Scott Francis is Technology Evangelist at PFU America, Inc. and Ricoh Document Scanners.
As the end of 2026 draws near, the central question for business leaders has shifted from asking whether their organization is adopting and deploying AI fast enough to assessing whether AI is performing as expected. With IT budget planning for 2027 well underway, organizations should use the waning months of this year as a reality check that will serve to separate organizations positioned to scale in 2027 from those still catching up.
There’s no question that the demand for AI is skyrocketing. What is unclear is whether organizations are meeting the projected return on investment thus far. Gartner predicts AI spend will reach $2.59 trillion globally this year, a substantial 47% increase over 2025.
With the end of the year quickly approaching, there is an opportunity to take stock and refocus on overcoming challenges to achieving ROI and readjust where necessary. Rather than taking a scattershot approach to AI deployments, organizations should take a measured, strategic approach on where to apply AI; for example, using one workflow, establishing KPIs and then tracking those quarter-over-quarter before expanding to more departments or enterprise-wide.
Carefully Planning And Preparing For The Agentic Transition
As the saying goes, “life moves fast.” And we’ve seen from technological advancements over the last few years that AI moves just as fast. Many companies are already in the planning phase of employing agentic AI, according to McKinsey’s report, with 40% of companies surveyed already scaling agents, up from 27% in 2025.
Despite the positive uptick in deployments, organizations should prioritize getting current AI deployments right, determining what’s hindering ROI and preparing for the next phase of digital transformation.
Priority One: Fixing Data Quality
AI systems are only as trustworthy as the data foundation, which is a major reason why prioritizing data quality is imperative. Flawed data takes on a new meaning in the AI era because it has the potential to undermine the models built on it.
The “2026 State of Data Integrity and AI Readiness” survey of data and analytics leaders puts a spotlight on data quality challenges: 43% of respondents cite data readiness as the most significant barrier to AI alignment with business objectives. Creating proactive strategies to fix bad data, including data errors, redundancy and more, is still one of the most inexpensive approaches to protecting against downstream damage. This imperative step provides organizations with the confidence that their AI deployments will meet performance expectations and transform data into the dependable foundation that reliable, high-performing AI requires.
Before expanding AI deployments and introducing agents, executives should audit whether the underlying systems can actually support agentic workflows, including clean data pipelines, defined APIs and integration between the systems agents need to touch.
Investing In Change Management And Reskilling
Preparing teams well before AI is rolled out across the organization is critical. Skipping this step can breed employee resentment and pushback and result in lowering morale.
Workplace tension is already a challenge, and for good reason. A WRITER study found that “92% of the C-suite are actively cultivating ‘AI elite’ employees, while 60% plan layoffs for non-adopters.” Business leaders should not view AI as a replacement for skilled personnel; it should be used to surface insights and recommendations, while consequential decisions, especially those affecting employees or customers, remain in human hands.
Training, communication and genuine feedback loops are what separate adoption from chaos, making change management well worth the investment. Reskilling and preserving the human element build trust and catch the blind spots LLMs can’t.
Planning For Agentic Transition Amid Geopolitical And Regulatory Volatility
IT leaders should be aware that AI is not immune from volatility, and a backup plan is mandatory. One well-publicized example of AI disruption is Anthropic’s suspension of global access to Claude Fable 5 and Mythos 5 following a U.S. government export control directive. Access was later restored under a modified compliance structure; however, business leaders should take this incident seriously and acknowledge that government-initiated suspension or access restriction can occur without advance notice. Business leaders can address regulatory instability by identifying which models and vendors they rely on and identifying fallback options before a suspension forces an unplanned migration.
There is also a growing regulatory divide between jurisdictions. The U.S. export rule’s workforce application means enterprises with frontier AI API integrations and global engineering teams may be exposed to compliance risk regardless of where their servers are located. Gartner projects more than a third of enterprises will run localized AI platforms by 2027, up from just 5% in 2025. For multinational firms, this means agentic AI strategy can no longer be a single global architecture; it will need to incorporate jurisdiction-based contingency planning.
Turning Momentum Into Long-Term Gains
The gap between AI pilots and AI payoff won’t close on its own. The executives who make real progress before year-end will be the ones connecting these three priorities to long-term success to achieve business goals and ROI goals. I believe it’s imperative for business leaders to focus on the fundamentals this year before expanding AI projects and incorporating agents, treating their data quality, teams, economics and risk as interconnected.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?