Artificial intelligence is rapidly embedding itself in professional services, with tax due diligence firms deploying AI agents to automate research, analyse vast datasets, draft reports and streamline workflows. But according to Taina Tech, many organisations are learning the hard way that managing AI agents demands a fundamentally different mindset from managing traditional software.
The instinctive approach, Taina Tech argues, is to dictate exactly how an agent should perform a task, mirroring decades of interaction with conventional technology through detailed process maps and rigidly defined steps. Yet AI agents are reasoning systems, capable of weighing objectives, spotting dependencies and recommending approaches. The counterintuitive lesson: rather than telling them how to do the work, ask them how they believe the work should be done.
Taina Tech recounts a firsthand experiment in building an agentic workflow for software development. The initial method was conventional, specifying every activity, checkpoint and documentation requirement. The results were acceptable but underwhelming. The agent followed instructions diligently but repeatedly hit friction points created by human assumptions, with some steps overly rigid and others built on dependencies that turned out to be unnecessary.
The turning point came when the approach was inverted. Rather than handing over a predefined methodology, the agent was given a simple objective, to design an effective lifecycle for agentic software development, and asked two questions: what information it needed, and how it would recommend structuring the process.
The difference was striking. The agent flagged overlooked information requirements around stakeholder objectives, testing expectations, governance and deployment constraints. It then proposed a workflow considerably more efficient than the human-designed version, built around iterative validation cycles, automated quality gates, continuous documentation and structured feedback, better suited to AI strengths and more adaptable to changing requirements.
For Taina Tech, the lesson is clear. Professionals often treat AI agents as junior employees needing granular instructions, when they behave more like highly capable specialists who thrive on clear objectives, context and well-defined outcomes. The human role shifts from process designer to outcome owner, though governance, oversight and professional judgement remain essential, particularly in tax due diligence where accuracy, compliance and defensibility are paramount.
The practical implication for due diligence teams is to open engagements with questions rather than instructions, asking agents what information they need, what assumptions they are making, what risks they see and what process would deliver the highest-quality outcome. These questions, Taina Tech notes, routinely expose blind spots and strip out unnecessary complexity.
For more, read the full story here.
Copyright © 2026 FinTech Global
Copyright © 2018 RegTech Analyst
