AI x QS = Opportunities

In my first article (The Rise of Machine Learning), I shared research into large language models (LLMs) which included what they are, how they work, their limitations and their risks. I have reflected on this research, and, in this second article, I will share how I consider LLM’s, and other forms of generative AI might help quantity surveyors improve their productivity and performance.

Technical / Regulatory Compliance

I remember watching the entirely avoidable Grenfell disaster unfold on the news and being horrified at the thought of people being trapped inside an inferno with no reasonable means of escape. I can’t help but feel that, as an industry, we collectively let down our ultimate client; the people who use, and rely on, the buildings and infrastructure we construct daily.

Dame Judith Hackitt, in her 2018 report, was, quite rightly, scathing in her assessment of the cultural issues in the construction industry which she described as “a race to the bottom” and having “insufficient focus on creating the best quality building possible”. The recommendations proposed by Dame Hackitt to solve these issues included: building as a system; risk-based approach; and transparency of information and audit trail.

The achievement of these outcomes becomes infinitely more achievable using generative AI. How can we be certain that technical / regulatory compliance has been achieved or that, as QS’s we have procured this compliance, given the volume of documentation a project produces? It is simply too much for a human to consume the vast number of technical standards, employer’s requirements, contractor’s proposals, BIM models, construction design and as-built information etc and resolve all compliance issues.

There has to be a better way, and maybe generative AI can help us find it.

Tendering

As QS’s we have a role to play in responding to Dame Hackitt’s recommendations. How often do you witness, or are you involved in, a tender where a huge document file transfer is issued with a request for a lump sum price in an unreasonably constrained timescale? The tendering contractor (or subcontractor) simply cannot, in the permitted tendering period, make its way through the full extent of the documents to prepare a robust programme and price which is based on the entire scope and risk profile. Consequently, whether we like it or not, we default to lowest price decision making because we cannot truly determine the value of each tender.

The use of generative AI (e.g. possibly Microsoft Copilot) could be used to ask questions of the data, which might include “What is the contractual risk profile?”, “What is the extent of our design responsibility?”, or “What is the extent of the scope?”. The initial benefit would be the reduction in time required to appraise the documents at the outset of the tendering process. In the evaluation process, we might spend more time focused on the value of the tender because the tender submissions are more consistent, and the pricing is more robust. This could lead us to a move away from lowest price tender selections.

Productivity Measurement

We are the UK’s worst performing industry when it comes to productivity. In my experience, this is hardly surprising given that the measurement of productivity in the construction industry is inconsistent at best, and non-existent at worst.

I struggle to understand the reason for this because it seems to be we have all of the jigsaw pieces to generate productivity calculations but, for some reason, we struggle to make the connections between various data sets and systems to generate robust productivity measures. It might be that we struggle to handle large data sets and make the connections between data which is often polluted or inconsistently structured.

The use of generative AI might yield insights which help us unlock the industry’s productivity challenge. This might be measuring the performance of operatives (something we just don’t seem to be able to consistently get right) or measuring the productivity of design / management staff (something we seem to largely ignore).

Project Financial Reporting

A subject very close to my professional heart, having grown up on the staple contractor QS diet of producing project financial reports for monthly business reviews (CVR’s, contract review reports, project review reports etc). However, I continue to be surprised (possibly even alarmed) at how very little has changed during my time in the industry when it comes to the preparation and frequency of these reports.

Why are we still doing monthly reports when the data and systems exist to, in theory, produce this information every day, and why are commercial teams taking days, sometimes weeks, to prepare this information? Despite the extent of this preparation, the output can sometimes fail to answer the question posed or, worse still, give a false impression of the project’s financial health. I see the frustrations mainly amongst business leaders who have prepared business plans, or measured against business plans, only to find out that the project financial information was unreliable.

It seems to me that there is far too much data handling and processing, and we have become comfortable doing this rather than using technology advancements to challenge and change the way we work. As QS's we need to embrace data science methods and encourage those joining the industry to develop these skills.

The ease of the generative AI interface could help unlock these skills challenges. The data handling and processing could be reduced which would create time to focus on solving the profitability or cash flow issues, and improving project performance. This information would assist business forecasting and allow timely interventions to be made where a project might move away from the business plan expectations.

Records, Records, Records

Lastly, the old contractor QS adage: “records, records, records”. I recall attending numerous training sessions where this message was drilled into me and, in case it escaped your mind just for a moment, it was drilled in a few more times for good measure.

It seems that the present and future state for records will change with the increased use of generative AI. The ease at which generative AI can produce life-like pictures and videos is exciting, but I see this could present problems for the quantity surveyor who evaluates substantiation presented by a subcontractor.

How do you prove that a photo you receive, say a progress photo on a given date, is what it purports to be? We may not be able to take the jpeg photo at face value. We will need to see, possibly even interrogate, the metadata to verify and validate the record as contemporaneous. This will mean developing our skills to ensure our investigations are alert to these possibilities.

Final Words

The improvement opportunities for quantity surveyors are limitless. I hope this sample of possibilities sparks some thoughts and intrigue about the opportunities that might be possible for quantity surveyors if we embrace generative AI.

It is apparent from my two articles that those in the profession, or entering the profession, will benefit from enhancing and developing their skills to make the most of the opportunity available to us.

These issues will be explored in future articles as Quantik looks at various tools available to the quantity surveyor, and how these might be used to improve the productivity and performance of the profession.

Back to insights