Thought Leadership

At CROP, we believe that innovation only becomes valuable when it helps our clients, our colleagues and the quality of our work. That is why our journey into AI did not start with the question: “Which tool should we buy?” It started with a more important question: “What do we want AI to help us achieve?”

During the PrimeGlobal Business Leadership Forum, we had the opportunity to share our experiences so far: what worked, what did not, and what we are currently learning as a mid-sized professional services firm working across audit, tax, legal and advisory services.

Why our position matters

CROP is large enough to have multiple service lines, complex client work and a strong need for quality and consistency. At the same time, we are not so large that we can simply build every AI solution ourselves from scratch. That position makes our AI journey both interesting and challenging.

Many off-the-shelf solutions do not fully meet the specific needs of all our service lines. Building a complete solution ourselves would take significant time, money and resources. So our challenge has been to find the right balance: being ambitious, but realistic; innovative, but careful; open to experimentation, but guided by a clear direction.

Starting small: useful, but not enough

Our AI journey began with small initiatives across the organization. Colleagues experimented with tools, tested ideas and explored possible use cases within their own teams. That was valuable. It created energy, curiosity and early learning.

But we also discovered the downside of starting small without a shared goal. Initiatives can become fragmented. Teams may optimize for their own needs, while the organization as a whole loses overview. Questions around compliance, data quality, professional standards and consistency quickly become important.

In other words: experimentation is necessary, but it cannot stand on its own.

That is why we made innovation a clear priority at board level.

We created an innovation board with representation from all service lines, worked on an AI policy and started defining a company-wide goal that could guide our decisions. The aim was not to stop local initiatives, but to connect them, learn from them and make sure they contribute to a broader strategy.

Our Tax AI ambition

Our goal is to use AI in a way that combines several important sources of knowledge: internal and confidential client data, reliable public sources, public professional literature and restricted paid professional literature. Based on that combined information, we want AI to support analysis, research and document creation in a tax and legal context.

That ambition is deliberately specific. For our type of work, quality matters. Context matters. Source reliability matters. AI can only be truly useful if it works with the right information and if people understand both its possibilities and its limitations.

Lessons from the market

Over the past two years, we have explored different solutions and providers: document management tools, professional literature providers, tax and legal AI start-ups, RPA solutions, Copilot and ChatGPT. Each option brought useful insights.

One important lesson was that no single solution automatically solves everything. Some providers are strong in one area but do not cover all our service lines. Others lack access to the professional literature we need. Some tools are powerful, but not sufficiently tailored to tax and legal work. And internally, we also had to be honest about our own readiness: the quality of our data, the way colleagues use AI and the expectation that AI will sometimes perform “miracles”.

The Lexgen proof of concept

Our most recent proof of concept with Lexgen has been one of the more promising steps so far. The reason is not simply the technology itself, but the way we approached the process.

We involved multiple service lines, included colleagues with different perspectives and tested whether internal and external sources could be combined in a meaningful way. We also looked closely at the role of suppliers of paid professional literature, because access to reliable sources is essential in our field.

The proof of concept helped us move from abstract discussions to practical experience. Colleagues could test use cases, share feedback and see where AI added value — and where it still fell short. That feedback was not always easy, but it was exactly what we needed. It helped us improve the process and made the conversation about AI more concrete and constructive.

Where we are now

We are still on the journey. There is no final blueprint, and the market is changing quickly. But we have learned several important lessons.

• First, a clear company-wide goal is important. Without it, AI initiatives risk becoming scattered experiments.
• Second, governance matters. AI touches client data, compliance, quality and professional responsibility. It cannot be treated as just another software tool.
• Third, people matter most. Training colleagues is not only about teaching prompts. It is about helping them understand when AI is useful, when it is not, and how to apply professional judgement.
• Finally, AI should strengthen the quality of our work, not replace the expertise behind it. For CROP, the goal is not to chase every new development. The goal is to use technology in a way that supports our clients, empowers our colleagues and prepares our organization for the future.