The viability gap: Why 42% of AI initiatives fail (and how to bridge it)

Board-level pressure is mounting. While the initial excitement surrounding generative AI captured the collective imagination, the honeymoon phase is ending.

CIOs are now facing intense scrutiny regarding measurable financial impact, yet the path to ROI remains elusive. In fact, a recent study reveals that the percentage of companies abandoning the majority of their AI initiatives has jumped from 17% to 42% in just one year.

The “move fast and break things” mentality, while popular in consumer app development, is proving insufficient – and often dangerous – when applied to mission-critical operations like loan origination or autonomous supply chains. To avoid the “pilot purgatory” of failed experiments, leaders must shift from “technology-forward” curiosity to a “business-back” strategic framework.

Through our work with global enterprises, Uvance Wayfinders, consulting by Fujitsu, has identified this shift as moving from the “art of the possible” to the “art of the viable.” To ensure AI investments deliver, every initiative must pass three critical filters:

Is it relevant?

Your innovation process must start by identifying “big problems” worth solving. Instead of “letting a thousand flowers bloom” – a strategy where most projects inevitably wither – identify 5–10 game-changing solutions anchored in high-impact outcomes. A pilot without a defined path to scale isn’t innovation, it’s deferred waste. Relevance should be judged by its ability to eliminate a major cost driver or unlock a measurable commercial upside. Ask: How will the era of AI upend the cost basis, product offering, or customer value proposition of your industry?

Is it realistic?

Can you economically deliver and maintain the solution? High-value AI often requires a complex integration of multiple technologies, combining Large Language Models (LLMs) with machine learning, computer vision, IoT, etc. Furthermore, mission-critical AI requires a higher bar for accuracy; you must account for “failover” processes and the long-term costs of monitoring for “model drift.” Unlike isolated pilots, these integrated systems must interact with existing infrastructure and be supported by robust, well-governed data.

Is it Practical?

Can your organisation actually adapt? Even a technically perfect solution will fail if it lacks operational readiness. Success is not found in the code alone, but in the adoption by your employees and customers. If a solution is too fragile, creates friction in the user journey, or requires constant manual intervention, it will never reach the scale required to justify the initial capital expenditure.

The competitive stakes

We may be heading into a winner‑take‑all phase. Companies that achieve automation at scale could reshape the cost structures and value propositions of entire industries. By narrowing focus to three to five high‑impact, viable initiatives, CIOs can mitigate risk, accelerate time‑to‑value, and deliver the step‑change results the board demands. The goal is no longer just to innovate, but to build systems that endure.

Don’t let your AI strategy become a statistic. Access the full blueprint for securing AI ROI and learn how to transition your mission-critical operations from pilot to production.