AI could also be extensively used, however drives enterprise at simply 25% of companies

AI adoption stays uneven. Whereas many organizations are experimenting with AI — together with information science and machine studying (DSML), generative AI and agentic AI — enterprise-wide deployment stays beneath 50%, based on current analysis by Dresner Advisory Providers.

Reflecting that uneven maturity, solely a few quarter of the five hundred respondents to the “Particular Report: Agentic and Generative AI” stated AI was a major driver of enterprise technique on the finish of 2025. Nonetheless, that determine greater than doubled in contrast with the primary half of 2025 — a reminder of how shortly expectations are shifting.

A bigger share — 55% — report that AI influences strategic planning however is just not but central to it.

Solely 16% of organizations say they continue to be primarily centered on studying what AI can do, suggesting most have moved past experimentation — even when they haven’t but scaled AI throughout the enterprise. 

As for why they’re investing in AI, organizations cite tackling long-standing enterprise challenges (49%), the danger of trade disruption (26%), and sustaining aggressive parity (8%) as their major motives. 

Associated:Who actually units AI guardrails? How CIOs can form AI governance coverage

“Regardless of the hype, a majority of organizations are nonetheless early of their AI journeys, experimenting selectively quite than deploying AI at scale all through their core enterprise processes,” stated Brian Lett, vice chairman at Dresner Advisory Providers. 

“Nevertheless, for these which can be prepared, AI has develop into an integral a part of technique that’s worthy of funding,” Lett added. “For these organizations, AI is not a skunkworks initiative or speculative expertise. Their AI adoption happens in operations and processes, and hyperlinks on to concrete enterprise outcomes.”

The strategic divide

Taken collectively, the info counsel a market in transition. The rising divide is not between organizations experimenting with AI and people that aren’t. It is between these which can be strategically embedding AI into ruled, production-grade processes and people utilizing AI tactically to enhance work. 

In my conversations with distributors over the previous yr, information maturity constantly got here up as the first bottleneck to scaling AI. Roughly half say they’re constructing instruments geared toward accelerating what has been a sluggish, multi-year course of of knowledge preparation and governance. 

Dresner’s findings reinforce that constraint: With out production-grade information and governance, AI initiatives stall on the pilot stage quite than shifting into full manufacturing. 

What organizations put into manufacturing reveals how far their AI efforts have progressed past experimentation. It additionally highlights the distinct roles totally different types of AI play now. 

Associated:How AI can construct organizational agility

The place AI is being utilized

Information science and machine studying stay probably the most mature types of enterprise AI, centered on optimizing selections and producing operational perception. Frequent functions embrace churn modeling, forecasting, A/B testing, personalization, anomaly detection and useful resource allocation. 

Generative AI has gained traction primarily via use circumstances centered on workforce productiveness. Its worth lies primarily in empowering workers to enhance their each day work. Whereas beneficial, these features alone do not essentially translate into enterprise transformation. Bettering particular person output is just not the identical as redesigning how work will get completed. 

Agentic AI combines analytical fashions, generative capabilities and workflow automation to execute multi-step duties throughout programs. Reasonably than stopping at perception or content material era, these programs act. They set off workflows, replace data and resolve points guided by outlined insurance policies. In contrast to DSML fashions that optimize selections or generate predictions, agentic programs carry these selections ahead. The place DSML informs generative AI assists, agentic programs function. 

Generative and agentic AI adoption

Associated:AI disruption and the collapse of certainty

On the finish of 2025, barely greater than half of organizations reported actively experimenting with generative and agentic AI. Nevertheless, manufacturing deployment stays extra restricted: — 34% for generative AI and 15% for agentic AI — although each charges have greater than doubled since 2024. Finances alignment can also be accelerating, with 72% allocating cash to generative AI initiatives and 66% to agentic AI. 

The hole between funds allocation and manufacturing deployment means that a few of this spending is just not but translating immediately into scaled functions. And whereas generative AI attracts vital funding, enterprise leaders say a portion of that funding is directed towards foundational information work required to help superior circumstances. In different phrases, AI budgets are quietly underwriting information modernization. 

As one college expertise chief famous to me, groups could start with less complicated use circumstances, however creating an AI utility that delivers a single view of the coed or identifies at-risk college students depends upon unified, well-governed information environments.

Information maturity as a constraint

Dresner analysis on agentic AI reveals a constant sample: Organizations which have moved agentic programs into manufacturing sometimes report earlier success with BI, and information modeling and machine studying. They’re additionally extra more likely to have a clearly outlined information chief. 

In different phrases, AI adoption correlates with established information self-discipline. Organizations which have already invested in modernizing analytical information infrastructure, enhancing information high quality, strengthening governance and decreasing information silos are higher positioned to operationalize AI at scale. Agentic functionality tends to comply with information maturity — not the opposite approach round. 

Organizations that progress past experimentation are inclined to comply with a structured path. 

Steps to AI maturity: Experimentation to execution

For CIOs and information leaders, the precedence is obvious: transfer AI from experimentation to embedded execution. That shift requires self-discipline in use-case choice, governance and a dedication to information.