Construction AI: Between Building and Buying

Earlier this month, Microsoft put $2.5B and a 6,000-person engineering workforce behind the assertion that enterprise AI does not work when you buy it off the shelf. Its new Frontier operating business embeds Microsoft’s own engineers inside customer organizations to design, build, and run personalized AI onsite. This move suggests that off-the-shelf implementation has contributed to a problem that has been haunting the AI industry: the fact that 95% of organizations get zero return on GenAI investments (MIT Project NANDA, 2025).

Off-the-shelf AI includes Claude, ChatGPT, Gemini, and other apps that do not feature personalized user integration. These tools can be highly effective for individuals, but AI at the enterprise level must be scalable and repeatable across a firm. All employees must use it, and they must use it in the same way.

If you’re a general contractor looking to work AI into your workflow, this move is useful for understanding just how intensive a process it is to integrate AI that will lead to tangible change. AI tools are too detailed for a contractor to build and scale without a dedicated team that is 100% focused on that specific tool alone; the need for firm-specific functionality and integration is too great for such tools to be bought off-the-shelf. In order to be effective, enterprise AI must function consistently at high levels across an organization, and they must be tailored to a firm’s exact needs and existing workflows. Otherwise, integration and functionality pains can result in the AI causing more trouble than it saves.

Buying AI Should Look Like Hiring a Partner

The SaaS industry is moving in a highly personalized and personally integrated direction. With new tools being created for specific industries at the firm level and the growing prevalence of the Forward Deployed Engineer integration format, off-the-shelf software is giving way to embedded expertise. The questions that decide a deal become data ownership, IP, and governance. AI starts to look less like a product than an ongoing, service-based partnership.

Construction is uniquely primed for this format. With subcontractors, architects, engineers, and all the rest, construction has always brought in specialists operating at their peak comparative advantage, and managed all the moving pieces. An AI partner that reviews submittals and sends you an email notification once it’s done is more a helpful assistant than a clunky tool you have to learn how to operate.

The Problem with DIY

Considering the significance of tailoring AI to your firm’s needs, as well as the pains of integrating an external tool into your workflow, it’s tempting to create your own AI tool internally.

The problem is simple: construction firms are not software engineering firms. One person can certainly experiment with AI to automate inbox review, to-do list creation, and other everyday tasks. Personalization at the individual level is entirely achievable, especially for processes that do not require a nuanced understanding of priorities or a high degree of tacit background knowledge. However, when it comes to scaling a tool to the organization level, personalization takes on a different hue. Scalable, repeatable, and consistent intelligence is half the battle. Accuracy and the capacity for critical judgment is the other half.

Even if general contractors do have internal software teams, the truth of the matter is that the best tool will arise from companies whose entire focus is that tool. Rather than expend internal resources on an AI tool that is adjacent to the firm’s goals, general contractors should maximize their efficiency (as well as their payroll) by doing what they do best, and leaving the AI stuff to experts who work with AI all day, every day.

The Price Tag on Custom AI, and How to Get the Best of Both Worlds

The problem with custom-created AI is that it’s expensive to have a software engineer arrive, embed themselves into your company, then build and maintain a custom AI fleet from scratch. It’s no surprise, for example, that all of Frontier’s listed clients are multi-billion dollar companies.

The option that grants you the best of both worlds: Find a tool that’s been developed and tested against the exact pain points of your industry. Ensure that their integration technique is seamless against your workflow. Then, work with their forward-deployed-engineer to build it into your organization. (While you’re at it, tell them about a few more of your pain points, then wait a few months to see them tackle those issues next.)

If you’re a general contractor without hundreds of millions of dollars to spare, the best bet for high-quality, deeply personalized AI tools are companies that work specifically in construction tech, and, better yet, specifically on the very problem you’re trying to solve.

With so many tools available, affordable personalization is closer than you think. The next step is just to select which one fits you best.