ethical or sensitive personnel actions, and tasks involving sensitive or confidential information. Only 22% say they are aware their employer prohibits AI use for these tasks.

That number should worry leadership teams more than the overreliance figure. It suggests that, where policy exists, it is not being translated into operational judgment. Less than half of IT leaders say their company has an AI policy, and even when policies exist, employees and IT leaders view them differently: 77% of employees say the policy needs improvement, compared with 47% of IT leaders.

A policy that employees cannot apply at the moment of work is not a governance system. It is a compliance artifact. This is why in Management by Design I harp on the value of policy and practice. A policy without practice is inert. A practice without policy may prove unguided. Policy instantiated through practice provides rigor for both.

AI governance needs decision patterns, examples, escalation paths, red lines, and consequences. It also needs positive permissions. Employees need to know not only what they cannot do, but what they can do confidently, with which tools, under which conditions, and with what level of review. For IT, this means moving beyond publishing a PDF to co‑designing playbooks with HR, legal, operations, and line managers that map policy to specific roles and workflows: to the entire work experience.

The IT‑employee disconnect is the management story

GoTo’s report fresh insights into the disconnects associated with AI implementations. The report finds that employees are more likely than IT leaders to say their company is not sufficiently encouraging responsible AI use, needs a better AI policy, is underutilizing AI, lacks practical AI understanding, and is not properly training workers. The report is infused with touchpoints that reflect responsible AI, policy, practical use, knowledge, and training disconnects.

This gap is not surprising. IT leaders focus too much on portfolio, procurement, security posture, and tool roadmap. Employees see task friction, ambiguity, quality risk, and social consequences. Both perspectives are valid, but they are not equivalent, and neither occurs in isolation. Work happens at the edge, at boundaries, in the space where security choice is governed by a policy that isn’t implemented in a tool. If employees report confusion, misuse, weak training, and low practical familiarity, then AI governance has not reached the frontline.

For Serious Insights, this supports a recurring argument: AI governance cannot be owned solely by IT. IT has critical responsibilities for platforms, data protection, security, integration, and observability. But practical AI adoption sits across HR, legal, operations, learning, knowledge management, line leadership, and the employees doing the work. Governance needs to be federated without becoming vague.

Human skills are not soft skills in an AI workplace

GoTo reports that employees rank judgment, creative thinking, emotional intelligence, communication, adaptability, and leadership as the human skills that matter most as AI takes over more work. Sixty‑five percent say employers are failing to equip people with the human skills they need.

That finding should be read in light of data misuse. The problem is not that employees lack prompt tips. Prompting appears in the report, but the higher‑value skills are accuracy and bias checking, knowing when to trust outputs, creativity and judgment, explaining AI results, and keeping up with new tools and practices. Those are not tool skills. They are epistemic skills: how people know, question, validate, explain, and act. The State of AI 2026 framework would place this under capability development rather than training. Training teaches use. Capability development changes how work is performed, reviewed, shared, and improved. AI makes that distinction unavoidable.

What this means for the 2026 State of AI work

The GoTo report fits into the 2026 State of AI narrative as evidence that the AI market has entered a post‑adoption phase. The question is no longer whether organizations are using AI. The question is whether AI is fostering the development of new capabilities or is merely underpining another round of productivity theater. GoTo’s data suggests both are happening at once.

The positive story is clear: employees are finding value, IT leaders continue to invest, and there is substantial untapped work where AI can reduce friction. The negative story is just as clear: overreliance, workslop, weak measurement, policy gaps, and uncertain accountability are turning AI into an unmanaged layer of organizational behavior. The report’s best use is not as a warning against AI. It is a warning against unmanaged AI.

Serious Insights analysis: the next phase is AI work architecture

The phrase I would attach to this report is AI work architecture. Organizations need to define how AI fits into the structure of work, not simply where it can be inserted. That architecture should include:

• Role‑based use cases and patterns of use.
• Human review thresholds for different risk levels.
• Knowledge capture and reuse practices.
• Approved tool configurations and escalation paths.
• Quality standards for AI‑influenced work products.
• Outcome measurement that links AI use to business metrics.

In other words, AI work architecture is the design of how AI fits into work, not just where AI tools are available. The GoTo data argues for a shift from access to enablement. Access answers the question, “Can employees use AI?” Enablement answers, “Can employees use AI well, safely, confidently, and in ways that improve organizational capability?” Most organizations are still over‑rotated toward the first question.

For IT leaders, this is an architectural brief. It calls for joint design sessions with HR, legal, operations, and line leaders to define role‑specific patterns; updates to portfolio and integration roadmaps to support observability and measurement; and investment in knowledge practices and human skills that keep people at the center of AI‑enabled work. The future of AI is not more magic from the foundation models but more engineering from those who adopt them.

Bottom line

GoTo’s Pulse of Work 2026 reinforces the central argument of the State of AI 2026: AI adoption has outrun organizational readiness. The report’s numbers show real productivity value, but they also expose the costs of treating AI as a tool deployment rather than a redesign of work. Overreliance, workslop, misuse, weak policy, limited measurement, and employee anxiety are not side effects. They are symptoms of incomplete work experience design.

The opportunity remains substantial. But the organizations that capture it will not be the ones that simply buy more AI or encourage more AI use. They will be the ones who build the practices around AI: judgment, accountability, measurement, knowledge sharing, role‑based enablement, and governance that employees can actually apply.

For more serious insights on AI, click here.

All images via ChatGPT from a prompt by the author.

Did you enjoy this analysis of GoTo’s Pulse of Work 2026? If so, please like, share, or comment. Thank you.