From individual to enterprise: What IBM Think reinforced about AI at scale.

Earlier this month, I had the opportunity to attend IBM Think in Boston, IBM’s flagship event for enterprise leaders navigating the next era of AI, hybrid cloud, and automation.  

For anyone operating at scale within a highly regulated industry like financial services, the conversations at Think are invaluable for their focus on execution. “What can we run securely, repeatedly, and at scale?” 

SEI has been on an AI journey for several years, and we’ve been intentional about how we frame it. We don’t want an “AI strategy” that sits alongside the business. The business must be “AI-conscious” from bottom to top for any meaningful change to happen in speed, quality, and scale. Speed is highly valuable to clients. Quality is king in financial services. And scale creates better overall capability and service that we can deliver to the market. The conference was particularly timely, as we begin our partnership with IBM Consulting to help us define and deploy the agents that will enable our operational workforce to run faster, deliver greater consistency, and minimize friction.

My biggest takeaway from Think? We need to move beyond AI-enabled individuals and toward an enterprise operating model for AI—one that blends people, platforms, governance, and measurement into the daily fabric of work. Below are a few reflections for leaders thinking about how to scale AI responsibly across technology and teams.

1. The nature of work is moving to orchestration. 

One concept I kept returning to is that work is shifting from “doing” to “orchestrating.” On engineering teams, some of the strongest developers are already becoming orchestrators—people who understand context deeply, break problems down effectively, and coordinate agent-driven execution.  

This transition is moving into operations. When operations professionals begin creating and using agents, we must support them with tooling, governance, and a way to share what works across the enterprise—similar to how engineering teams manage reusable components and CI/CD pipelines. 

At SEI, we’ve made significant progress enabling individuals with AI tools—from broad productivity capabilities to role-specific assistance for developers. But the next step is enterprise readiness: building the right guardrails and distribution mechanisms so useful that agents and workflows can be shared, governed, and improved as they scale. 

2. Governance may be the biggest challenge as AI scales.  

One challenge that stood out at Think is that execution is happening faster than governance models can keep up. 

Most governance models are human-centric and process-heavy. That worked when change moved at a slower pace. But when execution accelerates—when individuals can create workflows that affect client experience or operational controls—governance itself can become the bottleneck. 

That’s why a concept like “governance as code” is compelling. To keep up, organizations need policies that are not just documented in a playbook, but enforceable in the flow of work. Those controls must preserve their effectiveness without becoming so slow that teams must route around them. 

3. AI is rebuilding the software development life cycle (SDLC).  

Think was a reminder that software delivery is entering a new era—one where writing code is no longer the dominant constraint.

Historically, the “long pole” in delivery was development time. Teams built processes, staffing models, and timelines around that reality. But as AI accelerates code generation and prototyping, constraints shift to what comes next: validating requirements, automating testing, ensuring security, managing deployments, and keeping the whole lifecycle auditable.  

This is why tools like IBM’s “Bob” drew so much attention at the conference. IBM positioned Bob as a way to support work across the software development life cycle—planning through testing and operations—rather than as a narrow coding assistant. It’s a notable signal: Enterprises want AI that integrates into the SDLC with structured guardrails—not just faster code output. 

From SEI’s perspective, this lines up with what we’re actively evaluating: how to re-engineer the SDLC for the AI age, including what it means to scale testing and governance as prototyping accelerates. 

But here’s a non-negotiable: As governance and development become automated, quality and safety must become first-order engineering outcomes—designed into the lifecycle, measured continuously, and enforced through testing and security controls. To support quality, SEI uses human-in-the-loop review, keeping work product ownership focused on the individual. We’re also testing “competitive AI,” where multiple models and processes are used to validate conclusions, routing disagreements to an exception process.  

From Think to “do” 

Looking ahead, I’m focused on a few questions that will define the next 12–18 months for enterprise AI:

• Where can we remove friction in ways that clients can feel immediately, without compromising precision?
• How do we evolve our “AI harness” to support multi-model choice, consistent governance, and scalable reuse across teams? 
• What does measurement look like when productivity gains must be balanced with risk, control, and client outcomes—and when ROI models are still emerging? 

If we’ve answered these questions, we’ll have improved operations measurably over the next 12 months—and created a foundation for sustainable growth. 

The era of AI industrialization has begun. In financial services, that means assembling the right combination of workflow redesign, multi-model flexibility, rigorous governance, and safety-by-design engineering. It’s why our partnership with IBM is paramount—and why I’m optimistic about what we can deliver for our clients and our teams as we continue to bring agentic AI into the core of SEI’s operations.