Construction Needs to Learn What ChatGPT Can’t Do

This is a translation from chronicle written by CEO and co-founder Herman Smith, originally published by bygg.no.

I use ChatGPT every day. It shows how far generative AI has come, and I see a lot of untapped potential for the construction industry in this kind of technology.

But it’s precisely when you apply tools like this to construction that the limitations become obvious. Our industry is built on complex value chains, one-off projects, specialized processes, and strict competence and responsibility requirements. Each of these factors on its own makes it clear that generic, “one size fits all” AI isn’t enough.

That’s where the distinction between horizontal and vertical AI becomes critical. Over the course of 2025, something interesting happened in how we talk about AI: the conversation shifted from being impressed by broad, general-purpose models to understanding why industry-specific, vertical AI is necessary to actually change how work gets done in sectors like construction.

The realization has grown that general-purpose AI on its own cannot solve industry problems in a reliable, accountable way. Horizontal AI doesn’t automatically fit the unique constraints and risks of construction, and we need to understand why that is, and what a vertical approach can do differently.

Horizontal AI, Tools for Everyone

Think of a Swiss army knife, a tool that can be used by many people in many different situations.

This is horizontal AI: solutions like ChatGPT that are designed for general tasks and work across industries. They’ve opened up entirely new possibilities for writing emails, generating ideas, suggesting structures, drafting code, translating text, or summarizing information. They are useful, accessible, and getting better all the time.

But these tools have clear limitations when they’re applied to the specific challenges of a given industry. At least for two reasons:

1. Limited access to industry-specific data

2. The need to combine multiple technological approaches to solve complex problems

Construction generates enormous amounts of data. Most of this information is completely invisible to general AI models, which are typically trained on open and publicly available data. The result is that horizontal AI struggles with the specific needs of the industry, where unique terminology, exact references, and complex datasets require targeted analysis and deep domain understanding.

Even as horizontal AI continues to improve, it will still hit its limits in industry-specific processes that require:

• combinations of different technologies,

• adaptation to real-world workflows, and

• precise understanding of the sector.

Vertical AI, Built for Construction

If horizontal AI is the Swiss army knife, vertical AI is a specialized power tool, designed for a particular job, with the performance and features you actually need.

Vertical AI is developed to meet the needs and challenges of a specific industry. For construction, that means solutions that can:

• handle large volumes of complex text,

• optimize solutions and decisions, and

• work with project data in a trustworthy and secure way,

all based on the industry’s own language, standards, and specifications.

Where horizontal solutions know a little about a lot, vertical solutions need to know a lot about a little.

For project managers and engineers to truly rely on AI in decision-making and process improvement, the bar for trust and credibility is high. Vertical AI must deliver solutions that don’t just save time, but also provide real insight into complex problems.

Today, the construction industry is estimated to reuse only around 4 percent of the data generated in projects. Vertical AI makes it realistic to reuse more of that data, learn from it, and improve projects faster than we manage today.

In an industry with high demands on efficiency, safety, and accuracy, that matters.

The Next Digital Step for Construction

It’s natural that construction companies often start their “AI journey” with horizontal tools like ChatGPT. They’re easy to access, easy to experiment with, and a great way to build familiarity with AI.

But as the industry matures, the limitations become clearer, and the need for tailored solutions that address construction’s actual challenges grows harder to ignore.

AI is not magic. It is a powerful tool when we understand both its possibilities and its limitations.

To make real progress, we should start with two simple questions:

• Which problems are we actually trying to solve?

• How can we use our data more intelligently?

Horizontal and vertical AI are not competitors, they complement each other.

• Horizontal AI handles the broad, general tasks.

• Vertical AI provides precise insights and automates specialized work.

Used together, they give the industry a much better starting point than either one alone.