Companies planning autonomous IT operations still want a person to approve what AI does
Ninety percent of companies surveyed want autonomous IT operations within two years, yet 77% hesitate to let AI decide without human sign-off. The same respondents' readiness and visibility figures show most lack the groundwork even for agents that act under supervision.
The Watch · Security desk
What happened
- Most respondents want AI to propose workflows while people oversee key decisions and actions, and only a minority want IT that manages itself.
- Only 41% say their IT environment is prepared for autonomous operations.
- Technical specialists, the people deploying and running these systems, lean toward supervision more strongly than business and IT leaders, and a smaller share of them think autonomy will become the operating model.
- Security and compliance concerns top the list of obstacles, weighed most heavily by leaders, while specialists cite fragmented tools more often.
Why it matters
- decision Companies with two-year autonomy targets have to fund visibility and agent monitoring before they can drop human sign-off, because intent runs 49 points ahead of readiness.
- exposure Where nobody measures agent reliability, removing the approval step leaves no one in place to stop a bad automated change before it reaches production.
- constraint An agent can only diagnose the parts of the estate it can see. With unified visibility rare, that limits how much any company can trust it to fix things alone.
- contradiction The leaders setting autonomy plans rate their data and governance higher than the specialists running the systems do, so the readiness figures may be optimistic.
In this survey, intent runs 49 percentage points ahead of readiness [21]. The starting base is small. Respondents say 19% of their IT operations are automated today, so 81% are not [4][22]. Leaders give their organization's data a higher rating than the specialists who handle it daily. Help Net Security takes that to mean the people making the plans see a tidier picture than the people carrying them out [7].
The visibility gap is wider. Ninety-six percent say a view across networks, applications, endpoints and cloud matters for AI-driven IT [8]. Seventeen percent have that view unified across all of them [9]. The spread is 79 points [23]. According to the write-up, an agent can't diagnose an outage without a complete view, yet most companies keep that data fragmented or siloed [17]. The number of separate tools makes it worse. Most respondents say consolidating tools would cut operational friction, and most also say sprawl has increased their need to integrate systems [11].
From an incident seat, those two gaps add up to one concrete risk. An agent with write access that works from a partial view of the estate can change the wrong system, and a team that cannot measure the agent's reliability will struggle to explain afterwards why it did. I think the approval step is the control most of these companies can actually staff today. Help Net Security reaches a similar conclusion: without insight into how its agents perform, a company has a concrete reason to leave a person in charge of approval [18]. Many respondents also admit gaps in what they know about employee AI use and its cost. Specialists trust their AI governance less than leaders do [13].
The obstacles go beyond technology. Skills gaps and resistance to change also make the list [15]. One in three companies describe their IT team as already extremely effective at using AI to spot and resolve problems before employees are aware of them [14]. Almost all expect that within two years AI will alter what frontline IT and service desk teams are capable of [16]. All of these are self-assessments. Help Net Security's write-up does not give the sample size, the fieldwork dates or the firm that ran the survey [19].
What to watch
- Publication of the survey's sponsor, sample size and fieldwork dates, which would show whether the 90% and 77% figures hold across company size and sector.
- Whether a follow-up survey shows the 17% unified-visibility figure rising, since that view is what an agent needs before it can be trusted to act alone.
- Whether the 51% oversight gap narrows as companies add measurement of agent performance, reliability and cost.
Clarity's read
What the record supports and how the coverage leans. The claims behind it follow.
Reality
- Evidence35
- Adoption20
- Hype gap+25
- Incentives
- Insufficient
- Confidence35
Claim ledger
Ranked by verification strength, evidence, and original report placement.
- [1]
Ninety percent of companies want to move toward autonomous IT operations over the next two years, with agentic AI carrying out the work.
- [2]
77 percent say their own organization hesitates to let AI make an operational decision without a human approving it.
- [3]
Only 41 percent say their IT environment is prepared for autonomous operations.
- [4]
19 percent of respondents' IT operations are automated.
- [5]
A minority want IT that manages itself; most want AI to propose workflows while people oversee key decisions and actions.
- [6]
Technical specialists, the people who deploy and run these systems, favor supervision more than business and IT leaders do, and fewer of them expect autonomous IT to become the company's operating model.
- [7]
Leaders rate their organization's data more highly than the specialists who work with it every day, which the write-up says suggests planners see a tidier picture than the people carrying out the plans.
- [8]
96 percent of respondents say visibility across networks, applications, endpoints and cloud matters for AI-driven IT.
- [9]
Only 17 percent have visibility unified across networks, applications, endpoints and cloud.
- [10]
Security and compliance concerns top the list of obstacles; leaders weigh them most heavily, while specialists raise fragmented tools more often than leaders do.
- [11]
Most respondents say consolidating tools would cut operational friction, yet most also say sprawl has increased their need to integrate systems.
- [12]
Fifty-one percent say they lack enough oversight of AI agent performance and reliability.
- [13]
Many respondents say they do not know enough about how employees use AI or what it costs, and specialists are less confident than leaders that their governance of AI works.
- [14]
A third of companies say their IT team is already extremely effective at using AI to catch and fix problems before employees notice.
- [15]
Skills gaps and resistance to change also appear on the list of obstacles.
- [16]
Almost all respondents expect AI to change what frontline IT and service desk teams can do within two years.
- [17]
An agent diagnosing an outage needs the whole picture, and in most companies the data sits in pieces or in silos.
- [18]
A company that cannot see how its agents perform has a concrete reason to keep a person at the approval step.
- [19]
The Help Net Security write-up does not state the survey's sample size, fieldwork dates, or the firm that conducted it.
- [20]
Most of the companies mean autonomy with a person at the approval step, and few have the groundwork for even that.
- [21]
The share wanting autonomous IT operations exceeds the share calling their environment prepared by 49 percentage points.
- [22]
81 percent of respondents' IT operations are not automated.
- [23]
The share saying cross-domain visibility matters exceeds the share that has it unified by 79 percentage points.
Get this ledger inside ChatGPT. Ask about any of our stories and the answer brings its figure, each claim’s verdict and the sources. Free, no account.
Sources
1 independent publisher whose own reporting we read for this story.
- helpnetsecurity.comCompanies want autonomous IT operations but hesitate to let AI act alone
1 article · October 8, 2026
Topics and entities
Follow any of these and your For You feed starts watching them — no settings page required.
Topics
- Agentic AI and tool useFollow
- IT ObservabilityFollow
- Autonomous IT operationsFollow
- AI GovernanceFollow