Between enthusiasm and prudence, the Building, Public Works, and Construction sector is gradually mastering artificial intelligence. While generative AI tools are rapidly spreading across administrative functions, their integration on construction sites remains in its early days. A pioneering study by the Construction Jobs Observatory reveals the obstacles and opportunities of a transformation that seems inevitable.
Carried out in early 2025 with 621 company executives, the survey shows that fewer than 10% of organizations currently declare they are using AI solutions. A delay? Not really. More a prudent approach in a sector accustomed to gradual changes, where the failure of a tool can be measured in lives saved or lost, not in lines of code.
Because while 36% of companies say they are interested in a future deployment, the major moves have not yet begun. “We’ve seen some very interesting applications by people who work with historic monuments,” says the head of a small engineering firm in construction. “They fly drones that scan façades and generate plans to the millimeter. But so far it doesn’t engage many people.”
Generative AI Invades the Offices
If a revolution is on the horizon, it begins with the least valued tasks: office work. Assistants like ChatGPT or Copilot are spreading quietly across support roles. Writing emails, correcting documents, summarizing meetings: all administrative chores eased by these tools, accessible on smartphones.
“Site managers are logging into ChatGPT during construction meetings: AI generates the summaries and automatically prepares emails,” explains the head of a small business specialized in single-family home construction. “They’re first and foremost on the ground, with real value in supervising and managing teams. Here, clearly, it frees up time for them.”
The usage can become even more sophisticated. A small business accountant has thus created “little robots” with generative AI: now, when an invoice arrives in the document management system, the system opens the file, identifies the artisan and the site, creates the necessary folders, and files everything away. “A real paradigm shift in our daily management,” the executive notes with pride.
These “masked” uses or “Shadow AI” nonetheless raise the question of corporate governance. Without a clear policy, everyone tinkers in their own corner with consumer tools whose data security is not guaranteed.
On the Construction Sites, Promises Are Still Distant
Beyond the offices, the practical applications remain largely exploratory. AI could, however, profoundly transform the planning of works. With predictive algorithms, some solutions analyze in real time the progress of work, weather conditions, and resource availability to anticipate delays and automatically readjust schedules.
Safety is another promising area for experimentation. Computer vision systems, integrated into helmets or machinery cameras, can detect when PPE is not worn or when people are present in restricted zones. The goal: move from reactive prevention to a predictive approach capable of anticipating accidents before they happen.
Yet these innovations are still confined to large groups and a few pilot projects. “AI will only be effective if it is applied in an environment that is somewhat more standardized and organized than the ordinary site in our sector,” cautions a PME executive.
Barriers to Adoption: Technical, Economic, and Cultural
First hurdle: data quality. AI can deliver its potential only if it relies on reliable, well-structured information. Yet in many construction firms, data is scattered, poorly organized, and stored in heterogeneous formats. “To have exploitable data, it must be structured, organized, clean. In the Building world today, we’re very, very far,” stresses a technical director of a mid-sized company.
Second hurdle: the lack of interoperability between the sector’s numerous software tools. The different digital tools used—estimating, BIM, site tracking—do not communicate well with each other, making data sharing challenging. “I have lots of different solutions: for estimates, invoicing, banking, accounting… but nothing is interconnected,” laments another executive. “I spend 70% of my time building bridges between tools.”
The economic equation remains uncertain, especially for small and medium-sized enterprises, which account for 94% of the Construction workforce. Deploying an AI solution for these firms requires a substantial budget with a ROI that is difficult to quantify. “What our companies need to see to take the leap is the return on investment,” emphasizes a professional organization. “And for now, those who come to pitch products don’t really provide much of that.”
Finally, cultural barriers are not negligible. In a sector historically attached to manual know-how and transmission through the apprentices’ system, the idea that an algorithm could guide technical decisions triggers resistance. “Know-how is concrete, a gesture, memory, intuition,” summarizes a business owner. “If we delegate too much to automatic tools, we risk losing the field intelligence.”
The seniority of leaders is another hurdle. In a sector where many principals are approaching retirement, investing in technologies whose full fruits they may not see does not seem a priority. “We have an entire category of aging leaders who feel they’ll soon be gone, so why bother integrating innovation,” observes a training organization.
Skills, the Key Stone of Transformation
Faced with these challenges, upgrading skills appears the top lever. But before even discussing AI, the sector must raise its overall digital literacy. “Companies aren’t ready because they haven’t done this upfront structuring,” notes a trainer. “There’s a whole data governance policy to implement before integrating all this.”
In 2024, Constructys funded 1,762 AI training actions, a figure that jumped to 4,406 in 2025. But an analysis of the course titles reveals an offering still focused on general awareness, with little contextualization for the construction trades. Most trainings favor discovery of AI rather than operational upskilling.
Identified needs revolve around three levels. First, a minimal digital culture foundation: mastery of document management rules, data governance, and awareness of collaborative uses. Next, intermediate skills in usage engineering and change management, to identify relevant use cases and mobilize teams. Finally, advanced competencies in interoperability, automation, and system security.
A Transformation in Stages
The Construction Jobs Observatory study sketches the contours of a transformation that will unfold in steps. Support functions will be the first affected, with a shift toward roles of supervision and control of automatically produced content. Design professions, already familiar with BIM, will progressively incorporate AI tools to automate repetitive verification tasks and error detection.
On the sites, site managers become gradually “local data managers,” monitoring dashboards while preserving their hands-on expertise. Maintenance professions evolve toward predictive maintenance, where anticipating faults takes precedence over reactive repairs.
“We’re in a phase where the tools are there, but we must learn how to use them,” sums up a technical director. “AI is not an autopilot: it’s an assistant. The real know-how remains in human coordination.”
Source: Observatory of Construction Jobs, Study on the perception and integration of artificial intelligence tools in construction companies, January 2026
