Kyndryl says firms lag on AI readiness despite gains

Kyndryl says 64% of business and technology leaders now report positive returns from AI and machine learning investments. Its latest readiness report also found that only 24% say their organisations are fully prepared to manage external risks.

Overall readiness fell seven percentage points from the previous year, with cyber security readiness showing the steepest decline, down 11 points.

The findings paint a divided picture for large organisations as AI moves beyond pilots into wider deployment. Executives at the IT infrastructure services group say many companies are seeing value from AI but remain exposed by weak data foundations, uneven governance and slow workforce redesign.

"AI is creating real value. But AI is raising the bar for readiness and sustaining and growing. That value will depend on the readiness of everything around it. Again, it's not just technology, it's people, it's process and how we do what we do," said Martin Schroeder, Chairman and Chief Executive Officer, Kyndryl.

Schroeder said fewer organisations now describe themselves as being in the experimentation phase with AI, suggesting they have moved from testing the technology to broader use. But, he added, that shift has exposed a wider operational problem: adopting new tools does not mean the business around them is ready to support sustained use.

Kyndryl framed the issue around three areas: the underlying technology estate, how people work and the skills they need, and control and governance over AI systems and agents.

Data Gaps

Kim Basile, Kyndryl's Chief Information Officer, said chief information officers are under pressure to move quickly while protecting the core systems businesses rely on every day. Many leadership teams, she said, mistake a promising pilot for proof that the wider organisation is ready.

"A successful pilot can prove that an idea works. It doesn't tell you whether your data, your architecture, your controls, your skills, or even your operating model are ready to support it across the business," said Kim Basile, Chief Information Officer, Kyndryl.

Basile said the report shows four in 10 leaders believe their data is not ready for AI at scale. Those organisations report AI-related failures at twice the rate of companies that believe their data is ready, according to Kyndryl. She said the gap points to a hard lesson many companies have already learned after projects failed in production.

She also highlighted a contradiction in views of IT modernisation. Kyndryl found that 71% of leaders say they have mostly or fully modernised their technology estate. At the same time, 58% say that estate needs significant re-architecture before it is ready for AI at scale.

For Kyndryl, that means a modern estate is not always an AI-ready one. Basile said older assumptions about architecture, workflows and controls do not always hold when companies try to run newer AI models across large, complex environments.

The workforce picture is also lagging. Basile said 86% of respondents last year expected AI to transform roles within 12 months. This year, only 27% said they had formally redesigned roles around AI.

"We need to empower people with the tools, the training, and the change management structures to actually use AI effectively," said Basile.

Companies need clear answers on who makes decisions, when a person should intervene, who owns the outcome and how performance is measured, she said. Without that, businesses may see broad internal AI use without the results they expected.

From Pilot To Scale

Ismail Amla, Global Consult Lead at Kyndryl, said many organisations run into problems when they try to move from pilot projects to enterprise deployment. The main issue is usually not the model itself, but the business around it.

"The real challenge is rarely deploying the technology itself. It's rewiring the organization. Organisation to capitalize on the value AI can bring," said Ismail Amla, Global Consult Lead, Kyndryl.

Amla said AI often lands in organisations with existing systems, established workflows and fixed accountability structures. Costs become harder to predict once AI is added to live systems, he said. Some businesses also simply attach AI to already inefficient processes, limiting any improvement.

That can produce a narrower form of progress: a company may become more efficient without becoming more competitive or adaptable. Amla said the report suggests leaders feel less prepared for competition from outside their traditional sectors. In his view, AI is lowering some of the barriers that once defined industries, including who can build technology and how quickly they can enter adjacent markets.

That raises a broader strategic question for boards and senior executives: whether AI is being used to improve the current operating model or to reshape the business itself.

"Are we using AI to reinvent how work gets done? Or since implementing the existing model faster?" said Amla.

He said organisations making the strongest progress tend to start with a specific business problem. They identify where decisions stall, where handoffs create friction and where customer or employee experience breaks down. They then align the technology, data and controls around that workflow rather than waiting for every part of the enterprise to be modernised.

Governance And Security

Governance featured heavily in Kyndryl's assessment. Basile said it should not be treated as the opposite of speed. In her view, governance provides the structure that allows companies to make fast decisions safely and repeatedly.

"We need to stop treating governance as the opposite of speed," said Basile.

She linked that point directly to cyber security. AI is changing the balance for defenders and attackers at the same time. Security teams can use AI to improve detection and response, but threat actors can also move faster and increase the volume of attacks. Basile said that means architecture, governance and cyber security must work together from the start.

Kyndryl also used the event to outline its internal approach to AI deployment. Basile referred to the company's Kyndryl Agentic Framework, which she said is used to design, deploy and manage AI agents with governance, security and human oversight. She also said Kyndryl is trying to push AI experimentation closer to subject matter experts rather than keeping it solely within a central technology function.

Amla pointed to Kyndryl Bridge as another part of that approach. He said enterprises need tools that can observe how live systems and AI agents behave, track them in a catalogue, and apply policies and risk categories across those environments. He said the shift in modernisation is away from one-time overhauls and towards continuous change in systems that are already running.

Sovereignty Pressure

Another issue running through the report is digital sovereignty. Schroeder said customers, particularly in Europe, are focusing more closely on where data sits, where technology runs and who controls operations.

"Sovereignty is really about control," said Schroeder.

He said Kyndryl can already address parts of that issue through data location controls and local staffing requirements in sensitive environments such as government. Amla added that a growing concern is what he called "sovereignty debt" - technology decisions that reduce a company's ability to switch providers, move workloads or operate differently across geographies.

That concern sits alongside an older issue in enterprise IT. For years, many businesses pursued consolidation and standardisation. Amla said that improved efficiency but also increased concentration and dependency in parts of the technology stack. In the current geopolitical and regulatory climate, he said, companies are reassessing those trade-offs.

Schroeder said that reassessment is also affecting architecture choices. He pointed to moves by some organisations towards different operating systems and open-source options where they believe they can retain more control.

On business measurement, Schroeder said many AI projects are still justified by productivity gains, including faster processes, lower costs and improved service delivery. He said some customers are now attaching growth ambitions to those same projects as they become easier to do business with.

Inside Kyndryl, he said the company tracks whether AI leads to better decisions and stronger learning among engineers working on customer systems.

Asked what leaders should do next, Schroeder returned to the workforce issue. He said companies should focus on role-based education and skills development because the technology is advancing quickly and some employees remain uneasy or openly resistant.

"The technology works. So it's really now about people and about changing the process," said Schroeder.