Editor’s note: The following is a guest post by Loren Absher, Americas AI advisory lead at.
AI and data are at the heart of digital transformation agendas. But for many enterprises, progress in those critical areas is faltering.
Despite substantial investments in generative AI pilots, data platforms and automation efforts, scaling value remains elusive. What emerges instead is a troubling pattern: increasing complexity, inconsistent data structures and a portfolio of AI use cases that rarely make it beyond experimentation.
The result? Organizations remain stuck in the AI hype cycle — excited by potential, but unable to realize it.
The CIO agenda for the remainder of 2025 is clear: simplify data governance to support real-time AI decision-making, harmonize fragmented data ecosystems and design AI programs around measurable business ROI. Instead of chasing the next breakthrough, leaders must reap the value already within reach — and lay a foundation for the next evolution: agentic AI.
1. Surviving the hype, thriving on execution
The primary barrier to AI at scale is not the lack of tools but the absence of a coherent value realization strategy, ISG research shows. Enterprises have over-indexed experimentation and under-invested in foundational capabilities, particularly in data and orchestration.
Too many generative AI initiatives live in isolated pilots, disconnected from core workflows, measurable outcomes or unified data layers. This experimentation-first mindset must give way to execution-first thinking.
This means prioritizing use cases by ROI rather than novelty. CIOs should ask: What decisions drive our business KPIs? Where can AI enhance those decisions today? How can we measure and iterate on those outcomes? Designing for ROI, not for technology, is the critical mental shift for the second half of the year.
2. Fixing the data layer to enable real-time AI
Most AI failures stem from brittle data foundations. As AI matures from deterministic models to generative and agentic systems, the data challenges intensify. Legacy governance models are buckling under the weight of real-time, decision-centric demands.
Agentic AI will require systems that perceive, decide and act autonomously in dynamic environments. That level of autonomy demands continuous access to high-quality, semantically aligned, ethically governed data. But most organizations struggle with fragmented taxonomies, stale data lakes and compliance models designed for batch processing.
To enable real-time decisions, CIOs must shift toward AI-ready data governance, replacing data management models with strategic architectures that align to the needs of AI.
3. Designing for measurable value
To move AI from experimentation to enterprise value, CIOs must design for how ROI is measured. Too often, value is declared rather than demonstrated, with pilot sponsors grading their own homework and defining success by adoption instead of impact.
True business value must be quantifiable, repeatable and independently verifiable. That starts with defining success metrics at the outset. Like cybersecurity or compliance, CIOs must treat measurement as a first-class design principle, embedding it into the architecture.
This effort requires baseline metrics to quantify the “before” state and isolate AI-driven lift; objective, cross-functional KPIs tied to business goals and transparent value reporting that surfaces results to leadership.
When AI value is measured rigorously and shared transparently, it earns credibility and builds the confidence and budget needed to scale. Measurement is an accountability mechanism that keeps innovation grounded in reality and aligned to strategic priorities.
4. Advancing pilots to enterprise scale
The shift from experimentation to execution requires discipline.
CIOs must build scalable pathways from pilot to production, with clear criteria for pilot graduation, AI Centers of Excellence to codify best practices, integration bridges into enterprise systems and embedded MLOps capabilities to maintain and monitor deployed models.
Crucially, this shift requires reimagining talent structures. Fusion teams — blending domain experts, data scientists, and operational leaders — must become the norm, because they are the engine for translating AI potential into business outcomes.
5. Building trust, governance, and speed
CIOs often worry that accelerating AI might compromise control. In fact, the opposite is true.
Without proactive governance, AI becomes unscalable and risky. Governance should be embedded at every layer, with design-time principles for bias mitigation and explainability, run-time controls for human-in-the-loop and escalation paths, and audit-time oversight for decision traceability and impact analysis.
Trust and agility are co-requirements for scaling AI.
6. Laying the foundation for agentic AI
Agentic AI represents the next leap in enterprise intelligence — systems that don’t just analyze or recommend, but plan, decide and act on behalf of the organization. But the ability to harness that potential depends entirely on what CIOs architect today.
That agentic AI foundation begins with clean, connected real-time data. But just as critical is the architecture it feeds, agentic systems require event-driven, goal-oriented workflows that support autonomous decisioning across enterprise functions.
Agentic AI raises the bar on governance. These systems will drive decisions at scale—often without human initiation. That level of autonomy necessitates embedded guardrails: explainability, auditability, ethical alignment and continuous human oversight.
In essence, CIOs must now design for an AI future that is interoperable, accountable and ROI-driven from the start. Rather than a moonshot, it’s a structured evolution. By acting now to unify data, modernize architectures and codify measurement and governance, CIOs position their organizations to move confidently into the agentic era.
