What the Construction Giants are saying about Tech | Q4 2025 Earnings Report Review

We read all the earnings reports so you don’t have to.

If you are in London next week, I’ll be joining the Constructech Meetup hosted by Seray Wright on Thursday the 26th of March at 6pm - 8pm. There’ll be pizza and drinks, would love to see you there! Here’s the link to register.

---

Every quarter, the largest publicly traded construction and contech companies release quarterly earnings reports and host conference calls, providing insights into their business and an outlook on the future of construction.

At Last Week in ConTech, we review these through the lens of technology and innovation, attempting to understand how the ‘giants’ are preparing for and shaping the future of our industry.

This quarter, we reviewed the conference calls of 18 companies across the project lifecycle and technology stack from EPC, to GC’s, Trades to Technology providers. From this, we have provided excerpts of the following 7 companies which had notable discussions on AI and technology adoption:

• AECOM (EPC)

• WSP (EPC)

• Jacobs (EPC)

• Fluor (GC)

• Autodesk (Tech)

• Procore (Tech)

• Trimble (Tech)

From these calls two key themes emerged:

1. EPC firms are actively trialling, building partnerships with and deploying AI software at an accelerated response due to AECOM’s acquisition of Consigli, outlining their strategy for future growth.

2. Large ConTech companies are actively deploying AI solutions but finding it challenging to correctly price them, balancing traditional pricing with consumption based systems.

In this article we’ll be breaking down each conference call, highlighting the key quotes and providing further commentary where relevant. Of particular note were the WSP and Autodesk calls where they clearly laid out their AI strategies.

Let’s dive in.

---

AECOM

EPC Firm, Market Cap: $11.7 billion USD

AECOM’s presentation began by highlighting the outlook for the industry noting:

Rapid urbanization has pushed 45% of the world’s population into cities today, with projections showing dramatic further growth. Even in advanced economies, aging and inadequate systems are under severe strain. The U.S. alone faces a $3.7 trillion investment gap over the next decade.

They touched on their acquisition of Consigli stating:

We’ve completed the integration of our September acquisition. We have already doubled the size of our team…The technology is now live on our projects, and the initial performance results achieved have matched our expectations.

The call continued highlighting challenges they now face around pricing and structuring their contracts:

Our clients are recognizing that we might not have a way of contracting that makes sense. And so they’re bringing up in the dialogue how they would like to move to a method of contracting that recognizes that value. So, you know, moving away from something like cost plus to something that looks more like a fixed fee.

As discussed previously, it’s challenging to see how AECOM navigates this transition.

If a project was priced at 500 hours at a rate of $1000 per hour ($500,000 project value), the cost of delivery has reduced as AI reduces the number of hours. It’s unlikely clients will allow them to continue to ‘price to market’ as their CEO mentioned in the last earnings call.

Instead they’d either ask for:

1. A lower rate as the cost of delivery is reduced

2. Increased scope to reflect that more can be done in the same hours

The question I have is what is the long term strategic vision AECOM has with the acquisition?

As the first mover, they recognized that AI is structurally changing the project delivery model from cost plus to fixed fee. But how will they use their proprietary technology to create a competitive moat?

My hypothesis is that they should use their technology capability to create market dominance in specific complex market verticals such as nuclear power plant, bridge or stadium design where their expertise will compound.

For example, in the US, AECOM is one of the leading stadium designers, having worked on more than 90 stadiums and arenas. Given the complexity and cost of these projects, clients tend to go to firms with demonstrated experience. It means a small number of firms have:

1. Domain and technical expertise in delivery

2. Access to proprietary data

Both of these are required to build AI models for this segment. So if AECOM is able to deploy their technology quickly enough in the complex megaproject segment, they’ll be able to outbid competitors due to improved margin. As they win future work, they own the data and expertise to continue to improve and validate their models.

Over time this becomes an unassailable moat that a new entrant can’t replicate due to the lack of data and expertise availability.

WSP

EPC Firm, Market Cap: $21.7b USD

WSP started the call directly addressing the AECOM acquisition by stating:

In recent months, many actors have painted all professional services firm with the same AI brush, worrying that we are entering an era where advanced AI will replace firms like WSP.

They contrasted how WSP’s business model is different to other consultancies and industries which face disruption.

Our 83,000 experts design and manage complex physical projects, including bridges, transit systems, water treatment plants, energy facilities, environmental remediation, and so much more…

…It represents service work that blends advanced domain expertise, inherent know-how, technical analysis, field execution, and professional judgment, along with massive amounts of proprietary data, knowledge, and experience that is not publicly available.

This means that AI models can’t be trained to replace their expertise as their domain expertise isn’t publicly accessible in the way code is on GitHub repositories. They went on to say:

Large language model today are far better at automating digital, virtual, and repetitive tasks such as coding, data entry, and document generation, that are performing deterministic tasks and providing guaranteed correctness in the physical world. In the engineering world, that simply cannot occur.

To WSP, their competitive moat against AI is on based on:

1. Expertise in physical sciences and engineering

2. Decades of proprietary intellectual property and design standards and best practices

3. Their professional accreditation, legal standards, and obligations which establish a barrier to entry.

They then highlighted a number of key factors which make it hard for AI to disrupt their model:

1. Every project is unique and requires understanding of local physical variables and needs to incorporate local community and stakeholder input.

2. This means that infrastructure environmental projects can’t have a one-size-fits-all solution:

   We can’t simply feed a prompt into a computer and get a turnkey design…The answers depends on the specific terrain, traffic patterns, stakeholder inputs, regulatory reviews, river flows…
   
   …Our experts combine technical data, sometimes with AI assistance, with on-the-ground observation and stakeholder dialogue to get it right. This is why we say AI augments our capabilities, but is not a decision-maker in our field.

3. Accountability is also non-negotiable in the industry as their projects involve public safety, building codes and environmental laws that differ from jurisdiction.

   Licensed engineers are those who can certify that a design meets all those specific codes and can be built safely…AI-generated outputs are always subject to rigorous human oversights, thorough quality control, and professional accountability.

They went on to say:

We fully expect that some tasks in design and consulting will be automated..To the question, can AI design an asset on its own? The answer is no. AI can help with preliminary sizing and drafting, parametric optimization, cut code lookups, scenario generations, but it cannot guarantee compliance, verify safety, carry legal liability, produce deterministic proofs, ensure physical correctness, explain every step with guaranteed traceability, sign, seal the drawings, engage with the physical world and all stakeholders.

The key takeaway from WSP is that AI is coming but it won’t replace engineers, rather it will be embedded in workflows to augment them. So how are they doing this?

WSP shared two examples:

1. Microsoft Partnership
   They’ve partnered with Microsoft to enable one of the largest agentic and AI deployments to make sure their frontline knows how to responsibly use AI.

2. Ecosystem Strategy
   They’ve partnered with firms like UrbanLogiq, Fathom, Google and small startups. These firms come to them as they want domain expertise at scale as AI models compound when combined with their knowledge across projects, geographies and the data volume they uniquely have access to.

They are also incredibly specific about protecting their IP stating:

we’ll work with the ones that are willing to work with us on protecting our domain expertise while driving value to our clients. In other cases, we’re building internally our own proprietary models that will remain within our parameters, so we can retain that IP and the value we bring to our clients.

This strategy works for them as more than 60% of their work is fixed price.

Clients, more and more, are not looking for price…80% of our qualification criteria are qualification-based, are not price-based.

And they made this interesting point on AI:

The AI models are getting actually commoditized…In order to create value, these models need that domain expertise.

[The] technology player, coming in the power sector with no domain expertise…they’re missing 75% of the solution.

Jacobs

EPC Firm, Market Cap: $15b USD

Jacobs has had a differing strategy with respect to AI.

They discussed where they are using AI today:

[For projects] the schedules and the delivery model for these can’t be done without the use of the AI platforms. And when I say AI, machine learning, the automation of tasks that we put into play…

…in the field, we’re using some strong predictive analytics. It’s a platform called Acuity in order to really get out in front of field-level issues that are coming up in real time. And that’s been a real game changer for us. We’ve got Acuity deployed across all of our end markets in the field program management work.

we use Replica as our digital twinning…in the water sector..in the manufacturing sector and the data center sector [and] is allowing for us to get to the data insights and the simulation technologies.

Fluor

Contractor, Market Cap: $6.53b USD

Fluor describe themselves as an early adopter of AI:

We began our AI journey in 2018 by developing a predictive analytics platform built on data from more than 200 of our largest EPC projects. This foundational work allows us to benchmark schedule, planning, and cost performance using proven historical outcomes, so projects are planned with greater accuracy and discipline from the start.

They have now deployed AI across the project lifecycle from:

predicted analytics on capital projects to intelligent pricing insights across the supply chain.

It’s also been implemented across “individual functional roles, including HR, finance, legal, and procurement.”

Autodesk

Tech, Market Cap: $52.4b USD

Autodesk discussed their overall platform strategy by explaining their customers’ demand for convergence.

They view convergence as the next stage of digital transformation, where previously separate technologies, processes, and data silos merge to create new, agile ways of designing and making.

Convergence reduces risk, increases quality, and optimizes costs and resource use during the design and build phase of an asset. It enhances efficiency, resilience, and reuse during operations

Transforming in this way prepares them for “an agentic AI world” and they discussed what is needed to build this future:

1. Data Quality

   AI agents need large quantities of high fidelity, contextual, geometry-rich, two and three dimensional physical world data that represents physics and engineering-based principles so AI can learn how to drive, design and make decisions (which Autodesk uniquely has access to).

2. Context

   AI agents need context as when making inferences, agentic AI has to operate inside a live project where the correct answer depends on the design intent, current model state, regulations and standards, constraints, dependencies, permissions, and approvals. Decisions must be compliant, coordinated, traceable, and reversible.

3. Expertise

   To build AI agents, teams require specialized AI expertise to generate unique and valuable intellectual property. Autodesk has been building this talent pool for a decade.

The key takeaway was:

Data and context fuel the knowledge graph, which is foundational to any artificial intelligence. Data scarcity and context complexity make the knowledge graph hard to replicate for our industries.

And they said this about general AI development:

it’s not our goal to compete with the core capabilities of what the frontier models are good at. What our goal is to ensure that the combination of what the frontier models do, what an LLM does, and what our proprietary foundation models do, is always better than what a frontier model can do alone.

They also discussed their $200m investment into World Labs:

World models are important for physical AI because of their ability to spatially reason about physics, and also respond to real-world changes because of their awareness of what’s going on in 3D. We see this as a fun, foundational, kind of horizontal technology that’ll power lots of solutions.… It’ll go deeper into initial architecture design. It’ll end up going into areas associated with digital twins, with factory automation, robotics, all of these things associated with that.

One concern analysts note is that AI may reduce jobs and ultimately revenue, however Autodesk refutes this point:

[In our industry] we have a fundamental capacity problem…we absolutely want fewer people per project because we want our customers executing more projects. There’s plenty of demand for projects out there

AI then is beneficial as:

Speeding up modeling activity and things like that’s kind of improving the core value of a seat of software. We don’t expect seats to go away anytime soon. There will be a solid core of seats, but the task-based automation is going to add to the value of that seat. That seat is going to get more valuable as we enable one person to execute on more aspects of a project.

Again, it’s fewer people per project, more projects executed at the task-based level and at the seat-based level.

This works for them as they “already deliver project-based pricing around construction.” Now they are going to:

“monetize more of that project activity through consumption. As we reduce the number of people are working per project, we’ll monetize other aspects of the cross-pollination of the project.”

Procore

Tech, Market Cap: $8.8b USD

Procore has been deploying ‘Procore AI’ and they believe they will lead the AI era due to four main reasons:

1. Data
   They have 3m active users with a large proprietary dynamic data set.

2. Trust
   Their scalable, enterprise grade infrastructure ensures that AI actions are secure, compliant and contextually relevant.

3. Network Effect
   Procore is the central hub and system of collaboration creating a flywheel network effect.

4. Procore Agentic Solutions complete critical tasks
   Their AI isn’t just insights, it will work as an orchestration layer and a digital coworker.

With regards to pricing they stated:

We price on project scale rather than seat count, [so] our traditional revenue remains insulated from headcount fluctuations as AI drives industry efficiency.

They’ll also:

“be including some component-based, some consumption-based components.”

Trimble

Tech, Market Cap: $15.3b USD

Trimble discussed how they are looking to monetize AI:

We also monetize through the tier…And so we put AI capabilities in those best of the tiers…there’s a hybrid of where it’s both recurring and consumption.

Of course, agentic AI, generative AI does have a variable cost associated with it. It’s not for free. So the unit economics are, of course, different in an AI-forward world.,

Trimble has experience with consumption based pricing mentioning that 60% of revenue through transportation is consumption.

---

If you’d like a firm included in these summaries, please reply and let me know!

The material presented on Last Week in ConTech’s website are opinions only and are provided for informational purposes and should not be construed as investment advice. It is not a recommendation of, or an offer to sell or solicitation of an offer to buy, any particular security, strategy, or investment product. Any analysis or discussion of investments, sectors or the market generally are based on current information, including from public sources, that I consider reliable, but I do not represent that any research or the information provided is accurate or complete, and it should not be relied on as such. My views and opinions expressed in any website content are current at the time of publication and are subject to change. Past performance is not indicative of future results. Please consult an investment specialist before investing.