Artificial intelligence is becoming increasingly embedded within tax due diligence, helping professionals analyse transaction data, assess tax positions, identify potential risks and support reporting throughout the deal lifecycle. However, the ability to unlock value from these tools may depend less on the technology itself and more on how effectively professionals communicate with AI agents.
According to analysis from TAINA Technology, one of the key lessons emerging from early AI adoption is that agents perform best when given clear, direct instructions. Rather than asking AI systems to explore a topic or interpret broad requests, professionals can achieve stronger outcomes by assigning specific tasks with defined objectives.
This approach challenges traditional professional communication habits, where users are often encouraged to ask questions carefully, provide flexibility and allow room for interpretation. TAINA Technology’s analysis, written by Richard Kent, suggests AI agents operate differently, producing more reliable results when instructions are structured around clear actions and expected outcomes.
The difference becomes particularly apparent in tax due diligence workflows. A vague request asking an AI agent to review deferred tax calculations and identify potential concerns may leave too much uncertainty around what should be assessed. A more precise instruction directing the agent to analyse calculations, identify inconsistencies, highlight errors and examine unusual assumptions provides a clearer framework for generating useful insights.
This focus on precision is especially important in areas such as tax return reviews, historical liability assessments and transaction structuring, where unclear instructions can create inefficiencies and reduce confidence in AI-generated outputs.
However, TAINA Technology’s analysis does not suggest that prompts should simply become shorter or more rigid. Kent suggests that objectives and constraints remain critical factors in determining the quality of an AI agent’s response. Once those foundations are established, task-based instructions can help agents deliver more consistent and relevant results.
There are also situations where professionals understand the desired outcome but are uncertain about the best methodology. In these cases, TAINA Technology recommends maintaining an imperative approach while allowing the AI agent to determine the most effective process. For example, rather than asking an agent for general thoughts on FATCA compliance testing, users can instruct it to identify the most effective method for assessing compliance across a dataset.
As AI becomes increasingly integrated into tax due diligence processes, firms will need to develop new capabilities around managing and directing AI agents. TAINA Technology’s analysis highlights that the organisations gaining the most value from AI may not necessarily be those using the most advanced tools, but those that learn how to communicate with them effectively.
Kent explains that the shift required includes providing clear instructions, defining objectives and using action-oriented language. TAINA Technology’s analysis suggests that improving how professionals interact with AI agents could become a key differentiator in determining whether these systems deliver meaningful value or simply generate additional output.
