From Gut Feel to AI: Tolis Chatzisymeon on the Future of Cost Estimating

In the rapidly evolving landscape of construction technology, few areas have remained as resistant to change as cost estimating. Yet even this traditionally conservative domain is now experiencing a profound transformation through artificial intelligence.

We recently had the pleasure of hosting Tolis Chatzisymeon, founder and CEO of Nometech, on the Project Flux podcast. His insights into the intersection of AI and estimating offer a fascinating glimpse into how technology is reshaping one of construction's most critical functions.

Tolis's Journey: From Bridges to Business

Tolis's path to becoming a construction tech entrepreneur began with a passion for structural engineering. "I was always fascinated by bridges," he shared. This fascination led him to pursue a Master's degree in Construction Engineering and Product Management in the United States, where he gained exposure to both the technical and management aspects of construction.

The pivotal moment came in 2004 when Tolis noticed a fundamental disconnect in the industry. "I was working on a project where we had the planning team and the estimating team," he recalled. "The planning team was using Primavera P3 at the time, and the estimating team was using some Excel spreadsheets. And there was no connection between those two."

This gap wasn't merely a technical inconvenience—it represented a significant barrier to project success. "The planning team was planning the project in a certain way, and the estimating team was estimating the project in a completely different way," Tolis explained.

The result was a misalignment that often led to cost overruns and schedule delays.

Recognising this problem as an opportunity, Tolis founded Nometech to bridge the divide between planning and estimating. "We built a system that could take the planning data and match it with the estimating data," he said. This seemingly simple concept addressed a profound challenge in the industry and set the stage for Nometech's growth into a leading provider of estimating solutions.

The Evolution of Estimating: From Secrecy to Data

The traditional image of an estimator—the older professional sitting in a corner, sucking between their teeth and getting quite grumpy—isn't entirely without basis. As Tolis confirmed, "From my experience, we still see this happening, right? But it's not as bad as when I started Nometech."

In those early days, estimators often jealously guarded their domain. "I had to face all these old guys all the time," Tolis remembered. "And they were very overly protecting their kingdom. Nobody knew what they were doing." The secrecy was so extreme that some estimating systems even included features to quickly hide screens: "They used to have some game that you would play and switch screens instantly, so they were playing like card games and then switch flip it instantly. So nobody knew what these guys were doing."

This culture of secrecy is gradually giving way to a more open, data-driven approach. "Today you have modern generations coming in, so they understand data. They start to understand about BIM as well, so it's becoming a bit more open," Tolis observed. However, he noted that change is still incremental: "Still the old guy is taking the decision, he's signing off the estimate."

One of the most persistent challenges in estimating has been the reliance on "gut feel"—the subjective judgement that often overrides detailed analysis. As James Garner pointed out during the interview, it's not uncommon for someone to go to great lengths with detail and data analysis to produce an estimate, only to have a senior estimator say, "That doesn't feel right," and make arbitrary adjustments.

Nometech's products aim to address this issue by providing data-driven alternatives to gut feel. "Cosmodeler... basically gives you that gut feeling, but it's documented," Tolis explained. "Your historical estimates get in a repository and you can do cost per square metre analysis. You can compare across history." This approach doesn't eliminate professional judgement but rather grounds it in data and makes it transparent.

AI Applications in Modern Estimating

Nometech has embraced AI as a core component of its estimating platforms. "Within Omnidire, we do lots of R&D in AI," Tolis shared. "The latest release we have at Costos 8.5, our estimating platform, we use AI... to basically chain LLMs, large language models."

One practical application is in matching items from bills of quantities to a contractor's database. "When you upload the bill of quantities, the BUQ in the system... [it's] searching to your local database to find the best matches that you have within your own data and scores them," Tolis explained. This capability allows estimators to quickly identify relevant pricing information from their historical data, significantly accelerating the estimating process.

Computer vision is another area where Nometech is applying AI. "We have implemented the computer vision for automatic counting objects, let's say how many doors, how many floors," Tolis noted. The company is also working on taking off quantities directly from 2D drawings—a capability that remains important despite the promise of BIM. "We thought that with BIM that's not necessary, but still the industry is resisting," he acknowledged.

The accuracy of these AI applications depends heavily on data quality. "If your data is pretty accurate, then you can achieve pretty decent accuracy, even up to 10, plus 20, 30%," Tolis stated. However, he emphasised that this level of accuracy is only possible "provided you have recorded everything and you have the attributes."

When asked about the potential for AI to eventually automate the take-off process entirely, Tolis was cautiously optimistic. "I've seen it even with my guys, I've seen this happen. I've seen with my own eyes, uploading and drawing and measuring the areas, the walls, the stuff," he said. However, he noted that current solutions still require significant human intervention: "Still it's not accurate and still the amount of clicks that you have to do to correct it... is more than just doing it manually. But I think we'll reach a stage."

The Standardisation Challenge: ICMS and Industry Adoption

A significant barrier to effective AI implementation in estimating is the lack of standardised data. Tolis has been actively involved in addressing this challenge through his work on the International Construction Measurement Standards (ICMS).

"I was on the committee that was drafting the standard," he shared. "And my first role was to make sure that the standard is data driven. So somebody can report in that standard no matter how your estimate is structured." The goal was to create consistency in how construction costs are reported and classified, making it easier to compare projects and leverage historical data.

This represented a fundamental shift in thinking about measurement standards. "The challenge I faced when I entered there was in the QS's mind, they were thinking like I have to create a big booklet of pages where the more pages I produce, I'm doing a better job," Tolis recalled. "We had to take out that mentality. We had to think that, yeah, we're talking about data and how we can put the machine doing the reporting, the coding and the structure of the estimate in something meaningful."

Despite these efforts, adoption of ICMS remains a challenge. "Still there is a way to go," Tolis acknowledged. "We are trying now to work on a mega scale with mega scale customers, Middle East to comply with it. Still the owners are resisting." Part of the problem lies with consultants who prefer to maintain the status quo: "The consultants that help them don't guide them in that direction. Because it's more work for them doing it the way they're used to it than reporting ICMS."

The solution, in Tolis's view, is for project owners to drive adoption: "It has to be the owner that dictates ICMS and then the consultants will comply. Not the other way around." This parallels the adoption challenges faced by BIM, where client mandates have proven more effective than industry-led initiatives.

The Human Element in an AI World

Despite his enthusiasm for AI, Tolis maintains that human judgement will remain essential in estimating. "There's always a human in the middle," he emphasised. "There's always assumptions."

Part of this stems from the inherent complexity of translating real-world value into monetary terms. "Even the fundamental of estimating by itself, even money, you're transferring something to money, value that you're gonna give me to transfer to the project. Money itself is something that is changing," Tolis observed. Unlike physical constants, monetary value fluctuates based on numerous factors, including confidence, scarcity, and economic conditions.

\This complexity is compounded by external factors that can dramatically affect costs. "How do you model tariffs?" Tolis asked, citing a recent example: "I was talking with some customers in the US. The next day, they didn't know how much it's going to cost them the project they already started... Everything, the steel structures are going to change because they're importing from China."

When asked whether AI might eventually become better than humans at detecting signals of change and volatility, Tolis was sceptical. "I think when it comes to money, that's going to be always ambiguity. That's my feeling," he said. "AI can do a research, scan the web, get the sources it needs, still they do not know the deepest, like what's the decisions are made at the central banks and stuff."

Perhaps most importantly, humans provide accountability in a way that AI currently cannot. "The human puts an oven and takes responsibility, right? He can explain it. Whereas AI will give you the sources, but you can't blame AI," Tolis noted. This accountability remains crucial in an industry where cost estimates form the basis for major financial decisions.

The Future QS: Skills for an AI-Enhanced World

Looking ahead to the next five to ten years, Tolis sees AI reshaping the role of quantity surveyors, but not eliminating it. "AI will definitely improve the speed and the quality of the work that the QS is delivering. AI will help to make faster decisions, try something faster, more accurately maybe," he predicted.

This evolution will require QSs to develop new skills, particularly around understanding AI. "The skills that the QS needs to understand is now how AI made this recommendation, this assumption," Tolis explained. "So you need to understand a bit how you train a model, how you [use] some machine learning or LLM model. So to be able to understand why this is proposed by AI and what does it mean."

However, Tolis emphasised that fundamental knowledge remains essential. "When I estimate something... you need to run at least once the calculations, right, to understand what calculations are taking place. Not just the number you rate, it's more than the rate," he said. "The first principles are quite important."

This balance between traditional knowledge and new technological skills presents both a challenge and an opportunity for the profession. "I think there will be still a role for QS," Tolis affirmed, while noting that in some countries like the United States, the distinct QS role doesn't exist. Regardless of job titles, "somebody needs to take care of that and be responsible."

Advice for the Next Generation

For graduates entering the field, Tolis had clear advice: embrace BIM. "The first thing I recommend is BIM," he stated. "If you're a young QS, you can start understanding BIM, what BIM is... the data you put in a 3D model in the design because that's key value for career."

To help newcomers gain practical experience, Nometech offers a tool called CellBIM. "If you're a graduate you can download it. We provide licences to the graduates at cellbim.com so they can open a model, inside this precedent, start experimenting with the data, the quantities, the measurements within Excel," Tolis explained.

Beyond technical skills, Tolis highlighted the value of being able to bridge the worlds of construction and data science. "That combination is so difficult to find," he noted. "[Someone] that has data and has construction experience, you can either find a data geek that doesn't understand the global construction or the other way around. Somebody that has a good experience but doesn't get it with data sets."

These "translators" between technology and construction are increasingly valuable as the industry continues its digital transformation. As James Garner observed during the interview, "If you've got that role, you're a bit of a unicorn... because you've got very, very specific skills."

Conclusion: Bridging Technology and Expertise

The conversation with Tolis Chatzisymeon reveals a nuanced view of AI's role in construction estimating. Rather than replacing human expertise, AI serves as a powerful tool that can enhance accuracy, speed, and transparency. The most effective approach combines the pattern-recognition capabilities of AI with the contextual understanding and accountability of human professionals.

For the construction industry to fully realise the benefits of AI in estimating, several challenges must be addressed. Standardisation efforts like ICMS need broader adoption, particularly through client mandates. The quality and accessibility of data must improve to provide AI systems with reliable inputs. And professionals need to develop new skills whilst maintaining their fundamental understanding of construction processes.

As Tolis's journey demonstrates, the most valuable innovations often come from recognising and addressing the gaps between different aspects of construction management. By bridging planning and estimating, traditional expertise and modern technology, Nometech exemplifies how the industry can evolve without losing sight of its foundations.

The future of estimating will not be defined by AI alone, but by how effectively we integrate these powerful tools into our existing processes and expertise. As Tolis put it, "Estimating is going to be there always... people will always be interested to get the value of something, because they're going to make decisions based on it." The question is not whether estimating will continue to matter, but how it will transform to meet the challenges of an increasingly complex and data-rich construction landscape.

Want to stay updated on the latest in construction technology and AI? Subscribe to the Project Flux newsletter for weekly insights, tools, and expert perspectives. You can also connect with Tolis Chatzisymeon on LinkedIn and explore Nometech's estimating solutions.