Responsabilidade social
Kyndryl Foundation grantees drive global opportunities
artigo
8 de abr. de 2026
By Kim Basile, Chief Information Officer at Kyndryl
CIO conversations about AI almost always start in the same place: great pilots, impressive demos — and then real frustration when it’s time to scale. The technology isn’t the problem. Confidence, readiness and guardrails usually are.
We see this play out clearly in the data. In a survey of over 1,000 business leaders, Kyndryl found that without workforce readiness, AI efforts don’t scale or drive business benefit. While 95% of those surveyed are using AI, 71% are not workforce-ready, and only 42% are seeing a return on investment. Meanwhile, governance is often a critical enabler of innovation. Leaders in AI implementation were 51 points more likely to have fully implemented a governance framework.
In other words, transitions from pilots to production are failing. Most organizations can build impressive AI demos or small pilots, but far fewer can roll them out safely, consistently, and at scale across the business. At Kyndryl, we are harnessing our Agentic AI Framework to drive measurable adoption and usage of AI to reinvent our internal workflows. Throughout the firm, we’ve been very deliberate in how we’ve approached AI across three dimensions: tools, people and governance.
95%
of businesses are implementing AI across their global enterprises
71%
of business leaders say their workforces are not ready to fully benefit from AI
42%
reported seeing a positive return on their AI investments
Tools: Enabling innovation across the enterprise
As we’ve operationalized the Kyndryl Agentic Framework for our needs, we’ve had to take a hard look at the stack we used to bring it to life. And that brought us to our first strategic priority, which was to enable innovation throughout our enterprise, not just centrally. We are equipping our subject-matter experts to innovate within the processes they own and know best.
The foundation of our ecosystem is a solid set of strategic platforms. Within these tools, we’re making investments to add vendor-specific AI capabilities. We call this approach “batteries included,” and continue to evaluate and implement as we see fit. The result has been an environment of continuous learning and improvement. Customers shouldn’t have to cobble systems together to get value. Our goal is to make it easy for them to get started, iterate, and apply AI in practical and productive ways.
Kyndryl maintains a broad inventory of market‑leading AI assistants that we’ve designed to work alongside people as they move through their day. These tools are powerful, adaptable, and able to evolve with the market.
Just as important, we’ve created space for people to tinker. We designed our AI Garage Lab specifically for that purpose. It’s a safe environment where Kyndryls can experiment with AI, try ideas out, learn what works, and understand what doesn’t — without fear of breaking something or getting it wrong. People don’t learn AI by reading about it. They learn by using it.
All Kyndryls have access to Microsoft Copilot, enabling secure, everyday use of AI. Over the past six months, we’ve seen more than 45 million Copilot actions performed. Here, AI isn’t just for developers – any Kyndryl can build agents. Over fifty percent of Kyndryls have been creating agents and seventy-five percent of those employees are in the non-technical domain. This is really important as we think of AI being for everyone. Kyndryls are able to improve both their own productivity and work tasks. What matters most to me isn’t the numbers themselves—it’s that people who never thought of themselves as “technical” are building agents and using AI confidently in their work.
In terms of AI agents at Kyndryl, fifty percent are focused on productivity and the other half are transforming workflows including how we enable our sellers, deliver our services to customers, procurement and more.
We continue to grow our AI portfolio as new tools become available and evolve in the market. To complement these tools, we provide best practices, accelerators, and guardrails so people can build safely and effectively at different levels of scale. We’re also deploying AI to speed up modernization projects for our customers. Outside of our technical staff, we want to give business users the ability to build and evolve their own agentic workflows using natural language and low- or no-code tools, accelerating delivery while reducing dependency on IT. This approach echoes our overall technology ecosystem approach to align the right technologies with the right workloads.
People: AI as a key capability, not a special skill
AI shouldn’t feel like a special skill. It should feel like something anyone can pick up, try, and use in their own work. My goal is for every Kyndryl to feel confident using AI — whether that means streamlining everyday tasks, building agents, or helping customers think differently about their operations. That confidence only comes when people have the right tools, the right learning, and permission to experiment.
We’ve built a comprehensive, internal AI Learning Framework that blends foundational knowledge, role- specific training and practical application. This begins with required foundational learning — covering essential AI concepts, ethics, and responsible use — to ensure every Kyndryl understands the principles behind AI and its impact on our business. From there, we deliver role-tailored learning designed to meet each Kyndryl where they are. Rather than pushing generic or overly broad curricula, we offer training aligned to individual roles, skill levels, and customer needs.
To support AI adoption in daily work, our AI@Work hub gives all Kyndryls a centralized place to build the AI skills they need. It offers practical resources — including prompting guides, case studies, and governance aligned with strategy — to help employees integrate AI tools to streamline tasks, build agents, and focus on higher-value outcomes. Kyndryls can access curated courses, technical certifications, and resources that support both foundational growth and deeper specialization.
When we integrate governance and architecture from the start, we give people the confidence to develop agents that help close the gap between pilot projects and the creation of real business value.
Governance: Control that enables scale
As Kyndryls build agents and agent‑driven workflows, those agents need predictable, secure access to enterprise systems and data. That access has to be structured in a way AI can consume reliably — and in a way that stays within policy.
Kyndryl’s Agentic AI Framework provides this foundation by combining AI-ready infrastructure, governed data, embedded security, and modern platforms. Kyndryls remain in the loop to ensure agents behave predictably, stay within policy, and avoid risks from opaque or unconstrained systems.
A core element of the framework is risk-tiered, human-in-the-loop control:
- High risk: human-led, agent-assisted
- Medium risk: agent-led with human oversight
- Low risk: autonomous execution within preapproved guardrails
We reinforce these controls through continuous monitoring, bounded behavior, role-based access, and geographic data-residency guardrails. This is what allows us to apply AI confidently in mission‑critical environments.
From pilots to production
Scaling AI safely requires more than technology alone. That’s why we pair our technical architecture with strong governance structures, including an AI Governance Committee, AI Board, AI Security Standard, and AI Systems Inventory, all aligned to Kyndryl AI Policy.
The framework acts as a flexible “core,” while governance provides the protective layer — enforced through an agent gateway handling authentication, authorization, observability, auditing, kill switches, anomaly detection, and integrations with security operations centers, or SOCs.
Policy as Code, an important part of the Kyndryl Agentic Framework, underpins this model. The capability translates organizational rules, regulatory requirements and operational controls into machine-readable policies that govern how agentic AI workflows execute. Human-defined rules, such as approval thresholds, become enforceable controls that agents cannot bypass. Because guardrails are only as strong as the governance behind them, Kyndryl combines mission-critical IT expertise with a governance-led architecture to help build agents for security, compliance, and scale — enabling faster, safer adoption of agentic AI.
When we integrate governance and architecture from the start, we give people the confidence to develop agents that help close the gap between pilot projects and the creation of real business value.
