A key step in the construction process is when the architect and engineering design steps are complete, and contractors and suppliers are invited to submit bids. Referred to as the takeoff process, contractors use design documentation to extract materials and quantities to be procured, critical in the overall cost and pricing analysis. In general, the documentation stack addresses all aspects of the construction project - material, structural, landscaping, roadways, plumbing, water & sewage, HVAC, etc., and is typically shared in a PDF format running into 100s of pages. Estimation experts at different contractors need to go through this in detail to extract BOMs (Bill of Materials) and quantities, as well as identify critical path items. Based on experience, they select suppliers, request quotes, and use all this information along with labor estimates to understand project costs and schedules. These, along with competitive intelligence and profitability metrics, are used to provide a price quote or a bid.

Civil engineering and construction has traditionally relied on manual expertise from design, to bid, to execution. This is surprising given that construction is a $30T/year business, 10% of which is in North America. By comparison, the global semiconductor business is ~ $1T and automotive business at ~$3.5T. Both these industries rely on significant amounts of AI and autonomy in their operations.

The use of manual techniques has several drawbacks - lack of an experienced workforce (~200K estimators exist, 50% approaching retirement), accuracy issues (in the documentation or in the analysis), and the time it takes to complete. As a result, many contractors pass on business, without even bidding - they don’t have the people, bandwidth or the time. AI (Artificial Intelligence) tools are slowly making headway in such domains. A recent article discussed the use of such agents and tools to select and manage the supply chain in complex construction projects like urban hospitals. This article continues the thread, analyzing what happens earlier in the process as contractors evaluate designs, and provide pricing bids to win the business.

Attentive.ai - Making AI the Operating Backbone of Preconstruction

Having secured a recent Series B $30M raise (prior Series A raise was $12M) the company’s flagship product, Beam.AI, “ automates takeoffs for mechanical , concrete , steel , civil , utilities , roofing , and more. Contractor & suppliers cite time savings approaching 90% and a meaningful increase in how many bids they can submit, often close to 2X in a quarter.” The company serves customers like General Contractors (GCs), subcontractors, and suppliers in high-growth segments like mechanical, concrete, civil, utilities, and more.

Shiva Dhawan is the co-founder and CEO of Attentive.ai. A mechanical engineering graduate from the premier engineering institution in India (IIT, the Indian Institute of Technology), the original focus of the company was in insurance and mapping. It pivoted to the construction sector in 2021, when opportunities to address the takeoff process in landscaping and construction presented themselves.

The thesis is to train their AI agents using foundational AI models along with historical data from their client base of ~1,200. This needs significant human curation, all done with a staff of ~600 people, a majority located in India, with the remainder in California. This staff also provides a final quality check before presentation of the output to the client.

There are strict controls on the proprietary raw data shared by Attentive’s clients - it is used strictly for training the AI agents, and never shared. When competing contractors use Beam AI for a specific project, they all get the same BOM and quantities. The strategy is to buy the material at lower cost, and use their individual labor estimates and profit targets to price the job competitively. Attentive currently has revenues of $13M/year (average of $11K/client/year) through licensing contracts. According to Mr. Dhawan, “Construction has never had a demand problem. The real constraint has been estimating capacity. Contractors can only win the work they have time to bid, and that is where AI can create long-term business impact. Our Beam AI product helps teams move from manual takeoffs that take days to hours, while keeping accuracy and human review at the center of the workflow”.

Steel West and National Whole Sale Supply (NWS) are 2 of Attentive’s customers.

Steel West - CONSISTENT QUALITY, TIMELY DELIVERY, SOLID VALUE

Based in Idaho, the company was founded in 1970, and currently led by the third generation family owner and President, Colter Sears. The company fabricates site-erectable structural steel assemblies for residential and commercial construction. The assemblies are transported by a logistics vendor who delivers it to the site, typically in the mid-west, in a 500 mile radius around Idaho. After the takeoff process is completed, raw steel (multiple specifications) is procured, machined and assembled in their 15,000 sq. ft. workshop (Figure 1). The equipment includes plasma and mechanical drills (computer controlled), bandsaws, presses and welding machines.

The legacy takeoff process was highly manual and typically took 24 hours for a 400 ton steel proposal. With only 2 estimators on board (Mr. Sears is one), this restricted them to bid only on ~4 projects/week. An average of ~10 proposals/day are not bid on - partly because of the takeoff bottleneck, and also because of capacity constraints in the workshop. Project size ranges from $10K to $2M - this is another criteria used to decide which projects to bid on, the idea being to have at least 2 large projects on hand at all times.

According to Mr. Sears, using Beam AI has increased their bids by 50% - from 4 to 6/week, at which point they are completely capacity constrained. Using Beam AI also frees up time for other tasks like project management. As workshop capacity expands, they will be able to use Beam AI to execute even more takeoffs. Figure 2 shows how Beam AI is quickly able to absorb complex bid documents and translate it into a clear BOM in a matter of hours.

Another advantage of Beam AI tools is it is very accurate and dependable. In some cases, the bid documents have errors and inconsistencies - these are highlighted and flagged by the software for correction, before proceeding. According to Colton Sears, “Utilizing Beam AI for material takeoffs has increased our monthly bid volume by 35-50% while simultaneously, allowing us to focus more of our time on bid refinement and project controls.”

National Whole Sale Supply (NWS) - We Have the Material You Need

Founded in 2002, NWS is a Dallas-based supplier of ~50,000 SKUs or parts needed for construction - plumbing, electrical, HVAC, etc. - kind of like an Amazon for building supplies. With locations in Texas, Lousiana, Oklahoma and Arkansas, the company serves clients (typically contractors or sub-contractors) through 4 divisions, one of which is the Waterworks Division based in Dallas. It is responsible for all parts related to water and sewage transmission (up to the construction site, like hydrants, pipes, valves, etc.). Corey Reynolds is the Director, and has worked at NDS for the past 17 years, the last 5 being in the Waterworks division.

According to Mr. Reynolds, prior to implementing Beam AI tools 3 years ago, the takeoff process was highly manual, with design documentation on the water and sewage aspect alone running into 100s of pages. People had to be internally trained to do this, which is not trivial due to career interests and turnover issues. Beam accelerates the process dramatically, and extracts the correct SKUs from the drawings in a matter of hours. This helps when the bid processes are highly time constrained and last minute - NWS with Beam is able to participate in emergency bids. Beam also improves the accuracy of the process and is able to pick up errors in drawings, which is highly impactful during construction. Using Beam AI has increased the Waterworks Division bids by 50%. Typical bids range from $25K for a gas station, $3M for a residential neighborhood, and $5M for commercial high-rise construction projects. Figure 4 shows an example of how BeamAI’s tools accelerates the takeoff process for NWS.

Mr. Reynolds, “The real value of BeamAI for us isn’t replacing people—it’s helping our team make faster, more confident decisions earlier in the process. In our world, the companies that can move faster and make better decisions early are the ones that win—BeamAI is becoming a key driver of that advantage".

Stack - Accelerated with AI. Driven by You

Stack bills itself as a company that helps general and specialty contractors “from project evaluation to completion, through use of cloud-based software to help run their business and maximize their profits. Our preconstruction solutions enable fast and accurate takeoff and estimation while our construction solutions power real-time field and project collaboration”.

Based in Cincinnati, Ohio and founded in 2011, it recently appointed Viyas Sundaram as CEO. With prior experiences in the automated workflow world, he has been brought on to lead STACK’s continued expansion and innovation, through use of AI and other digital tools. Stack creates specialized LLM (Large Language Models) tools for construction based on training using data from its clients on how to interpret drawings to extraction of BOMs and quantities for accelerating the takeoff process. It does this through manual annotations of millions of blueprints provided by its customers, and training AI systems using data science principles. The goal is to enable its customers to create the optimal bid, quickly and accurately. The company currently has 7,000 customers and 120 employees, is owned by a private equity company, and is profitable.

According to Mr. Sundaram, the expectations for digital AI agents in terms of accuracy and precision is approaching similar levels to physical AI - the former is driven by business outcomes, whereas the latter requires accuracy and judgement to ensure human and capital safety. Typical margins for contractors and suppliers in construction are in the 15-20% range. AI precision of 70% can reduce these margins by 50% or more, or worse, lead to losses. So focusing on getting precision and accuracy similar to physical AI (typically > 99%) is critical. The way to do this is developing AI models based on large and diverse data sets, and curated and tested by human experts. Figure 5 shows how Stack’s AI tools help its customers win more business - profitably.

Stack cites various cases studies in its blog articles regarding the ROI (Return of Investment) gains of its customers using Stack’s cloud-based tools. Specifically:

  1. 600% year–over–year revenue and 30% increase in win rates
  2. Takeoff times reduced by 40%
  3. Eliminating manual errors and rework, driving a 30% increase in win rate
  4. Creating leaner preconstruction workflows

According to Mr. Sundaram, “The reality is that time and revenue losses begin before the shovel even hits the ground. Manual methods or outdated, desktop software to keep track of important documents, generate estimates, and takeoffs isn’t functional. With physical AI, there is no room for error, as big machines move at high speed, and serious capital is at stake. Stack aims to bring digital tools to the same level of precision as physical AI.”

Construction, mining and agricultural machines have pioneered physical AI and the AoT® (Autonomy of Things) for decades. long before autonomous cars became a thing. Construction management and more importantly, pre-construction, are finally embracing the AI age, through the use of digital AI tools and lean methodologies to eliminate waste, rework and delays. These tools leverage vast amounts of mostly boring contractor data curated by humans for training. As opposed to physical AI where the data is jealously guarded, the construction industry does not seem to regard its data as competitive - primarily because it may be publicly available, and there are a large amount of smaller contractors who don’t consider it as a competitive tool. It remains to be seen if that will change.

AoT® (Autonomy of Things)  is a Registered Trademark of Patience Consulting LLC. Visit aot.llc for more details.