Humans will keep control as AI assistants tailored to the construction industry deliver value on many fronts, and soon.
By Johnny Clemmons
There may be an industry poised to gain more from AI-driven automation than the construction business, but I sure can’t think of one.
The sorts of problems that slow or stymie construction projects today happen to be what AI is best at tackling. It excels at corralling information regarding all manner of operational decisions and actions from all corners of a business and synthesizing it so that people — or, increasingly, AI itself — can make good decisions fast.
The vision of autonomous construction is not to remove people from the equation. Humans will stay in control. But ERP-based AI assistants made up of multiple, task-focused AI agents tailored to the construction business are now in rapid development. The first commercial products are slated for early 2027, with many more to follow.
AI assistants will boost construction-business bottom line
Construction businesses large and small should be paying attention. SAP estimates that AI-automating just three common construction workflows can deliver a roughly 2% boost to the overall business’s bottom line. That can add up to millions of dollars a year. Here’s what those three use cases would do.
- Fix fragmented field-to-finance data that typically suffers from masked cost deviations and schedule slippage from lagging financial reports. AI accessing real-time data can surface potential problems before they happen.
- Break through procurement and contracting silos caused by vendors, subcontractors, and materials being managed outside the core system. That hides cost escalation until margins take a hit. Autonomous procurement workflows that access real-time data from core and specialized systems flag vendor risk and potential maverick spend before they impact a project.
- Address administrative drag on project delivery because billing, payroll, and progress reporting happen outside the core system, pulling project managers offsite and slowing cash conversion. AI assistant-driven billing, payroll, and reporting lower the burden and boost cash flow.
Rapid development of construction AI assistants is happening now
Those are just a couple of the many AI assistants in development or envisioned by a single ERP provider working with construction firms on rapid-development pilots (60-90 days from concept to deployment). The vision is to develop AI assistants spanning the construction value chain, and it’s happening quickly. Some of the other assistants in advanced conceptual stages or already being developed include:
- Bid optimization assistants to increase program win rates and improve project profitability.
- Customer agreement assistants to reduce rework risk, revenue leakage, and ensure contract compliance.
- Project planning assistants to create profitable project plans and improve planning accuracy and efficiency.
- Project procurement assistants and project expediting assistants to improve fill rates and on-time deliveries, reduce idle time and material wastages, and save on sourcing.
- Project execution assistants to keep projects on schedule and on budget, boost billed amounts, improve cash flow and profitability, sharpen crew time-tracking and efficiency, and enhance jobsite safety.
- Project closing assistants to manage financial closing, punch-list management, and other aspects of closing to speed the closing process and improve client satisfaction.
- Maintenance assistants to track equipment and fleet maintenance, manage service orders, and schedule and dispatch to cut costs and improve asset uptime.
Let’s take a closer look at the AI agents behind a bid optimization AI assistant. A project RFP/tender agent tracks RFPs and tenders the business might pursue. A quantity take-off agent reviews what’s actually in the project, from the number of concrete pours it will need to the size, count, and brand of windows. An estimating agent takes that information and delivers cost numbers. Depending on the project director and team’s comfort level with the AI assistant, it can all happen with varying degrees of autonomy.
Comfort levels with AI will depend on agents’ accuracy, insightfulness, and consistency (so it is with comfort levels with human staff, too). That brings us to data.
Enabling reliable AI assistants and the agents they run on requires three broad types of data.
- A knowledge graph that connects master data, transactions, events, and what we call reference structures (in construction, things like project/WBS, work packages, materials, suppliers, and assets) so AI agents can access the business context they need.
- An industry ontology that connects the knowledge graph with industry context to let AI agents synthesize data, business processes, and expertise into actionable intelligence.
- A “company memory,” if you will, to make sure AI agents adhere to company policies, procedures, contracts, specs, and work standards.
The move to cloud platforms has already put many in the industry in a good position for agentic AI, data-wise. Autonomous construction’s payoffs look certain to justify additional work on that front. Construction businesses of a certain size have told me that they could derive tens of millions of dollars in annual value from a single one of these AI assistants.
Finally, it has become standard practice to offer reassuring words about the importance of the human element. For good reason: In construction, AI cannot and will not replace the knowledge, experience, and judgement of construction professionals. But it will relieve them of putting out the constant fires sparked by historically stovepiped teams and data.
AI will also reduce training needs and flatten learning curves related to new tools and workflows. That’s important, because releases of new features will happen much more quickly than in the past. Users will be presented with intuitive analytics views, and they will interact with the system conversationally, significantly speeding adoption.
The construction business is rife with complexities, contingencies, and interdependencies. It is exceptionally suited to having industry-native AI agents operating as teams under the direction of human-guided AI assistants. Those semi-automated teams will soon handle an array of construction-industry processes quickly, reliably, and at scale. Some of the world’s biggest builders are paving the way. I expect many others will soon follow.
Johnny Clemmons is SAP’s global vice president and industry head for construction.







