AI and the Delay Expert: Why Defensible Practice Matters

Artificial intelligence is rapidly becoming part of professional working practice. For expert witnesses, however, efficiency is only part of the equation.

Expert evidence must remain independent, transparent and capable of being explained under scrutiny. An AI-generated output may look convincing, but that does not necessarily mean that it is accurate, reliable or appropriate for use in expert analysis.

In January 2026, The Academy of Experts published its Guidance for Expert Witnesses on the use of Artificial Intelligence. It provides a timely and welcome framework for experts considering whether, when and how AI may be used in their work.

At the heart of the guidance is a straightforward professional principle:

“AI is not a substitute for the expert’s opinion.”

— The Academy of Experts, January 2026, Section A, paragraph 2.1

The expert remains ultimately responsible for their work product. That responsibility cannot be transferred to a software platform, an algorithm or an AI-generated response.

The question is not simply whether AI may be used

AI can offer genuine benefits to expert witnesses. It may assist with document organisation, information retrieval, summarisation, quality control, data review and communication.

The more important question is how that assistance is controlled.

An expert must be able to understand what the tool has done, verify its output and explain whether it had any bearing on the analysis or opinion. That becomes increasingly important where AI is used for more than administrative support and begins to interact with substantive evidence, analytical processes or expert reasoning.

The Academy’s guidance therefore distinguishes between lower-risk uses and activities carrying significantly greater risk. Administrative tasks and basic editorial review may be relatively low risk. By contrast, using AI to undertake material analysis, recreate counterfactual scenarios or contribute to an expert’s substantive opinion requires much greater scrutiny.

Why delay analysis requires particular care

Forensic delay analysis is a data-heavy and judgement-intensive discipline.

A delay expert may need to examine thousands of programme activities, logic links, constraints, calendars, progress updates and critical path movements. That information must then be considered alongside contemporaneous records, correspondence, technical evidence and the factual circumstances of the project.

AI may be able to process or organise parts of that information quickly. But speed does not remove the need for professional judgement.

An apparent critical path identified by an automated process still requires expert interpretation. Extracted programme data must still be checked against the native files. A summary of project records must still be tested against the source documents. Any conclusion about causation must remain the conclusion of the named expert.

There are additional difficulties where AI is used in scenario modelling.

Certain forms of delay analysis may involve counterfactual exercises, including “but-for” or collapsed as-built modelling. These methods depend heavily on assumptions, programme logic and the reliability of the underlying information. If AI is involved, hidden assumptions or unsupported changes could materially affect the outcome.

The Academy appropriately treats AI-assisted scenario modelling as high risk. The practical question for delay experts is therefore what safeguards are needed before any such output could properly inform expert evidence.

AI for Delay Experts: Turning Guidance into Defensible Practice

The Academy’s guidance is intended to apply across many different expert disciplines. It therefore establishes a broad professional framework rather than a technical procedure for forensic delay analysis.

My new paper, AI and the Delay Expert: Turning Professional Guidance into Defensible Practice, considers how that framework can be translated into practical controls for delay experts.

The paper examines:

• adequate human oversight in data-heavy delay analysis;

• validation of AI-assisted programme and critical path work;

• controls around counterfactual and scenario modelling;

• management of AI use by assistants and expert teams;

• matter-specific AI-use plans and registers; and

• preparation for disclosure and cross-examination.

A central recommendation is that experts should maintain a clear audit trail of significant AI use. This may include recording the tool used, its purpose, the source material considered, the prompts or parameters applied, the output generated and the verification undertaken before anything was relied upon.

That record is not merely administrative. It helps the expert demonstrate that AI use was controlled, proportionate and subject to independent professional review.

Assistance must not become substitution

The responsible response to AI is neither to reject it automatically nor to use it casually.

AI-assisted tools may become increasingly useful in complex disputes, particularly where experts are dealing with large volumes of project records and programme data. But usefulness does not remove accountability.

The expert must still be able to explain the methodology, identify the assumptions, test the outputs, validate the results and demonstrate that the final opinion represents their own independent professional judgement.

That is the distinction between AI-assisted work and the outsourcing of expert reasoning.

AI may assist the delay expert, but it must not replace professional judgement, evidential reasoning or expert independence.

Read the full paper

The full paper develops these issues in greater detail and includes practical safeguards, recommendations and a delay expert’s AI-use checklist.

AI and the Delay Expert: Turning Professional Guidance into Defensible Practice

VIEW THE PAPER