Boards to demand real-world returns from AI in 2026

Technology executives expect 2026 to mark a reset in corporate artificial intelligence strategies, as boards demand clearer returns on recent investments and shift attention from speculative breakthroughs to practical deployment.

Industry leaders across software, cloud and telecoms predict that companies will reassess expansive AI roadmaps, prioritise proven use cases and scale back experimental spending that lacks a direct link to revenue or efficiency gains.

Spending rethink

Enterprise AI budgets are already under pressure as finance chiefs review the benefits of projects launched during the recent surge of interest in generative models and automation tools.

Eric Ethridge, Senior Technical Account Manager at DoiT, said many organisations will start to question whether their investments have delivered on expectations.

"We spent 2024-2025 blowing up AI as 'the thing', and 2026 will be the year it has to prove itself.

"Forrester just predicted that enterprises will defer 25% of planned AI spend to 2027 - I'd say that number is closer to 40%. Organisations will have to ask themselves, were employees more productive? Did the AI models replace all that human capital companies said they would? Was it even useful?

"Besides ROI, we are likely to see a push for smaller, hyper-focused models that will start to invade all types of desktop and mobile applications. But ultimately, unless quantum computing makes significant progress, we won't see AI make any significant advancements," said Ethridge.

His comments reflect growing scrutiny from boards that backed wide-ranging AI initiatives in the last two years. Many of those projects sit in pilot stages or operate in narrow workflows.

Pragmatic focus

Some technology leaders expect a sharper divide between companies that continue to chase speculative advances and those that concentrate on embedding existing tools into day-to-day operations.

Vishwanath T R, Co-founder and CTO at Glean, said the shift will centre on execution rather than future breakthroughs.

"In 2026, the biggest shift won't be an AI "winter," it'll be a reckoning. Enterprises will realize the next wave of value isn't locked behind AGI; it's in mastering the tools we already have. The winners will be the ones focused on turning today's model capabilities into real products and ROI. The dreamers waiting for a quantum leap in intelligence will be left waiting, while the builders."

Many vendors now promote features that package existing foundation models into search, documentation, customer support and developer tools. These products often target specific workflows instead of broad transformation programmes.

Smaller models

Ethridge also expects organisations to look more closely at compact AI models that run closer to end users. These systems typically train on narrower datasets and support a limited set of functions.

Vendors have started to embed these smaller models inside productivity software, customer relationship management tools and industry-specific applications. This approach reduces reliance on large centralised models that demand more infrastructure and specialised skills.

Companies that adopt this pattern often aim for incremental improvements such as faster document handling or more accurate recommendations. They also gain more control over data residency and governance.

Telecoms adoption

In telecoms, network suppliers expect operators to move beyond trials and run critical infrastructure with AI-based orchestration. Many mobile and fixed-line carriers have tested machine learning in planning, assurance and fault prediction over recent years.

Sarit Assaf, GM, Amdocs Network & Technology, said operators are now preparing for broader automation across their estates.

"2026 will be the year operators stop experimenting with AI and start running networks with it. Operators are moving from AI in isolated tasks to AI orchestrating the entire network," says Assaf. "Pre-integrated autonomy platforms will replace custom deployments, enabling large-scale ROI and reducing operational risk."

Assaf also pointed to efforts to build a consistent software layer that manages AI components across networks.

"The industry will also see the emergence of a long-missing orchestration layer that manages trust, telemetry, and lifecycle governance across thousands of AI models. This 'AI substrate' will allow AI to be operated with the same rigor and predictability as traditional network software. As a result, operators may concentrate investments on a small number of high-impact use cases-autonomous capacity planning, experience assurance, and zero-touch service lifecycle operations." said Assaf.