The Human-and-AI Operating Model: How to Make It Work

In this edition of The Future of Work, I’m excited to share an article from Alexander Michael, Vice President, Global Practice Area Leader, Information & Communications Technology at Frost & Sullivan.

Alex wrote this piece on the topic he’ll bring to the stage at The Big Game on October 15: how humans and AI can work together on the front line. It's a great read whether you'll be joining us or not. Enjoy!

✦

The scale of AI adoption is no longer in doubt. Frost & Sullivan research shows that 85% of organizations are already engaged with generative AI in the CX space. Almost half of customer interactions are already handled by virtual agents. AI is already an established part of service delivery rather than a peripheral experiment.

However, AI adoption is advancing faster than the structures needed to govern it. Businesses add AI capabilities one use case at a time, while workflows and KPIs remain anchored to an operating model designed for an exclusively human workforce.

This is why successful AI transformation is about answering one hard question: how should people and AI work together to deliver better outcomes for everyone? That question will be at the heart of the Human-and-AI Operating Model: How to Make It Work panel at The Big Game 2026, where I will join representatives from Centrical and Firstsource to explore what it takes to make the model work in practice.

Frost & Sullivan’s research into back-office technology tells a similar story. Over 40% of organizations have conducted POCs with agentic AI, over a quarter have implemented at least one capability, and three-quarters expect to expand agentic AI into further use cases within the next 2-3 years. For me, the conclusion has to be that the human+AI operating model already exists in many organizations. The problem is just that it has emerged through successive technology decisions rather than deliberate design.

In my work with end-user organizations, I find that decision-makers are no longer necessarily asking for AI. They are asking for an overarching operating model that works better rather than requesting another isolated feature. This distinction matters to me. I do not think the starting question should be, “Where can we introduce AI?” It should be: How should work, information and accountability move through the organization, and where can AI make that model faster, more consistent and more adaptive? Because automating processes in a bad model does not fix it. It probably makes it worse.

This leads me to one of the central arguments I hope we will explore at The Big Game 2026: Automating a fragmented operating model can make individual steps faster without making the organization more effective. AI cannot compensate for unclear ownership. Nor can it resolve poorly designed hand-offs simply by making them happen more quickly.

I do not think a productive human-and-AI model should begin with a binary division between activities performed by people and activities performed by machines. We should allocate work according to comparative advantage. AI is good at speed, scale and precision. Humans are good at ambiguous situations, empathy and relationships. We all know this. So, our operating model must cater to at least three categories of work: human-led, AI-assisted and AI-executed. A healthy operating model must also recognize that every AI deployment changes the human job around it. People can no longer be evaluated against the same old KPIs, and their managers must coach them for a different type of work.

AI can improve the employee experience, but design determines the outcome. According to Frost & Sullivan’s research, AI implementation has improved agent satisfaction in almost 90% of organizations. This supports a human+AI model based on augmentation, because AI can absorb administrative effort and prepare human agents for the higher-value work they are best placed to do. But AI does not determine the employee experience. The operating model does. Imagine if our agents were made to face a relentless barrage of emotionally taxing, difficult interactions without the authority or recovery time they need ... This is why I believe the operating model is also about job quality. Efficiency is not a complete operating principle.

Organizations want better execution and AI can make that happen, but only when it forms part of an operating model designed around positive outcomes. For me, that is the real challenge for the human+AI enterprise. It is redesigning the organization so that people and AI can work together effectively, while people remain in control of the decisions and judgements that matter.

And I could go on and on because we also need to look at orchestration, governance, the role of supervisors, the incentive structure and fragmented technology stacks. Hopefully, we’ll have time to explore all that and more at The Big Game 2026.