Professional services firms adopting AI agents may be approaching the technology with the wrong mindset, according to TAINA Technology’s Rich Kent, who argues that organisations risk limiting results by managing intelligent systems like traditional software.
TAINA Technology’s Richard Kent recently delved into how organisations can unlock greater value from AI agents, exploring why professional services firms may be limiting the technology’s potential by managing it like conventional software.
Artificial intelligence is becoming increasingly embedded across professional services, with tax teams experimenting with AI agents to automate research, analyse large datasets, generate reports and streamline workflows. However, successfully deploying these systems requires a different management approach from traditional automation tools.
Unlike conventional software, AI agents are designed to reason through objectives, identify dependencies and recommend approaches. This creates a challenge for organisations accustomed to mapping every process step and expecting technology to execute instructions exactly as written.
The instinctive approach is often to provide detailed workflows, define each stage and specify how information should move through a process. While this can create consistency, Kent argues it can also limit an AI agent’s ability to identify inefficiencies and suggest alternative approaches.
Kent discovered this while developing an agentic workflow for software development. His initial approach involved creating a detailed process map covering activities, checkpoints, documentation requirements and information flows between stages.
The agent followed the workflow, but the results exposed limitations in the approach. Some parts of the process were unnecessarily rigid, while others relied on assumptions about dependencies that were not required. The workflow reflected the way a human had designed the process rather than the way an AI agent could optimise it.
Kent then changed the approach by providing the agent with a single objective and asking two questions: what information it needed to complete the task effectively, and how it would recommend structuring the process.
The agent identified information gaps around stakeholder objectives, testing expectations, governance requirements, feedback loops and deployment constraints. It also proposed a workflow based around iterative validation cycles, automated quality checks, continuous documentation and structured feedback mechanisms.
The experience highlights a broader challenge for organisations deploying AI agents: the technology requires a different form of collaboration.
Rather than replacing professional judgement, AI agents are shifting the role of humans from process designers towards outcome owners. The objective, oversight and accountability remain human responsibilities, particularly in areas such as tax due diligence where accuracy, compliance and defensibility are critical.
For tax teams, this could change how AI is applied across workflows including transaction reviews, risk identification, workpaper generation and report preparation. Instead of beginning engagements with detailed instructions, teams may increasingly need to ask what information an agent requires, what assumptions it is making and what approach it recommends.
The shift reflects a wider change in how expertise is defined in the age of AI. As intelligent systems become more capable, the advantage may not come from providing the most detailed instructions, but from knowing the right questions to ask.
TAINA Technology’s analysis highlights that the value of AI agents will depend less on how precisely organisations can instruct them and more on how effectively teams can collaborate with them. By providing clear objectives, relevant context and appropriate oversight, businesses can allow AI agents to contribute to process design while keeping human judgement at the centre of decision-making.
Investors
The following investor(s) were tagged in this article.
