The gap between Artificial Intelligence (AI) potential and implementation has narrowed considerably. In Europe’s tax and accounting sector, AI adoption has risen from 8 percent in 2024 to 42 percent this year according to our research. Firms are moving beyond experimentation to real deployment. Yet progress happens only when technology is guided by sector expertise and customer needs.
Firms have learned that effective adoption isn’t about budgets seeking use cases. Success starts with understanding workflow challenges. The key principle: attune AI systems to customer needs and sector expertise, not technology for its own sake.
This article draws on years of deployment experience: what creates value, where implementations fail, and how firms can build strategies that deliver real business impact.
The reality behind the AI hype
Despite rising adoption, firms face consistent challenges moving from pilots to scaled deployment. Three barriers emerge repeatedly:
Trust and transparency
Firms often assume clients want maximum automation. The reality differs. Accountants resist “black box” implementations even when technically capable. Take automated tax return flows designed to collect data, organise returns, and submit to authorities with one click. These accountants sought review checkpoints and control before submitting returns for thousands of clients.
The concern is visibility into decision-making. Finance professionals need to see what information was collected, understand what’s on the return, and control accuracy before submission. Trust requires explainability, confidence in results and control over key decisions.
While cybersecurity and data privacy are critical, limited visibility undermines confidence. Without transparency, professionals can’t feel in control.
Change management matters more than technology
Even the best tools can fail without proper change management. Providing new tools without addressing transformation creates pushback. Tax and accounting firms often favour familiar processes over efficiency.
We’ve seen this before. Software developers initially feared coding tools as job threats. It took significant effort to reframe them as career enhancement. The barrier is real. Those who don’t embrace these capabilities are at risk. Those who do gain efficiency benefits.
Firms must guide users through the transition. Training sessions, experimentation programmes, and clear change management are vital. Once people see the power, uptake multiplies. But expect initial resistance.
Starting with customer needs
Many implementations fail because firms start with budgets and search for use cases. This misses the mark. Generic tools lack specialised knowledge required for professional services. Without sector expertise embedded in these systems, outputs lack the accuracy that accountants require.
Building strategic AI maturity
Addressing visibility concerns, internal resistance, and misaligned implementation requires a deliberate approach grounded in four core principles:
1. Ground solutions in sector expertise
Successful implementation requires combining technology with sector-specific knowledge, trusted content, and specialised understanding. Focus on solutions addressing real use cases.
For tax and accounting, this means leveraging intellectual property around customer workflows, regulatory content, legal requirements, and industry standards. By training systems on specialist-validated data, we create explainable outputs that cite sources, demonstrate reasoning, and provide the visibility accountants need.
2. Adopt human-centric design
Position these technologies as augmentation that helps staff gain efficiency benefits. Address the talent shortage by enabling accountants and tax advisors to handle more clients and focus on higher-value advisory work.
Build transparency and explainability into every feature. Users should understand what the system does, see data sources, and retain authority at critical points. Remove manual drudgery and repetitive tasks, freeing accountants for judgement-based, relationship-driven advisory services.
Work with clients throughout development, gathering feedback before launch. Refine based on real usage patterns. Invest in user experience to ensure features are intuitive and fit naturally into workflows. If implementation requires extensive training, the approach has failed.
3. Implement security and privacy frameworks
Protecting customer data is non-negotiable. Firms with established programmes have frameworks ready to apply to emerging technologies, building on principles of responsible practice.
Infrastructure choices matter. Opt for a privacy-by-design architecture that meets GDPR and local regulations to build data protection into AI and agentic systems from the start.
Clear internal policies define which tools employees can access and what information can be processed. Training on data privacy and security supports these policies.
4. Measure value through business outcomes
Define success through metrics: time savings on routine tasks, scalability, reduced error rates, and improved compliance deadline management.
Consider the numbers. If automated flows save time, practices could scale from servicing 10,000 to 15,000 customers with similar resources. These systems ensure tasks are completed on time, even during peak periods. Moreover, machines trained on millions of sources can achieve greater accuracy than individuals—especially when guided by the right human oversight.
The key here is to focus investment where staff spend most effort.
The evolution to agentic systems
Beyond conversational interfaces and traditional machine learning, the next chapter in AI evolution involves autonomous agents capable of executing complex task sequences independently. Agentic capabilities include proactively managing operations, anticipating needs, and executing multi-step processes.
For example, an agent might identify an upcoming tax return deadline and recognise missing client information. After requesting and receiving details, it assembles the return for tax adviser or accountant review. This vision of a “zero-touch tax return” sees automation handle end-to-end operations with human oversight at final approval.
Although traditional algorithms continue handling core functions like optical character recognition (OCR) data extraction from invoices, generative capabilities improve user interaction, visibility, and orchestration. We’re in early days for implementation. The primary challenge is understanding complete customer operations to apply agentic capabilities where they add most value.
Moving from experimentation to maturity
Achieving AI maturity requires moving beyond experimentation to deliberate implementation. Firms that scale effectively ground capabilities in sector expertise while prioritising human-centric approaches with visibility. They focus on real customer pain points, invest in change management, and iterate based on feedback.
Value emerges when technology, particularly solutions that blend established and generative features, is grounded in domain knowledge and customer trust. At the end of the day, the true measure of success is how effectively technology serves its users.
