Enterprise AI has reached a pivotal moment. Years of experimentation, pilot programs and false starts are giving way to a hyper-focused era of AI demand. The conversation has moved beyond a general sense of enthusiasm for generative AI as more pressing, hard-to-answer questions emerge. Frustrated by an increasingly obvious gap between the money spent on GenAI tools and the technology’s quantifiable benefits, more organisations want to know: how can we scale AI to deliver measurable business value? And, for some, is AI even the right solution?
Deloitte’s 2026 AI Report highlights the emerging industry consensus that successful AI adoption is about focusing ambition through the lens of tangible results. Executive teams increasingly measure success by ROI, good governance, workforce readiness, security and the commercial implications of their AI adoption.
The opportunity is significant, yet the gap between experimentation and enterprise-scale deployment remains challenging to bridge. Many enterprises still lack the necessary data, processes and governance foundations to scale AI effectively, while others worry about becoming locked inside a new and costly vendor ecosystem.
This readiness gap is increasingly becoming the difference between organisations that scale to achieve measurable outcomes with their AI programs and those that remain trapped in the pilot stage. At the heart of this issue is that, not only is AI being applied to systems and processes that aren’t in a position to best make use of the technology, but in the race to adopt, organisations are pushing AI into spaces where it isn’t needed at all.
Not every question requires AI as the answer. Business and technology leaders are approaching every challenge with an AI-first assumption, when instead they should be asking:
1. How can we fix the underlying processes?
2. Can we use traditional automation to solve this problem?
3. Then, and only then, should we ask: “Does AI make sense?”
The AI readiness gap
Despite widespread investment, relatively few organisations have successfully crossed the gap between being AI pilots and deploying AI at scale. According to McKinsey, just 1% of executives in 2025 believed their AI deployment had reached anything resembling maturity. McKinsey also found that fewer than 10% of organisations have successfully scaled AI agents.
While the potential of Agentic AI is difficult to deny, so too is the fact that there are still kinks to iron out. More than half (55%) of IT decision makers recently surveyed cited reliability and hallucination management as their primary concerns. CIOs also continue to voice concerns over the long-term economics of Agentic AI, particularly as stories of skyrocketing token consumption emerge, as well as the risk of vendor dependency and rising operational costs.
These findings all point to the obvious truth that adopting, much less scaling, Agentic AI is a thornier issue than it seemed a year ago.
AI isn’t always the answer: Taking a methodical, three-layer approach
When all you have is a hammer, everything looks like a nail. Tech and business leaders need to catch their breath and realise there are other tools at their disposal, no matter how shiny the hammer or how much pressure they’re under to start hitting things with it.
Approaching business problems under the assumption that AI is a magic wand will create more problems and costs than it solves. Agentic AI in particular needs strong foundations to be effective.
Many enterprises continue to treat AI as a technology initiative when it is fundamentally an operational capability. Deploying AI across poor-quality enterprise data and inefficient operating models rarely produces sustainable value. Agentic AI is a force multiplier for what already exists. Apply it to good data and clean processes, and the results speak for themselves. Fail to lay a solid foundation, and AI will accelerate and exacerbate existing problems by producing unreliable outputs, inconsistent decisions and ballooning governance risks.
For isolated use cases, AI can generate rapid productivity gains, but scaling AI across an enterprise requires much stronger foundations. Data quality, governance, process standardisation, security, and operating models all become prerequisites for success.
When approaching Agentic AI ERP, first fix your processes. Remove instances of waste, redundancy, and time-consuming manual workarounds. Not all technology problems require technology solutions — just good operational discipline.
Then, if possible, use traditional automation to streamline processes as much as you can. Rules-based, non-decision work like approvals, routing, and reconciliation don’t need an AI agent, just traditional automation that can take care of high volume, low complexity tasks.
Completing these two steps before having any serious conversations about AI puts an organisation in the best possible position to execute a successful pilot that stands a good chance of scaling organisation-wide. Target decision-oriented work where real ROI can be measured. Give your Agentic AI pilot clear success criteria supported by fallback paths and strong governance. Recognise that AI is just one tool in your toolbox. It’s not the solution to every problem.
Preparing for enterprise-scale AI
The enterprises that come out ahead in the next phase of AI adoption probably won’t be the ones that rushed to be first or that spent the most money on the largest, shiniest technology platforms. Instead, it’ll be the ones that firmed up their data quality before throwing AI agents into the mix, established strong governance frameworks, modernised their operational frameworks and retained the flexibility to pivot as AI capabilities mature.
Most importantly, they’ll be the organisations that realise AI isn’t always the answer, and that AI success relies upon setting your implementation up for success.
Enterprise leaders looking to be among the winners of the next tech race should get their priorities in order before committing to any AI roadmap:
- Do the necessary legwork to ensure operations and data foundations are in place to give AI the best chance of success.
- Take a critical approach to determining how (and if) AI should be deployed.
- Build on strong foundations of organisational discipline and automation that will allow AI to scale.
- Avoid unnecessary vendor dependency.
- Evaluate platform decisions according to measurable business outcomes, not technology trends.
The next 18 months will see the urgency among enterprises to move beyond pilot programs and into full-scale AI deployment grow. But I urge every enterprise leader who thinks it’s better to take a leap of faith than to be late to the party to take a breath. Real competitive advantage will belong to the organisations that are genuinely ready to scale, and that do so in a considered, careful way. Without firm foundations, it doesn’t matter how high or fast you build; it’ll all come crashing down.
James Harvey
James Harvey serves as Theater CTO, EMEA at Rimini Street, where he advises clients on strategic innovation initiatives that align technology with long-term business goals across enterprise applications and architecture. He brings 25 years of leadership experience across banking, financial services, energy, logistics, FMCG, retail, and technology, with deep expertise in strategy, operations, AIOps, security, ERP, SRE, DevOps, and observability. Before joining Rimini Street, he was Executive CTO, EMEA - Observability at Cisco, and earlier held senior IT leadership roles at organisations including RBS, ABN AMRO, New Look, and BP.
