Understanding How AI Works Under the Hood, Leads to Better Business Outcomes

Artificial intelligence is transforming enterprise operations, but technology alone won't determine who succeeds. In this blog, Guido van den Boom explores how organizations can move beyond AI pilots to build production-ready AI systems, strengthen long-term client partnerships, empower employees with agentic AI, and create measurable business value. His message is clear: the companies that thrive will be those whose leaders combine technological understanding with the vision to reinvent how work gets done. 

The AI revolution is moving more slowly than many headlines suggest, yet its long-term business impact is still widely underestimated. While multinationals are quick to attribute workforce reductions to AI, we need to focus on the quieter revolution unfolding under the hood, and to the fundamental questions every executive should be asking now. "In the end, AI does not decide what happens to your company. That comes down to the vision and creativity of the people making the C-level decisions. 

You really have to take the reports of multinationals now cutting thousands of jobs 'because agents are taking over the work' with a very large pinch of salt. At the same time, I see developments on the horizon that will seriously upend our work and our world in the years ahead, not only at our own company and our clients, but certainly across society as a whole. We need to prepare for these developments now. 

As chief commercial officer of Xebia, I help shape the firm's commercial strategy across one of its largest markets. Xebia partners with enterprise organizations across financial services, retail, healthcare, media and technology to solve complex business and technology challenges. With 4,500+ professionals worldwide, a significant portion of Xebia’s workforce is dedicated to software development, placing them squarely in the firing line of Claude Code and similar AI services. Steering Xebia through that minefield is part of my role. 

Over the past several years, Xebia has evolved from a traditional consulting model toward long-term strategic partnerships with its largest enterprise clients. Seen from the Netherlands, those include ING, dsm-firmenich, Rituals, ASML, Sligro, Rabobank and Orbia/Wavin, as well as digital natives such as Flow Traders and Mollie. Working alongside these organizations for many years has given Xebia a front-row seat to the rapid evolution of enterprise data, cloud and AI. That long-term perspective now enables Xebia to help clients move beyond experimentation and embed AI into core business operations.  

From Technology Delivery to Strategic Advisor  

Xebia has worked for ING for more than 20 years. The bank was always focused on innovative digital technology, particularly after former CEO Ralph Hamers set out the ambition of becoming "an IT company with a banking license". Our role in helping ING introduce the Spotify model for agile ways of working reflects how many of our client relationships evolve. We are often brought in not to define the strategy, but to help organizations operationalise it at scale. 

Initially, the bank asked a strategy consulting firm to develop an agile operating model. When it came to embedding that model across the organization, ING turned to Xebia. We deployed a team of ten agile coaches to support management and team leaders throughout the transformation. That's where we create the greatest value: turning strategic ambition into operational reality. 

We can do that because we have developed deep knowledge of our clients' technology landscapes over many years. Historically, our role focused on ensuring technology platforms operated reliably and efficiently. Today, as data and AI become central to business strategy, our role has expanded far beyond implementation. As a result, we work alongside executive leadership on strategic business decisions rather than purely technology execution.  

Moving Enterprise AI Beyond Traditional Chatbots  

ING’s customer service chatbot originally relied on decision trees and structured knowledge bases. But the deflection rate, meaning the percentage of inquiries resolved without human intervention, was far too low, increasing costs and reducing customer satisfaction. The rapid rise of LLMs suddenly created the opportunity to fundamentally rethink that approch. 

A bank like ING holds a treasure trove of historical customer communication: years of dialogues, emails, complaints and the answers to them, stored across various databases. Collectively, this represents one of the organization's most valuable knowledge assets. Where a classic chatbot can only fall back on a predefined decision tree, you can train a language model, in this case Google Vertex AI, on all those real past customer interactions. The model then learns not only what ING answers, but also how: in the right tone of voice, and in line with all the rules and regulations. 

In the highly regulated environment of a bank in particular, that is no simple matter. A key challenge is that LLMs are non-deterministic: they can go in any direction, producing the most unexpected and undesirable answers. So you have to "tame" such a model and govern its behavior. But if you set out to assess every answer manually for quality, it will simply take you years to train and steer such a model. 

Building Trustworthy AI Through Systematic Evaluation 

Following extensive experimentation, we built our own evaluation framework: a systematic way to assess and monitor the model's output at scale. Today, an evaluation framework like that comes as standard when you roll out agents from the major cloud providers. Back then it did not yet exist, so it meant a great deal of extra work. At the same time, it also yielded a great deal of valuable insight. 

How exactly are these models put together, and how do you make sure they do precisely what you want? The better you understand what is happening under the hood, the greater the chance of a strong business result on the bottom line. Today, the ING chatbot is among the few production-grade generative AI customer service solutions operating autonomously within a highly regulated banking environment. For Xebia, building that expertise has fundamentally changed our role—from implementing technology to helping enterprise leaders make confident AI decisions grounded in operational experience. 

The Future of AI Is Connected, Contextual and Autonomous 

Today, most organizations are still using AI to optimize existing processes rather than fundamentally redesign how work is done. At a global chemicals group, for example, we are helping the marketing department to identify market segments they do not yet serve and can move into. For a large FMCG company, we have set up internal document classification, where the system generates the relevant insights itself. Work that previously took weeks can now be completed in hours while significantly reducing operational costs.  

At a global cosmetics player, we are already taking a step further by extending AI beyond knowledge workers to frontline employees. They have to know all kinds of things: how to open and close the store, how to deal with a customer, what to do in the event of theft. Thanks to our new "store buddy", they will soon be able to ask any question in natural language to a tablet. Behind the scenes, a "super-agent" with a small army of other agents makes sure every answer is pulled in real time from systems such as ServiceNow and Salesforce. 

Connecting enterprise data across systems has become one of the most important foundations for successful AI adoption. Many AI initiatives fail to deliver meaningful business value because the models lack sufficient organizational context. By feeding internal data to agentic AI systems in "bite-sized chunks", the results improve at remarkable speed.

Helping Organization Prepare for an AI-Native Future 

One of the sectors facing the greatest impact is software development, which is our core business. Tasks that once required large engineering teams can increasingly be completed through smaller teams augmented by AI agents. And when I look at our clients, I see the role of people is shifting from executing every task themselves to directing, validating and orchestrating AI-assisted workflows.  

Investing in the development of your own people is therefore more important than ever. Our developers, for instance, will have to grow into a role as "orchestrator". Rather than spending every hour writing code manually, they increasingly guide AI agents, validate outputs, establish guardrails and ensure software meets business, quality and security requirements. That calls for a new kind of professional: people who understand precisely both how AI systems work and how technology creates measurable business value for the client. 

Leadership Determines AI's Business Impact 

The same applies across the organizations. IKEA, for example, rolled out an AI chatbot that reportedly handles almost half of all customer conversations. The management could have used that as an argument to cut a large part of the workforce and generate more profit for shareholders. Instead, they looked precisely at what customer pain point the bot could not handle. Turned out, customers seeking advice on furnishing their homes, still preferred human taste and a listening ear. 

As a result, IKEA retrained a large share of its customer service staff as interior advisers. That transformation generated more than €1bn in new revenue in the first year. Instead of asking "How do we save as much as possible with AI?", IKEA looked at new possibilities.  

AI will have a major impact in the years ahead on companies, their people and the way they work together. In the end, AI does not decide what happens to your company. That comes down to the vision, creativity and business acumen of the people making the C-level decisions. 

Profile: Guido van den Boom

Guido van den Boom studied history but began his career in consultancy more than 25 years ago at Magnus Management Consultants, where he led major SAP transformations. After a spell at HEMA, where he set up the first web shop in the Netherlands, he spent ten years at Accenture, leading Accenture Interactive (later Accenture Song) in the Netherlands, among other roles. Next, he joined Capgemini, where he was head of digital customer experience. Guido joined Xebia in 2018 to build up its digital strategy practice. He has been Chief Commercial Officer of the Benelux since 2025. 

This article has originally been featured in Dutch on Businesswise.nl. Click here for the Dutch version.