why leadership, data and business models must evolve too

AI is changing more than technology:

AI is reshaping business models, the work people do and the role of leadership teams. Organisations must therefore learn, adapt and transform faster than the markets they operate in, says Matthijs Mons, Director Experience, Software & Development at Conclusion Benelux. Sticking with the traditional time-and-materials model is simply not a realistic option.

What do you sell when AI enables you to complete the same assignment in a fraction of the time? That question affects every professional services organisation whose revenue depends on billable hours. Clients expect to benefit from greater efficiency, while providers need to invest heavily in technology, training, new roles and updated processes. At Conclusion, this challenge sits at the heart of a broad AI programme.

Mons sees tremendous opportunities, but also significant risks for organisations that fail to adapt quickly enough. Conclusion aims to move, learn and adapt faster than the market around it changes, because everything that seems relevant today could be outdated within six months.

According to Mons, AI makes it increasingly difficult to link revenue growth solely to headcount and billable hours. This affects professional services firms across industries, from IT and consulting to legal services.

As a result, Conclusion is increasingly adopting alternative commercial models, including fixed-price engagements. By analysing projects thoroughly upfront and applying AI intelligently, the organisation can estimate delivery costs more accurately. The challenge lies in combining an attractive proposition for clients with sustainable margins.

Conclusion is also exploring models where compensation is linked directly to the value delivered. What outcomes does a client achieve, and can the service provider share in that value creation? Mons sees significant potential in this approach, although clients are not always ready for it yet. At the same time, subscription-based services and recurring revenue models are becoming increasingly important.

As Mons explains, revenue growth and FTE growth will become increasingly disconnected from one another.

Among clients, Mons sees both AI frontrunners and organisations that are still taking a wait-and-see approach. Enthusiasm is widespread. Employees are building solutions and teams are experimenting with new applications. However, embedding those applications into day-to-day operations is a very different challenge.

According to Mons, getting an AI solution into production is one thing; delivering measurable business improvement is something else entirely.

To achieve that, core processes need to be redesigned, systems integrated and employees equipped with the right skills. Data quality and clear ownership are equally essential. While the potential of AI is significant, only a limited number of organisations have so far succeeded in generating value at scale.

That value extends well beyond cost reduction. Organisations can use AI to serve customers faster and more effectively. Mons therefore encourages leaders to critically assess their existing services. How easy is it for clients to get support? How useful are the answers they receive? And how can AI help improve that experience?

The next step is to prioritise which processes should be transformed first. Conclusion often develops business cases for individual processes to assess the feasibility of change, the required investment, the expected benefits and the implementation approach.

According to Mons, an employee who generates significant value with AI may well justify several hundred pounds or euros in additional monthly costs. The real issue arises when heavy usage generates little or no value.

The worst situation is when people use AI extensively without really knowing how to use it effectively.

Repeatedly performing the same task with slightly different prompts drives up costs unnecessarily. Training and awareness are therefore just as important as technical controls.

Model selection matters too. Conclusion is developing self-hosted open-source models alongside advanced models from external providers. Not every task requires the most sophisticated solution. As Mons notes, a plumber does not need to know Latin.

Similarly, not every process requires an AI agent. For structured, repeatable tasks, conventional software may be a more efficient option. Sometimes, the best solution is simply to build software, which can itself now be developed wholly or partly with the help of AI.

Organisations must also prepare for growing adoption. Even if the cost per token decreases, total costs may still rise.

AI exposes existing data challenges, from outdated information and inaccessible systems to unclear ownership structures.

Mons points out that while organisations may want to move quickly with AI, the technology itself is rarely the limiting factor. More often, the constraint lies in the underlying data.

That does not mean organisations must first solve every data challenge before getting started. Instead, they should focus on a specific process and identify the data required to support it.

Conclusion refers to this as a data product.

Organisations do not need to unlock every dataset for every AI agent, nor can they realistically build every agent at the same time. Mons recommends developing both the application and the supporting data product in parallel, with clear agreements around quality, access and ownership. In some cases, a well-designed dashboard can already be a valuable first step.

To encourage knowledge sharing and accelerate adoption, Conclusion has developed its own AI platform. The platform allows its operating companies to share use cases, components and expertise while reducing duplication, providing cost transparency and supporting secure use of AI.

As Mons puts it, there is little value in reinventing the wheel hundreds of times.

The platform is also designed to provide flexibility, allowing organisations to switch models or technology providers when necessary. A vendor leading today may not remain the market leader tomorrow.

Oversight remains essential, particularly when deploying AI agents. Testing and governance frameworks alone are insufficient if nobody monitors outcomes in practice. Organisations must continue to review performance, monitor behaviour and intervene when necessary.

Conclusion uses four employee profiles and three capability levels: beginner, advanced and expert. An assessment helps determine the most appropriate learning path. In addition to self-study, AI coaches support employees within projects.

According to Mons, people learn most effectively by doing their work and adapting the way they work.

AI is putting the greatest pressure on standardised work. This is not simply a matter of young versus experienced employees. Organisations will continue to need junior talent, but AI raises new questions about how future expertise is developed.

Mons believes that one of the major challenges for organisations is ensuring that today’s junior professionals have opportunities to gain the experience required to become tomorrow’s senior experts.

Looking ahead, he expects the traditional organisational pyramid to evolve into what he describes as a diamond supported by a platform of AI agents, with a fundamentally different distribution of responsibilities and human contribution.

Conclusion’s ambition is to continue developing and retaining as many employees as possible. Mons believes that most people have little reason to fear for their jobs, provided they continue investing in their own development.