In this edition of AI Risk & Strategy we look at why 'use cases' and bottom-up experimentation is exactly the wrong way to think about Agentic AI. It builds on our last article on the status of Agentic AI maturity.

The 46% Failure Problem

The statistics tell a sobering story about the state of enterprise AI strategy today. According to IDC organizations have running an average of 37 AI proof-of-concepts, with some large firms juggling hundreds of pilots across siloed teams. Yet executives who proudly announced pipelines of AI experiments are now confronting a harsh reality: nearly half of these initiatives are destined to fail.

A recent survey of over 1,000 North American and European firms by S&P Global revealed that 46% of all AI proof-of-concepts get axed before reaching production deployment — what industry insiders call "pilot purgatory." Even more alarming, 42% of companies have scrapped most of their AI projects entirely, a dramatic jump from just 17% the year prior. This represents millions of dollars in wasted investment per company, with larger enterprises burning through tens of millions on experiments that never see the light of day.

Gartner's analysts have seen enough. They predict over 40% of agentic AI projects will be cancelled by 2027 due to escalating costs and unclear business value. The culprit? What they describe as a frenzy of early-stage experiments "mostly driven by hype" and misaligned to true business needs. This scattershot experimentation often yields disappointing outcomes, with reports noting high failure rates and poor ROI from the proof-of-concept overload.

For industries collectively spending billions on AI transformation, a 46% failure rate isn't just disappointing — it's a strategic crisis that's about to separate winners from losers.

More at www.ai-risk.co/insights

A Fundamental Misunderstanding: "Use Cases" Are Yesterday's Framework

The core reason so many Agentic AI projects are on track to fail isn't technical — it's conceptual. Companies are trapped in the mindset of chasing individual "use cases" rather than executing holistic strategy. This fragmented thinking treats Agentic AI initiatives as a checklist of tactical experiments that never add up to competitive advantage.

Our own experience in enabling effective adoption shows that companies accelerate ROI when they move beyond isolated initiatives and use Agentic AI to rethink existing workflows within a unified strategy. In contrast, the typical laundry list approach — one pilot for marketing, another for HR, another for IT, all developed in isolation — creates silos of experimentation that never scale or connect.

We propose a shift in perspective: from seeing AI as software tools, to deploying AI agents as a digital workforce. Asking "What are the use cases for AI agents?" is like asking "What are the use cases for hiring people?" — it completely misses the bigger picture. Today's advanced AI agents can be given broad goals and operate with increasing autonomy. As one client put it to us recently, "AI is much more than a tool; it acts like a teammate," logging into enterprise systems, collaborating on Slack, accessing data, and making decisions much like human staff members.

The organizations which successfully leverage Agentic AI will be those which treat it as transformational capability anchored to core business objectives, not a grab-bag of tech demos. The frontrunners we work on this topic focus on transforming core processes and pursuing ambitious outcomes rather than scattering resources across minor efficiency use cases. The key message: viewing AI agents as strategic "digital workers" that augment key business processes is far more effective than piecemeal experimentation.

www.ai-risk.co

The Cost of Bottom-Up Chaos: Why Random Pilots Fail

A bottom-up approach, where disparate teams launch AI projects ad hoc, creates chaos and hidden costs that executives rarely see coming. Without top-down coordination, companies discover different departments solving the same problem in parallel, wasting massive resources. As one tech executive observed to us, teams end up "trying to solve the same foundational problem many different ways" — five chatbot projects here, redundant prediction models there, each reinventing the wheel.

This duplicated effort doesn't just burn budget; it creates incompatible systems and technical debt that make integration a nightmare. The organizational confusion runs deeper. Priorities become unclear when there are 50+ experimental projects underway. Employees get excited about AI prototypes that never materialize into usable tools, breeding cynicism. Meanwhile, executives lose sight of strategic goals amid the noise. One CTO noted that the "Wild West" of AI experimentation is finally calming as leaders realize unfocused pilots were a dead end.

The real killer is opportunity cost. While you're tinkering in endless pilots, competitors are leaping ahead with scaled AI deployments. Our research shows that AI-leading companies have achieved 1.5× higher revenue growth and 1.6× higher shareholder returns over the past three years — but the counterintuitive truth is that they succeeded by pursuing fewer initiatives (about half as many) while scaling more than twice as many solutions enterprise-wide.

Microsoft's excellent recent research on "Frontier Firms" tells the same story. Companies that embed AI everywhere report 71% of employees saying their company is thriving, versus only 37% at less AI-driven firms. The companies stuck in pilot purgatory are losing ground every day that competitors are standardizing AI agents in production.

The Agentic AI Maturity Model

The Strategic Alternative: Top-Down Transformation

The antidote to pilot chaos is hypothesis-driven strategy that starts with clear executive vision of where AI agents create the most business value. This means treating AI as a company-wide transformation, not a tech sandbox.

Bank of New York Mellon provides a useful template for this approach. Rather than random experiments, BNY Mellon's C-suite defined three strategic pillars — run operations better, enhance client service, transform culture — and aligned all AI agent development to those goals. Their enterprise-wide AI platform "Eliza" now supports 40 AI use cases in production and serves tens of thousands of employees. Over 22,000 staff have been trained in AI, with about half actively building new AI assistants and agents themselves.

This 'workforce' mindset unlocks agentic AI's real promise. Consider Lemonade's approach to insurance claims. Instead of piloting claim-processing use cases, they strategically "hired" an AI claims adjuster named AI Jim to handle a high-volume frontline role. AI Jim autonomously processes one-third of all claims now, set an industry record by paying one claim in 3 seconds, and delivered tangible outcomes: 30% cost reduction, 25% faster processing, and 90%+ customer satisfaction for AI-driven claims.

The scalability advantage is profound. Once an AI agent is developed and proven, it can be replicated at near-zero marginal cost. You can deploy 100 customer service agents as easily as one — impossible with human employees. This fungible labour means AI agents can tackle demand surges or expand to new regions instantly. While one company's support team is limited by headcount and training speed, another company's AI agent "team" can scale overnight to handle 10× the volume. (See this case study for the 'art of the possible').

The results speak for themselves. Gartner predicts agentic AI will make 15% of daily work decisions autonomously by 2028. Companies approaching AI as strategic workforce transformation are positioning themselves to capture this productivity revolution, while those trapped in pilot purgatory risk being left behind entirely.

The Practical Framework: Five Phases to Strategic Success

More at www.ai-risk.co

The antidote to pilot chaos is our proven Agentic AI Accelerator — a structured five-phase methodology that guides organizations from initial exploration to enterprise transformation. Unlike scattered experimentation, this approach creates a clear roadmap with measurable milestones at each stage.

Phase 1: Executive Illumination (1-2 weeks) begins with comprehensive assessment of your organization's readiness and opportunity landscape. Through structured workshops and data-driven analysis, we map strategic priorities against agentic capabilities, delivering an opportunity heat map that visualizes potential impact across functions and builds executive alignment on transformation priorities.
Phase 2: Strategic Alignment (2-3 weeks) translates strategic intent into concrete implementation plans. We define priority use cases with clear scope boundaries, establish measurable success metrics aligned with business outcomes, and develop a phased implementation roadmap that secures resources and leadership commitment.
Phase 3: Rapid Prototyping (4-12 weeks) moves quickly from planning to action, developing working agentic solutions in high-priority domains. Through rapid iteration cycles, we validate technical feasibility while demonstrating tangible business value, producing functioning prototypes that solve real business problems with performance dashboards tracking progress against success metrics.
Phase 4: Scale-Ready Strategy (2-3 weeks) develops comprehensive scaling frameworks before expanding beyond initial implementations. We address enterprise integration requirements, regulatory considerations, and change management strategies, delivering secure scaling roadmaps with detailed governance frameworks.
Phase 5: Enterprise Activation (Ongoing) transforms successful prototypes into production capabilities deployed across the organization through phased rollouts, comprehensive training, and continuous optimization—helping organizations realize the full potential of the Agentic Enterprise.

The Imperative for Leaders

The competitive timeline is accelerating. By 2026, operational excellence with AI will be table stakes in most knowledge-intensive industries. Organizations implementing strategic agentic approaches now position themselves ahead of this curve while competitors struggle to escape pilot purgatory.

The window for establishing decisive competitive advantage is closing. Organizations following our structured methodology typically achieve tangible business results within one or two quarters — a stark contrast to the endless cycles of promising pilots that never scale.

The future belongs to organizations that embrace the transformative potential of human-AI collaboration. The question isn't whether agentic AI will reshape your industry, but whether you'll lead that transformation or be left behind by it.

I lead AI Risk, the boutique advisory firm specializing in Agentic AI strategy and fast, effective and safe implementation. Our team has delivered some of the world's most advanced formed of Agentic AI, achieving industry-leading results across financial services and other knowledge-intensive sectors. For details of how you can exploit the Agentic AI Accelerator just reach out here: Simon Torrance

Why Your Agentic AI Deployments Are Doomed to Fail... Before They Start