In conversation with Jelle Dhuyvetter

How is AI changing BPO, and why does this moment matter? 

For me, AI‑enabled BPO is really about one thing: outcomes. Not software, not features, not hype: outcomes. 

Traditionally, vendors deliver a platform, help get it live, and then hand over the keys. From that point on, the client owns the day‑to‑day complexity. What we’re seeing now is a real appetite for something different. A model where AI automates and coordinates large parts of the regulated workflow, while humans remain clearly responsible for validation, decisions, and accountability. 

This is the right moment for that shift. The pace of new AI tooling is intense, and most firms simply don’t have the time or capacity to experiment endlessly or stitch solutions together themselves. What they’re asking for instead is help reducing manual effort, creating consistency, and getting to reliable results faster. 

At the same time, the technology has reached a point where it can genuinely support end‑to‑end workflows — but only if it’s deployed with strong governance. Especially in financial services, letting AI run unchecked is not an option. 

That’s why I prefer to talk about human‑governed automation. AI can run and enhance regulated processes at scale, but control, transparency, and trust stay with people. Autonomy isn’t the goal — confidence is.

“Agentic AI” gets mentioned a lot. How should regulated firms really think about it? 

The term is indeed used very loosely and that’s where confusion starts. 

In reality, there’s a world of difference between AI that assists a task, AI that supports an end‑to‑end workflow, and AI that takes autonomous decisions. In regulated environments, full autonomy is rarely appropriate — and often not acceptable. 

What does make sense is agent‑enabled workflows with humans firmly in charge. AI can execute steps, connect processes, flag issues, and make suggestions. But responsibility for review, intervention, and approval remains human. 

This accountability is much more than merely a nice-to-have design preference. Frameworks like the EU AI Act make it very clear: the more material the use case, the greater the need for oversight, explainability, and documentation. And while that legislation is European, the underlying principles are influencing regulators far beyond Europe. 

In practice, the most valuable use of so‑called “agentic” technology in our world is orchestration. Helping experts do more, faster, and with greater consistency — without losing visibility or control. Compliance should never be self‑driving. 

What does this mean for Professional Services? How does the role change? 

It changes the role quite fundamentally. 

Professional Services can no longer be just about implementing software and moving on. In an AI‑enabled BPO model, the work becomes continuous and outcome-oriented. You design workflows, train and fine‑tune them, put governance in place, and keep improving them over time. 

Crucially, the expert oversight layer can work in different ways. In some cases, our specialists provide that validation and control. In others, the client’s own teams retain it. The automation is consistent — the responsibility model is flexible. 

That opens up some very natural service areas. There’s still upfront work: understanding client processes, configuring agent logic, handling exceptions, and designing governance and compliance frameworks. But the real value sits in what happens after go‑live. 

Ongoing outcome assurance becomes central. Monitoring performance, validating results, maintaining audit trails, managing human review, and adapting workflows as regulations or business priorities change.

What are the biggest challenges in making this work? 

There are four that stand out. 

First, people and culture. This isn’t just a technical shift — it’s a human one. Experts need to get comfortable training and governing automated processes, not just executing tasks themselves. Teams need to think in systems, not silos. And organisations need roles that blend domain expertise, process thinking, and AI literacy. 

Second, operating models. You can’t just bolt AI onto existing ways of working and hope for the best. Governance, escalation, risk management, and success metrics all need to be rethought. Human oversight has to be deliberately designed into the process — not added as an afterthought. 

Third, commercial models. Many service organisations are still geared around selling hours and projects. This model is about selling outcomes and ongoing value, which can feel less tangible at first. Pricing, contracting, and even how value is explained to clients all need to evolve. 

Finally, regulation. In high‑stakes environments, things like explainability, logging, conformity assessments, and monitoring are part of the service itself. Regulation such as the EU AI Act only reinforces that. 

That’s why I strongly believe in starting small. Pick a focused, high‑value use case. Prove the governance and the business value. Then scale. Trying to do everything at once is rarely realistic.