AI agent adoption needs ongoing governance, not one-off projects
Most companies now run AI agents, but few can see, control or measure them. Recent surveys show where adoption breaks down and what teams can do about it.
Gartner expects the share of businesses deploying AI agents to rise from 17% to around 60% over the next two years.
A WalkMe survey of 3,750 people found employees lose 7.9 hours a week to friction with workplace tools, and 45% recently used unapproved AI tools.
Separate surveys found 79% of organizations run agents in production, but only 26% feel well prepared to resolve conflicts between them.
The common lesson is that training is a one-off event, while adoption needs ongoing support, visibility and clear rules for every agent.
Quick read · 1 min
AI agents, software that can carry out multi-step tasks on its own, are spreading fast at work. Gartner expects the share of businesses deploying them to grow from 17% to about 60% in two years, and one survey found 79% of organizations already run agents in production.
The catch is control. Surveys show companies often can't see or govern the tools they already have:
Leaders think staff use 35 apps, but the real number is 661, a WalkMe study found
45% of workers recently used unapproved AI tools, and 36% of them shared confidential data
Only 26% of executives feel well prepared to resolve conflicts between agents
The lesson is that training once isn't enough. Successful adoption needs ongoing help inside daily workflows, a clear map of every tool in use, and rules that tie each agent to a responsible person with limited access.
Executives think their companies use about 35 software applications. The real number, according to a study by software firm WalkMe, is 661. That blind spot sums up the biggest problem with AI agent adoption right now: businesses are switching on tools they can’t fully see, let alone manage.
AI agents are programs that can carry out multi-step tasks on their own, such as updating records, answering customers or writing code. If your employer is rolling them out, the way it handles that rollout will shape your workday, your company’s data security and whether the promised time savings ever show up. The research so far suggests most organizations are moving faster than their controls.
01
How fast AI agent adoption is growing
Use is climbing quickly. Gartner expects the share of businesses deploying agentic AI to jump from 17% today to about 60% within two years. Companies are mainly after automation of complex, multi-step workflows, better efficiency and higher productivity.
Other surveys suggest many firms are already further along. Identity security company SailPoint surveyed 340 security, IT and risk leaders and found that 79% of their organizations were running AI agents in production. So far, the most common uses have been focused tasks in software engineering, customer support and operations.
The value hasn’t kept up. WalkMe’s “The State of Digital Adoption 2026” report, based on 3,750 respondents worldwide, found that 40% of all spending on digital transformation falls short of expectations, mostly because of problems getting people to use the tools well. WalkMe sells software that guides employees through apps, so it has an interest in this finding, but the pattern shows up in other research too.
02
Why training isn’t the same as adoption
This is the distinction most companies miss. Training is a time-limited course that teaches people how a system works. Adoption is the long, ongoing process of people actually changing how they work so the tool delivers value.
Skipping the second part is expensive. WalkMe found the average employee loses 7.9 hours a week, roughly a full workday, to friction with workplace tools:
3.69 hours from a lack of guidance on how to use tools, including 1.03 hours spent decoding unclear instructions
1.88 hours reworking AI prompts and waiting for broken tools to be fixed
1.34 hours typing the same information into several apps
About an hour on workarounds when tools don’t talk to each other
Ofir Hatsor, who leads WalkMe’s sales across the EMEA region, says the gap between the value companies expect and what they actually get will shrink, but over the next three to seven years rather than right away. In some companies, he says, it will never close.
03
The visibility gap and the rise of shadow AI
You can’t govern what you can’t see. Beyond the 35-versus-661 app gap, WalkMe found leaders guess their teams use 21 AI-powered tools when the real figure is 80.
Much of that hidden use is shadow AI, meaning AI tools used outside a company’s approved policies. In WalkMe’s survey, 45% of workers admitted using unapproved AI tools in the previous 30 days because the official systems didn’t fit how they work, and 36% of those put confidential company data into them. Building simple apps with AI coding tools is now easy enough that employees make their own and share them, often without IT knowing.
Honesty is part of the problem. WalkMe’s AI Pulse Survey found 32.5% of workers have passed off AI-made work as their own, and 28.3% have pretended to understand AI in a meeting. Faced with that uncertainty, many organizations are slowing their agent rollouts until they can see and reduce the risks.
04
Governance gaps that slow AI agents down
Confidence is running ahead of readiness. A survey of more than 850 IT decision-makers by software company OutSystems, with KPMG and CIO Dive, found 82% of executives believe they can keep up with governing, connecting and monitoring agents. Only 26% feel well prepared to sort out conflicts between agents, systems and automated workflows, and just 5% have automated controls, audit trails and formal human review points in place.
Identity is the other weak spot. Agents log in to systems with credentials, tokens and keys, much like people do. SailPoint found only 2% of organizations use tools built specifically to manage agent identities, and only 15% can grant that kind of non-human access in real time.
Cost is murky too. Many companies don’t know how many AI tokens, the units AI services bill by, they are consuming or what they cost, which can add up fast. And the business case often assumes headcount cuts, when the real gains may come from better quality work and helping people rather than replacing them.
What smart AI agent adoption looks like for your team
For employees, this means the AI tools at work should come with real support, not just a one-hour training session. For teams and managers, the research points to a few practical steps:
Map what’s actually in use. Find every app and AI tool people rely on, including unofficial ones, before adding agents.
Fix the reason people go rogue. Shadow AI grows when approved tools don’t fit real tasks. Ask staff what they need.
Give help inside the workflow. Guidance at the moment someone is stuck works better than a manual they read once.
Tie every agent to a person. SailPoint’s advice is to link each agent to an accountable employee and give it only the access a task needs, for only as long as it needs it.
Track tokens and outcomes. Measure what agents cost and what they improve, not just hours saved on paper.
Above all, start with the people the agent is supposed to help. Agents that make someone’s day harder will be quietly ignored or worked around.
06
Quick answers
What is AI agent adoption?
It is the ongoing process of getting AI agents, software that completes multi-step tasks on its own, into everyday work so they deliver real value, beyond a one-time rollout or training course.
Why do AI agent projects fail to deliver?
Surveys point to poor visibility into tools, shadow AI, weak governance, unclear costs and too little ongoing support for the people expected to use them.
What is shadow AI?
Shadow AI is the use of AI tools outside a company’s approved policies. It can expose confidential data when employees paste company information into unapproved services.
The machine learning veteran says Microsoft AI wanted bigger models, not research. Here is what happened to AI Frontiers, his cost-cutting idea and the AutoGen framework.
7 min read
We use cookies to understand how readers use Talk With Tech, so we can make it better. Is that OK? Privacy policy
The Daily Brief
Today's biggest tech stories, in 5 minutes
Every morning, the news that matters from AI, phones, apps and the people shaping tech. Explained in plain English. Free.
One email a day. No spam, unsubscribe anytime. Privacy policy