Beyond the Hype: What AI Actually Looks Like in a Dispute

Every technology company in the construction disputes space is talking about artificial intelligence. The marketing materials promise automated delay analysis, instant quantum assessments, and AI-generated expert reports. The reality is more nuanced,and, for practitioners who understand the difference, more useful.

The first distinction to draw is between genuine analytical capability and simple document processing. Most tools marketed as 'AI for construction disputes' are performing document classification, keyword extraction, and basic pattern matching. These are useful functions,being able to rapidly sort one hundred thousand project documents into chronological, thematic, or party-based categories saves significant analyst time. But they are not performing forensic analysis.

Forensic delay analysis requires an understanding of critical path methodology, the contractual framework governing extensions of time, and the factual matrix of the specific project. No current AI system can independently determine whether a delay event was on the critical path, whether concurrent delay applies, or whether the contractor's mitigation efforts were reasonable. These are matters of professional judgment that require human expertise.

What AI can do, effectively, is prepare the evidence base for human analysis. At Meritus Via, our proprietary tools ingest programme data,Primavera P6, Asta Powerproject, Microsoft Project schedules,and automatically map the critical path, identify float consumption patterns, and flag schedule anomalies. The system cross-references correspondence dates with programme updates to highlight potential areas of interest. It does not produce conclusions. It produces a structured evidence base that our senior practitioners can interrogate.

The same principle applies to quantum analysis. Our tools can parse bills of quantities, variation accounts, and cost reports to identify mathematical inconsistencies, duplicated claims, and unsupported figures. The output is a flagged dataset,not a damages assessment. The assessment remains the responsibility of the quantum expert who understands the contractual mechanism, the applicable formula for head office overheads, and the evidential burden for each head of claim.

This distinction is critical because the output of any technology-assisted analysis must be disclosable. Under CPR Part 35.3,¹ the expert's overriding duty is to the court,and that duty extends to full transparency about methodology. Practice Direction 35² requires experts to state the substance of all material instructions and the methodology employed. The Academy of Experts' guidance on AI use (2025)³ makes clear that experts must disclose any use of AI tools and are personally responsible for all material produced in their name, regardless of AI involvement. An expert who relies on an opaque AI system to produce conclusions will face significant difficulties under cross-examination. An expert who uses technology to structure and accelerate their own analysis,and can demonstrate exactly how,has a significant advantage.

The judiciary has taken an increasingly clear position. Lord Justice Birss, addressing the Expert Witness Institute Annual Conference (2025),⁴ emphasised that CPR 35.1 restricts expert evidence to that 'reasonably required to resolve the proceedings',a principle that demands focus and selectivity, not the indiscriminate volume that AI tools can generate. The SRA Code of Conduct⁵ has already been applied in cases involving AI-generated inaccuracies in legal proceedings, and the regulatory direction is unambiguous: the expert, not the tool, bears responsibility.

The gap between what technology companies promise and what practitioners actually need is substantial. The practitioners who will benefit most from AI are not those who hand over their analysis to a machine. They are those who use AI to work faster, more accurately, and with a more complete evidence base,while retaining full ownership of their professional opinions.