EWJ August 62 2025 web - Journal - Page 61
between drawing revisions; verify the validity and
extent of the variation claims; link drawing changes to
the claims pursued in the dispute; and providie visual
presentation of the evidence that can be incorporated
into the report.
of a new set of legal, ethical, and practical guidelines to
ensure its implementation with caution and clarity.
Below are some key concerns, illustrated with
examples relevant to construction disputes.
1. Confidentiality and Data Protection
One of the primary concerns when using AI in expert
work is the confidentiality of the project data. As many
AI tools operate in cloud-based environments, experts
must upload cost ledgers, drawings, or contractual
documents to external servers to obtain AI assistance.
For instance, using ChatGPT, Jasper, or Microsoft
Copilot without an enterprise agreement could result
in data being processed outside of a jurisdiction, potentially breaching confidentiality clauses or privacy
laws. This risk is amplified when parties have not obtained consent from clients or stakeholders to handle
sensitive records through third-party platforms.
2. Automated Quantity Takeoffs
AI-driven software like Togal.AI5 and Kreo6 can
automatically detect and measure elements such as
walls, doors, windows, and areas on construction
drawings and classify rooms and finishes in seconds,
helping to reduce manual measurement time. This
can also help reduce the time required to measure
variations, generate visual summaries of the analysis
performed by the expert, reduce manual input errors,
and enhance reproducibility.
3. Document Review and Data Analysis
Microsoft Excel CoPilot (GPT integration) and Power
BI with GPT plugins, which are integrated with
spreadsheet tools, allow users to ask natural-language
questions about ledger data, simplifying the data interrogation. AI can also generate pivot-table summaries, charts, and statistics from raw cost entries.
Chat GPT Advanced Data Analysis modes can import
large CSV ledgers and perform a comprehensive statistical review in one step.
2. Admissibility and Transparency
In formal proceedings, the admissibility of a quantum
expert’s report hinges on several factors, with transparency, logical reasoning, and openness to scrutiny
being particularly significant. Therefore, if a quantum
expert presents conclusions based on AI analysis but
cannot clearly explain how the AI tool arrived at its
output, the evidence may be deemed inadmissible.
Therefore, any analysis conducted using AI assistance
must be accompanied by supporting documentation
that clearly outlines the data inputs, logic paths, and algorithmic assumptions to avoid a tribunal considering
it a “black box” output. The is explored further below.
These tools allow quantum experts to move beyond
sample analysis and assess full datasets, such as complete cost ledgers with over 100,000 entries, providing much higher levels of rigour and eliminating
potential sampling biases.
3. Overreliance on Technology
While AI can efficiently surface patterns or outliers, it
cannot determine contractual relevance or practical
context. For example, a Power BI dashboard may flag
a 20% cost spike in a specific trade, but only a human
expert can interpret whether this is due to an agreed
variation, late delivery, or misallocation. Relying solely
on statistical output without expert interpretation risks
drawing conclusions that are irrelevant or misleading.
4. Valuation and Costing
Tools like Causeway Estimating7, PriceAL (by C-Link)8,
and Buildxact enable the quantum expert to achieve
improved speed, accuracy, and consistency in pricing
and valuation exercises. For example, by using these
tools, the expert can benchmark the assessment
against historical project data to validate the reasonableness of claimed rates.
5. Report Drafting and Documentation
In line with other professional disciplines, we are also
seeing the increased use of AI-driven tools such as
ChatGPT Enterprise,9 Claude,10 Jasper11 (among others) to generate structured, readable, and consistent
report narratives. Whilst this may accelerate the drafting of repetitive sections and ensure a consistent structure across multiple claims, the same caveats apply as
the quantum expert remains the author of the report.
4. Quality and Consistency of Source Data
For AI tools to generate accurate results, it is necessary
to feed them with clear and structured data. However,
in many construction disputes, the project documents
may be incomplete, inconsistent, or misclassified. For
example, subcontractors might submit invoices in various forms, or labour entries might be recorded under
different cost codes in different months. Unless carefully cleaned and verified by the expert beforehand,
feeding disorganised data into an AI model (whether
via Excel Copilot or custom Python scripts) might
result in skewed outputs and flawed valuations.
Challenges and Limitations in Using AI for
Quantum Expert Work
Although there are significant opportunities to
increase the effectiveness, precision, and scalability of
quantum expert work through AI, it is crucial to understand the limitations of its application to lower the
associated risks. Since the role of an expert witness remains based on independence, and professional accountability, it is crucial to carefully control the use of
AI to assist experts in data analysis and report drafting, thereby preventing potential challenges to the
credibility or admissibility of the expert’s opinion.
Rapid AI development necessitates the introduction
EXPERT WITNESS JOURNAL
5. Lack of Industry-Specific Training
The majority of commercially available AI systems are
not designed for construction-specific scenarios and
are instead trained on large datasets. For instance,
ChatGPT may misunderstand an expert's request to
write a narrative on ‘time-related costs’ as referring to
general inflation rather than project prolongation
costs as specified in the relevant contract. Similarly,
GPT-powered Excel products may incorrectly group
59
AUGUST/SEPT 2025