2026 will herald a significant turning point for GenAI adoption in the enterprise. If the last two years were about running pilots in specific business functions or demonstrating proof of concept, then this will be the year that companies start to scale GenAI capability to deliver real value.

The latest findings from Deloitte AI Institute’s ‘State of the Enterprise’ report demonstrate that businesses are already starting to move GenAI adoption beyond experimentation.

One quarter of surveyed companies have shifted at least 40% of GenAI pilots into production - and that number is expected to double over the next three to six months.

What do CIOs and CTOs need to know as they look to make the transition this year? Before thinking about moving beyond the pilot phase though, it’s important to first consider why many pilots don’t succeed in the first place.

The reality check

The companies making real progress on GenAI adoption are not those running the most pilots, but those making deliberate choices about where to apply it first.

Successful adopters have focused on a small number of high-value problems where impact is clear and integration is achievable, using those early wins to build capability and confidence.

A great example is Walmart, which has focused on high-impact areas - both behind the scenes in supply chain and inventory management, and also building agentic commerce into its core retail offering through the ‘Instant Checkout’ partnership with OpenAI.

However, not all firms’ adoption journeys have been so smooth. The much-publicized MIT study ‘The Gen AI Divide: State of AI in Business 2025’ found that 95% of GenAI pilots fail, largely because they are never properly integrated into real business processes.

The most common mistake is launching dozens of pilots across functions, none of which receive the attention needed to scale. Too often, organizations attempt to deploy generic LLMs across vastly different domains - from marketing and supply chain to finance and legal - without properly adapting to how work is actually done.

AI is treated as a layer on top of existing workflows rather than embedded into them, while the realities of skills, data readiness, governance, and change management are underestimated.

The result is predictable: pilots that look promising in isolation but don’t work at scale, leaving organizations stuck in a cycle of experimentation without impact.

Lessons from the front line

Instead, the key is to design pilots with impact and the potential for scale in mind. At Reckitt, our first forays into GenAI solutions were confined to a small number of high-value use cases in the marketing function, where AI tools could reshape outcomes, not just speed up tasks.

Crucially, instead of relying on generic chatbots, Reckitt invested in purpose-designed GenAI solutions, built specifically for marketing workflows - from insight generation and content creation to product innovation.

These tools were trained and tuned to Reckitt’s brands, data, regulatory requirements, and creative standards, enabling AI to be embedded in how work happened, not bolted on as an extra step.

Equally significantly, success of the pilots was defined upfront. Reckitt aligned teams around clear outcomes - faster concept development, better creative output, and strong team adoption - to ensure the success of the pilot was quantified in business impact.

The results were clear: product concepts were generated up to 60% faster, and marketing efficiency improved by more than 30%.

These lessons are not unique to Reckitt. If GenAI is rethought not as a set of tools to deploy, but as a capability to build with focus and intent, they are repeatable for senior leaders across the business landscape.

Three steps to remember

1. Don’t try to AI everything

Despite unprecedented investment, only around 4% of companies take a focused, structured approach to GenAI adoption. It’s far better to start with one high-value function where friction is obvious and impact is measurable.

Use that deployment to work through integration, governance, and adoption challenges, and build internal belief before expanding.

2. Build bespoke tools, not bolt-on experiments

Real value comes from purpose-built AI, trained on proprietary knowledge, embedded into existing tools, and designed around specific decisions and outputs.

The alternative is off-the-shelf GenAI tools, like ChatGPT-style interfaces, which don’t learn from company-specific data and therefore don’t adapt well to workflows or the specific functions where they are needed.

3. Define success upfront

Pilots often stall because no one agrees what success looks like. Whether the goal is time savings, quality gains, revenue impact, or cultural change, leaders must set clear objectives from the outset and align technology, business, and change teams around them.

Deloitte’s data shows the shift to value creation has begun - but only for companies willing to change how they build and deploy AI. The winners will not be those with the most pilots or models, but those with focused use cases, bespoke solutions, clear success metrics, and the determination to see them through.

We've featured the best AI website builder.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit