The challenge with enterprise AI has shifted from broad adoption to generating measurable impact. AI is already embedded across IT operations, software development and physical environments, but many organizations are still working out how to scale it consistently, manage cost and convert investment into value.

HCLTech’s The AI Impact Imperatives, 2026 report shows how far AI has moved into the enterprise. It also highlights the execution gap now facing business and technology leaders. While investment is accelerating and expectations are rising, the average expected failure rate for major AI projects initiated over the next 24 months stands at 43%.

Closing that gap requires more than technology deployment. Enterprises need the right foundations, governance and partnerships to turn AI into a repeatable business capability. As organizations move from pilots to production, third-party experts can bring external experience, reusable delivery patterns and greater cost discipline to AI programs.

For Ananth Subramanya, Executive Vice President, Digital Business at HCLTech, this is changing what enterprises need from AI partners. The focus is shifting toward outcome-led execution, workflow redesign and the ability to scale AI with greater consistency.

Partners are becoming execution accelerators

AI programs increasingly carry board-level expectations. Enterprises are expected to move quickly, demonstrate value and avoid the risk of costly, fragmented experimentation.

The report found that 80% of organizations are partnering with third-party experts to advance AI initiatives. Among those organizations, 90% say partners accelerate ramp-up time and time to value for AI projects.

“Partners have a broader sense of what’s happening in the industry that enterprises want to take advantage of. Their own understanding is perhaps limited to their use cases and scenarios, but partners have a broader lens of how to get that done.”

That broader lens matters because many enterprises are now asking how to move faster.

Subramanya also points to the human side of execution. “Change management is difficult, so typically inertia sets in and partners can actually bring a fresh thought process to enterprises,” he says.

The shift from input metrics to outcomes

The report’s finding that 89% of organizations say partners increase the business impact of AI projects that reach production reflects a wider change in enterprise delivery.

For years, technology delivery was often measured through capacity, activity and speed. In AI, those measures can be misleading if they are not connected to business outcomes.

“Just because you ran faster doesn’t mean you get to your destination, because it requires both speed and direction,” says Subramanya.

This is becoming a defining difference between traditional delivery models and the next phase of AI partnerships. The strongest partners are not simply helping enterprises deploy tools or increase individual productivity. They are helping define what better looks like in business terms.

“What successful partners are doing for enterprises is changing the conversation to outcomes,” Subramanya says. “I’m able to deliver a better outcome for you because I’m able to use AI and that leads to consistency, better quality and speed that lets you get your outcomes faster.”

For enterprise leaders, this changes the role of AI measurement. The focus needs to move beyond whether teams are using AI to whether AI is improving cost, quality, speed, resilience or customer experience.

Repeatability is becoming a cost-control discipline

The cost of AI is becoming an increasingly important part of the enterprise conversation. The report found that 86% of organizations working with partners help organizations manage AI-related costs better than if they had not sought help.

As AI consumption models evolve, cost control is becoming a design issue. Poorly designed prompts, agents and workflows can scale usage without scaling value.

“Today, all model providers across the world have moved to token-based pricing models from subscription-based pricing models,” says Subramanya. “Poorly written prompts, badly written agents are going to cost you a lot of money for very little accuracy of output.”

This is where repeatability becomes more than a delivery principle. It becomes a way to improve cost, accuracy and consistency across the enterprise.

“As partners see a variety of use cases, they are able to identify the right balance of accuracy per token cost,” he says. “As long as you’re able to use those patterns in a consistent manner, you’re able to get to similar outcomes without having to go through the same pains of learning.”

Physical AI raises the partnership requirement

The report describes Physical AI as the application of AI, including GenAI, Agentic AI, machine learning and autonomous decision-making systems, to perceive, understand and act upon physical environments and operational processes in the real world.

While 90% of respondents said Physical AI will be an important technology evolution for their organization to master over the next three years, deployment remains uneven. Partnerships appear to play an important role in Physical AI maturity. According to the report, 54% of organizations working with third-party experts have Physical AI systems in production, compared with 26% of those not using partners.

The reason is that Physical AI depends on a wider ecosystem than most digital AI use cases. It requires hardware, edge devices, algorithms, platforms and industry-specific operational knowledge to work together.

“In the Physical AI world, it is not just the delivery partners that matter,” says Subramanya. “Those delivery partners have dramatically better ecosystem partnerships with various Physical AI providers, be it chip companies like NVIDIA or end-consumer products such as camera providers or device manufacturers.”

He gives the example of port security, where AI can be embedded into cameras to identify adverse events and alert teams before disruption affects operations.

“The ecosystem of solution that I can bring with the hardware provider, the algorithm and the platform makes it much more likely that, on Physical AI, a partnership model succeeds,” he says. “Otherwise, enterprises would have to build it themselves, which can be both time consuming and expensive.”

Industrializing AI as a business capability

For AI to become a business capability rather than a technical program, enterprises need to rethink how work gets done.

“The way we do the work today does not have to be the way we do work with AI,” says Subramanya. “We tend to see optimizations on existing workflow as step one of AI implementation, but that really delivers very little value.”

The bigger opportunity comes from redesigning workflows around AI, rather than using AI to accelerate existing processes. In technology operations, for example, AI can identify a problem, create autonomous fixes, run those fixes through a pipeline and bring a developer into the workflow for review.

Subramanya says industrialization depends on three priorities: reimagining workflows, creating repeatable enterprise patterns and measuring outcomes.

“AI becomes a tool everybody uses, a tool everybody claims they’re more productive with. You get a bill, the bill is big, but I don’t think the outcomes for the enterprise change,” he says. “Having a measure on the outcome becomes extremely important.”

The message for enterprises is clear: AI value is not created by faster activity alone. Instead, industrialization comes from reimagined workflows, repeatable enterprise patterns and measurable business outcomes.