The number that went backwards in ServiceNow’s AI index

AI spending grew 110% in a year and agent adoption reached 59%. Over the same period, the share of organisations connecting AI workflows across business functions fell from 30% to 16%. Findings from 4,500 executives and 2,000 employees in 19 countries, including 100 executives in the UAE.

Every headline figure in ServiceNow’s Enterprise AI Maturity Index 2026 moved in the right direction except one – Investment grew 110% year on year. Agent adoption reached 59% of organisations, up from roughly a third. Overall maturity scored 51 out of 100 — a recovery, since the index fell to 35 in 2025 from 44 in 2024, so the 16-point rise restores lost ground rather than breaking new.

In the UAE, drawn from a sample of 100 executives, spending rose 105% and the maturity score reached 48.

Against all of that, the proportion of organisations that had streamlined and integrated workflows across business functions using AI fell from 30% in 2025 to 16% in 2026. Fourteen points down, in the year of the largest spending increase the study has recorded.

ServiceNow’s explanation is that organisations still run fragmented platforms and have now added a layer of agents on top of them. That reading holds up against the rest of the data, and it is the most useful thing in the report for anyone approving next year’s budget.

Adoption is running four to six years ahead of the plumbing

59% of organisations use AI agents. 16% have replaced fragmented legacy systems with an integrated platform. In the UAE the equivalent figures are 57% and 14%.

Agents are being connected to estates that were assembled for a different purpose, and the study is specific about the consequence. Executives name inadequate data accuracy, access and management as their single biggest barrier to AI adoption, at 71% globally and 77% in the UAE. Legacy integration follows at 47%, infrastructure at 45%.

The seven pillars measured make the same point in one line. Vision and leadership scored 57. Management and culture, and data modernisation, 54 each. Governance, 53. Talent and skills, 49. AI-enabled workflows scored 40, last of the seven. Seventeen points separate what organisations have decided to do from what their systems can carry out.

Three in five run agents. One in five can test them.

59% adoption sits alongside 20% who have implemented AI testing, auditing and risk assessment processes. In the UAE, 57% and 16%.

Break the adoption figure down and the exposure is easier to size. 41% of organisations use agents to assist individuals with no autonomous action. 13% are scaling what they already run. 5% are redesigning how work gets done. 9% have agents completing multistep work independently — 7% in the UAE. None of the 4,500 organisations surveyed has built anything that works across functions and improves itself.

So most of that 59% is running assistants, which is low-risk. The problem arrives as the 9% grows, and it is scheduled to: 20% expect autonomous multistep workflows within two years. The testing figure has no comparable trajectory attached to it.

Vijay Kotu, Chief Analytics Officer at ServiceNow, sets out what that costs in the report. Connected agents amplify each other’s errors as readily as they amplify each other’s value, complexity compounds faster than organisations anticipate, and by the time it surfaces the correction is expensive.

The top fifth got there by paying upfront

The study groups 21% of organisations as Pacesetters, averaging 74 out of 100 against 45 for everyone else. They report 160% average AI ROI today, projected at 194% within two years, and describe themselves as 5.6 times more productive, 2.7 times better at scaling and 2.6 times better at reducing risk.

The behaviours behind that are measurable:

Pacesetters are six times likelier to have testing in place and eighteen times likelier to have agents working independently. The sequence in which those two things happen is what the whole report is arguing about.
Saif Mashat, Area Vice President for the Middle East and Africa at ServiceNow, put the same case to regional buyers when the UAE numbers were released: “The organisations pulling ahead are no longer distinguished by how much they spend on AI, but by how effectively they operationalise it.” He added that “strong governance, connected data and orchestrated workflows are what translate investment into business outcomes.”

Executives are budgeting against resistance the survey does not find

The index surveyed employees for the first time this year, and the two groups disagree in one consistent direction.

71% of employees expect AI to improve their morale and job satisfaction. 45% of executives think it will. 67% of employees expect it to free them for higher-value work, against 46% of executives. Executives worry more about skills erosion than employees do, by 34% to 18%, and more about ethical concerns, by 37% to 23%.
Meanwhile the gaps employees actually report go unaddressed. 42% say they are not getting enough AI training. 59% of organisations have no long-term HR plan for the future of work. 21% have assessed AI skills across the enterprise. 58% of employees say their organisation is doing a poor job connecting AI workflows across the business — the workforce arriving at the same conclusion as the 16% integration figure, from the other end.

The country detail varies the emphasis without changing the shape. 42% of French executives say clients are satisfied with AI-enabled experiences, against 63% in the US. 57% of US employees think leadership is not keeping up with market shifts, and six in 10 Japanese employees say the same. India’s constraint is infrastructure digitised late and now tied to national platforms. Germany’s is legal liability. Across every market the study covers, the barriers are integration, data and people.

What a CIO should do with this

Four checks, each against a benchmark in the study.

Count your agents, then count what can test them. Sitting inside the 59% and outside the 20% is the largest unpriced risk on a 2027 plan. Pacesetters run at 61%.

Ask whether integration is funded in the same line item as adoption. 84% of organisations have not replaced fragmented systems. The 14-point fall in cross-functional integration is what buying agents on top of that produces.

Settle data ownership before the next deployment. 71% of Pacesetters have formal ownership and control policies against 22% of everyone else. It is the widest foundational gap in the research and the cheapest to begin closing.

Move the change-management budget from persuasion to training. Employees are 26 points more optimistic than their own executives about job satisfaction. The shortfall the data shows is skills.

Brian Solis, Head of Global Innovation at ServiceNow, argues in the report that every leader has an org chart and very few have a map of how outcomes actually move through the business — where people add judgment and where software can be trusted to finish a job.

Last year’s 110% increase moved vision to 57 and left workflows at 40. Whether the next 81% does anything different depends on which of those two numbers it is aimed at.