Kyndryl Expands Agentic AI Framework to Scale AI-native Operations Securely Across the Enterprise

Kyndryl introduces advanced agentic AI to move enterprises from pilots to production scale

Kyndryl announced new capabilities for its Agentic AI Framework that help enterprises orchestrate, secure and deploy AI agents across complex environments. The update pairs a structured design process with a hands-on engagement model to shorten the gap between idea and production.

The goal is straightforward: break out of isolated proofs of concept and ship AI-native workflows that improve efficiency and business outcomes across departments.

What's new

• Core orchestration: Converts an organization's tech footprint into agentic AI workflows that can be deployed and scaled.
• Agentic ingestion: Analyzes code, policies, data interdependencies and business goals, with inputs from Kyndryl Bridge, to inform how agents operate.
• Secure-by-design: "Guardian" concepts enable autonomous, transparent and compliant operation with auditability.
• Future workforce model: Defines which roles agents will play and how they collaborate with employees.

Why it matters for IT, dev, and ops

Kyndryl is deploying forward engineers, proprietary methods and IP through Kyndryl Vital to co-create solutions with customers. This reduces time from design to deployment in sectors like government, banking, insurance and manufacturing.

With an infrastructure-first mindset and decades of operating mission-critical systems, Kyndryl reports that roughly a quarter of its recent signings include AI-related work across data architecture, cloud and digital workplace services.

How the framework is applied

Kyndryl's model helps teams define human roles and skills required to deliver outcomes alongside agent counterparts. Using industry reference architectures and a catalog of agents and workflows, the agent builder lets enterprises design, test and deploy agents for tasks such as code generation, test execution and end-to-end process automation.

Agents are created and deployed in line with compliance standards and security protocols, prepared for mission-critical use from day one.

Use cases already in motion

• Insurance: Agentic actuarial workflows that generate regulatory filings, run proactive compliance checks and surface insights for real-time decisions.
• Government: Agents that connect processes across departments (tax, licensing, immigration, benefits), apply policy knowledge and act with civil servants to serve citizens and businesses.
• Banking: Intelligent agents embedded across onboarding-from application to vetting-to reduce cycle time and improve the client experience.

Practical next steps for your team

• Identify 1-3 workflows where decisions follow clear rules and data is accessible (e.g., testing pipelines, document validation, reconciliations).
• Inventory systems, policies, datasets and dependencies to feed an ingestion step. Flag sensitive data and required controls.
• Define an agent RACI: what agents can decide, what they recommend, and what must be escalated.
• Stand up a controlled runtime: identity and access, secret management, audit trails, allow/deny lists, logging, and egress controls.
• Measure outcomes: cycle time, error/defect rate, SLA adherence, unit cost per case, customer satisfaction.
• Pilot a single workflow, then scale via templates and reusable patterns across similar processes.

Governance that scales

Align agent behavior with recognized risk frameworks and control libraries. For example, the NIST AI Risk Management Framework can help structure risk policies, testing and monitoring.

Partnerships and talent

Kyndryl is working with its global alliance partners to build joint solutions and is collaborating with universities to develop the next wave of AI professionals. This ecosystem approach helps customers embed agents into core operations faster and with the right skills on board.

Where to learn more

Explore Kyndryl's services and industry solutions at kyndryl.com.

If your team is building skills for agentic workflows and automation, see focused learning paths and certifications here: AI Automation Certification.