Jason Rosenfeld, Chief Growth & Alliances Officer, NewRocket.
By the end of 2025, it became clear that enterprise AI did not stall because the technology underdelivered. Models improved. Infrastructure scaled. Investment surged. What stalled was the enterprise itself.
Despite an estimated $30 to $40 billion invested globally in generative AI initiatives, most organizations reached an uncomfortable conclusion: the return was negligible. In fact, recent enterprise surveys show that as many as 90 to 95% of organizations are seeing little to no measurable financial return from their AI investments.
For years, enterprises approached AI as something to pilot, assist or advise. Copilots were layered onto existing workflows. Automation was bolted onto legacy processes. Dashboards multiplied. But the underlying operating model remained unchanged. AI was asked to accelerate work without being allowed to execute it.
That contradiction is what finally caught up with organizations in 2025.
The Real Constraint Was Never the Model
Most enterprise AI programs were designed to be safe rather than effective. AI could recommend but not decide. It could summarize but not resolve. It could flag anomalies but not correct them. Humans remained responsible for stitching insights into action across fragmented systems and approval chains.
The result was predictable. Task-level productivity improved, but enterprise-level impact stalled. Leaders celebrated pilots while CFOs quietly questioned value. Only about 20% of finance leaders report satisfaction with the business impact of recent technology investments, and ERP upgrades and large platform migrations were cited among the lowest-value initiatives despite their cost and disruption.
This gap between effort and outcome forced leaders to take a step back and to examine what really was going on. Enterprises did not need more AI features. They needed a different relationship between humans, systems and decision authority.
Why Agentic AI Is Struggling for the Same Reason
Agentic AI entered the conversation as the next leap forward, but it is already showing the same fault lines. Gartner now predicts that more than 40% of agentic AI projects will be canceled by the end of 2027, largely due to unclear value, escalating costs and weak governance.
This does not mean agentic AI is overhyped. It means it is being misapplied.
Most early agentic efforts attempted to retrofit autonomy into environments that were never designed for it. Legacy systems were tightly coupled. Data was fragmented. Approval models were opaque. In many enterprises, 94% of CIOs say their core data requires significant cleanup before it can support AI initiatives, and only 7% believe their historical data is truly ready.
Autonomy cannot thrive in brittle environments. When organizations try to force agents into broken workflows and environments with poor data quality, costs rise and confidence falls.
What High Performers Are Doing Differently
A small but growing group of enterprises entered 2026 with a different approach. Instead of replacing legacy systems wholesale, they focused on orchestration rather than transformation.
Rather than rip and replace core systems of record, they are layering intelligent workflow, decision logic and automation around them. Nearly one-third of CIOs now prefer supplementing existing enterprise platforms with fit-for-purpose technologies instead of full migrations. This allows AI agents to act across systems without destabilizing them.
More importantly, these organizations are explicitly redefining decision classes. Low-risk, high-frequency decisions are delegated fully to AI-driven workflows. High-impact or ambiguous decisions are escalated with context, not friction.
Autonomy Is an Operating Model Shift, Not a Feature
The enterprises seeing real progress are not chasing autonomy everywhere. They are applying it where it compounds.
Gartner projects that by 2028, 15% of day-to-day business decisions will be made autonomously, up from essentially zero just a few years ago. At the same time, one-third of enterprise software applications will embed agentic capabilities.
Humans are moving upstream. Machines are handling execution at speed and scale. Oversight replaces micromanagement. Governance replaces gatekeeping.
This transition also explains why CFOs are increasingly involved in technology decisions. Nearly two-thirds of CFOs now see themselves as responsible for underlying technology choices tied to business outcomes, and only one in five are satisfied with the value delivered so far. The mandate is clear: innovation must translate into results.
The Consulting Model Is Shifting Too
This change is forcing a parallel evolution in consulting and services. Traditional models optimized for long implementations and manual intervention struggle in environments where AI executes continuously.
Enterprises heading into 2026 are demanding partners who can help redesign workflows, decision rights and operating assumptions, not just deploy tools. They want measurable outcomes, not more pilots.
This is where many AI initiatives stalled in 2025. Not because ambition was lacking, but because execution authority remained trapped in human-centric models that no longer scale.
The Inflection Point Is Now
Enterprise AI stalled because organizations tried to modernize intelligence without modernizing how work gets done. That era is ending.
In 2026, progress will come from allowing AI to act where it makes sense, integrating agents across legacy systems rather than replacing them, and redefining the human role from executor to orchestrator.
The technology is ready. The data is improving. The budgets are there. The outcomes are real. What is changing now is the willingness to rethink the enterprise itself.
Jason Rosenfeld is Chief Growth & Alliances Officer at NewRocket, where he orchestrates global go-to-market strategy and cultivates strategic partnerships with ServiceNow. A recognized voice on Enterprise Service Management, Asset Management and ServiceNow innovation, Jason has 15 years as a ServiceNow executive veteran and 25 years leading professional services organizations and brings unique insight from his previous role as CEO of a major ServiceNow Elite Partner.
© Arc, All Rights Reserved. Request academic re-use from www.copyright.com. All other uses, submit a request to TMSalesOperations@arc-network.com. For more information visit Asset & Logo Licensing.