The Responsible AI Summit in Chicago in late June provided a timely reminder that the conversation around AI has changed. Here we review some key thoughts arising from the summit discussions including the panel moderated by Chinmoy Bhatiya, Capco AI Lead.
Not long ago, many organizations were still asking whether they should use AI, where to experiment and how quickly they could move. Today, the questions are sharper and more consequential. How do we scale AI responsibly? Who is accountable when AI influences decisions or takes actions?
What made the summit especially valuable was the balance of perspectives in the room. This was not a conversation limited to technologists or policy experts. It brought together leaders from legal, risk, compliance, privacy, cybersecurity, data, product, operations and business strategy.
That mix matters because AI does not live neatly inside one function. It changes how companies serve customers, make decisions, manage risk, design products, train employees and meet regulatory expectations.
A clear theme emerged across the discussions: Responsible AI can no longer be treated as a set of abstract principles sitting on a website or in a policy document. Principles are important, but they are only the starting point.
Responsible AI is moving from aspiration to operation. Principles still matter, but organizations now need governance that can hold up in day-to-day business: clear ownership, practical controls, escalation paths, documentation and accountability.
Good governance should not slow innovation. Done well, it gives teams the confidence to move faster because they understand the rules of the road. When people understand what is allowed, what requires review, what risks must be managed, and who needs to be involved, they can move with more clarity and less friction. Responsible AI becomes the structure that allows AI to scale safely.
Chinmoy Bhatiya moderated a key panel discussion entitled, “Governing AI Agents: Technology, Employee, or Both?” The panel explored one of the most important questions now emerging in enterprise AI: when AI agents can access data, use tools, make decisions, and trigger actions across workflows, how should they be governed?
An AI agent is not just a chatbot answering a question. AI agents can adapt their behavior based on context, data and goals. That makes governance more complex and raises new questions about permissions, oversight, boundaries and intervention points. In many cases, an agent may be able to take steps on behalf of a user, interact with systems, recommend decisions, or execute parts of a business process.
The panel framed the issue around capability rather than labels. The question is not simply whether an AI agent is ‘technology’ or something more like a ‘digital employee’. The better question is: what can it actually do?
For example, what data can it access? What actions can it take? Can it affect a customer, an employee, a transaction or a regulated process? Can a human review or reverse its actions? Who owns the outcome if something goes wrong?
A key challenge here concerns the Identity & Access Management (IAM) systems that presently govern access by human professionals and service accounts. When an agent executes steps, and uses tools and data based on its own decision making, how should IAM be handled?
One idea discussed was whether agents should have their own non-human identities within IAM systems, rather than operating invisibly through a user’s credentials or a generic service account. That would make their actions easier to track, audit and govern.
The final lightning round from the panel brought the discussion into practical focus. Before organizations scale Agentic AI, they need documented governance, end-to-end accountability, clear behavioral expectations and strong human oversight. Governance must follow capability. The more an AI system can see, decide or do, the stronger the controls around it need to be.
Another important theme from the summit was AI literacy. As AI tools become easier to use, more people across an organization will influence how AI is adopted. That includes people who may not have technical backgrounds but still make important choices about use cases, vendors, data, outputs and customer impact.
Responsible AI cannot sit only with a central governance team. Business users, product owners, executives, technologists, lawyers, risk leaders and compliance teams all need a shared understanding of what AI can do, where it can fail, and how to use it responsibly.
The summit also reinforced the importance of monitoring AI after deployment. Approval is not the finish line. AI systems can drift, produce unexpected outputs, rely on changing data, or be used in ways that were not originally intended. Responsible organizations need ways to monitor performance, evaluate outcomes, capture evidence, escalate concerns and respond when something goes wrong.
What stood out was the shared sense of urgency and awareness among industry leaders. Across sectors, organizations are asking similar questions: How do we govern AI without stifling innovation? How do we create accountability across complex ecosystems? How do we manage third-party risk? How do we build trust with customers, employees, regulators and partners in systems that are complex and evolving?
For Capco, it was energizing to be part of that conversation alongside organizations, peers and industry leaders who are approaching AI with both ambition and care. The Responsible AI Summit made one thing clear: Responsible AI is no longer a side conversation. It is becoming central to how modern organizations build, compete and earn trust.