[ad_1]
Having developed business software for knowledge workers, I know that the biggest hurdle to an application’s success is asking people to manually enter existing data. If the data is already somewhere, why do I have to pick it up, perhaps interpret it, and re-enter it into a new system?
For twenty years, the industry has tried to digitize construction by asking people to enter more information into more forms and systems. BIM, CDEs, project management platforms, safety and quality systems, ERP systems, mobile apps, and IoT solutions all depend on people feeding them data. The bottleneck has never been the software. It has been the human effort required to keep information flowing.
A few days ago, I interviewed Ethan Ow, CEO and co-founder of Wenti Labs, a Singaporean startup. Ethan shared his views on data and digitalization and discussed how his firm is tackling its challenge with agentic AI. Instead of offering another SaaS, they are orchestrating existing systems.
What construction really runs on
Ethan remarked that construction runs on WhatsApp and Excel. It’s a humorous comment, but it looks like construction IT in Singapore and Finland has at least one thing in common. In Finland, WhatsApp is probably the most popular communication method on construction sites.
The point of Ethan’s notion is that official systems of record are rarely where information is first stored. Information originates in conversations, messages, photos, phone calls, and site observations. Most digitalization efforts have tried to force these informal workflows into formal systems, leading to manual re-entry.
Ethan’s startup has shown that agentic AI can change that. It can interpret unstructured data from various sources and introduce automation into informal workflows.
Shifting attention from systems to workflows
Historically, software vendors have sold systems, and users have adapted to the software. AI agents reverse this relationship: the software adapts to the workflow.
That may prove to be one of the most disruptive consequences of generative AI. If AI can bridge between systems, classify information, complete forms, and move data automatically, companies may no longer need to redesign their processes around software limitations.
The interaction becomes conversational and contextual rather than form-based. Wenti Labs’ May 1 press release offers an example. A site safety manager can photograph a hazard and send it via a chat app to their team. Wenti’s AI automatically logs it as a safety issue, categorizes it by severity, and notifies the relevant project team if the issue is not resolved in a timely manner, all without opening any additional applications.
Better data may be AI’s biggest outcome
Most discussions about AI focus on productivity. Ethan claims that the more strategic impact is eventually data quality.
Today, much project data is incomplete, inconsistent, or never recorded at all because documentation takes too much effort.
If AI captures information automatically and structures it consistently, the industry gains something it has struggled to achieve for decades: reliable operational data. Only then do concepts such as predictive project controls, safety analytics, carbon tracking, and autonomous workflows become practical.
Wenti Labs’s “agentic OS for construction”
Wenti does not require customers to adopt a new platform or redesign their processes. Instead, the startup studies each customer’s existing processes and integrates AI agents into the workflow. They call this “an agentic OS for construction.”
The company positions itself almost as a digital workforce. Ethan describes Wenti as an “internship company,” except that the interns are AI agents running on cloud infrastructure rather than people sitting in an office.
This philosophy extends to software integration. Rather than replacing systems such as Procore, HammerTech, Autodesk, or Aconex, Wenti serves as an intelligent layer that enables information to flow between them.
The company does not store the customer’s data, so there’s no “data lock-in” as is the case with typical SaaS firms. If the customer decides not to continue with Wenti, they can keep on running their workflows with the original data.
The challenge for agentic AI startups
This and other discussions with construction AI startups raise a critical strategic risk.
Ethan says they are not building their own model. They use OpenAI, Gemini, Claude, etc. Why should a contractor buy from a startup instead of building its own AI agents? Or replace Wenti with another startup’s agents since Wenti does not even have the contractor’s data.
The real product may be to know how to turn construction workflows into AI workflows. That’s not something every contractor wants to learn on their own, even if they’d have the tools and funds to do so. In fact, Wenti’s first customer was Who Hup, one of Singapore’s largest contractors with ostensibly the resources to build their own agents.
What is the future of SaaS?
Ethan’s interview makes it clear that companies are unwilling to revamp their workflows and give up their core IT systems because of AI.
Any software has a system record (database, business rules, permissions, etc.) and a user interface (forms, menus, buttons, etc.). Today, humans interact with the system through the user interface. AI is bypassing or, in some cases, operating on that layer.
This means that the traditional tools and systems remain relevant for some time. AI will not yet replace Procore; it may replace the need for humans to navigate Procore.
Wenti’s solution is realistic for the time being. However, I can see an alternative future in which AI agents can code the necessary capabilities themselves if they have raw data (photos, videos, sensor data) and do not need traditional tools to process it.
Removing the friction
The interview’s central message is that the greatest opportunity for AI in construction is not replacing engineers or creating ever more sophisticated digital models. It is removing the friction between physical work and digital information.
A housing contractor told me that it now takes more on-site hours per square meter to record and report compliance data than it did two decades ago. Add the increased complexity of construction, and it’s no wonder productivity growth is lagging.
When machines can track progress, compliance, safety, and other critical aspects without human intervention, sites can reclaim the time it now takes to type in all that information manually.