Across the real estate industry, a pattern is becoming hard to miss. Companies that would never have considered building their own software a few years ago are now developing AI tools internally. These include internal AI assistants that summarise communications, CRM enhancements that surface patterns in seller and buyer data, and automations that draft reports, extract information from documents, or flag administrative tasks that need attention.

None of this is surprising. AI has made it faster, easier and less expensive for companies to build their own tools and automate parts of their operations. A small internal team, or in some cases a single employee, can now produce in weeks what would once have required a specialist provider and a significant budget. That is a genuine shift, and for real estate leaders weighing their technology options, it deserves to be taken seriously rather than dismissed.

But the ease of building something is only the beginning of the story. What determines whether that investment pays off has far less to do with the initial build and far more to do with what happens afterwards. Building has become easier. Operating it has not.

There is an important difference between demonstrating a working application and operating one as part of how a business runs. A capable team can now produce something that looks impressive quickly: a tool that drafts responses, analyses records, or answers questions about the business. What that demonstration does not show is what the same application needs once people rely on it every day. This includes keeping information secure, controlling who has access to it, ensuring the information remains accurate, connecting with the systems people already use, and adapting when those systems change.

The evidence on this gap is now substantial enough to take seriously. Research from Gartner found that by the end of 2025, at least half of generative AI projects had been abandoned after the proof of concept stage due to poor data quality, inadequate risk controls, escalating costs or unclear business value. It highlights a growing tension for companies: the pressure to innovate with AI, and the reality of what it takes to successfully operate it within a business. Gartner has gone further with agentic AI, forecasting that more than 40% of agentic AI projects will be cancelled by the end of 2027 for many of the same underlying reasons.

The more useful questions sit 12, 24 and 36 months out, well past the point where the application first works. Who owns it if the person who built it moves to a different role or leaves the company entirely? Who is responsible for making sure its outputs remain accurate as the AI technology behind it changes? Who manages access and permissions as the number of people using it grows across the business? Who tests each update, and who is responsible when a connected system, such as the CRM, document platform or communications tool, changes and the connection stops working as intended?

These are not hypothetical concerns raised to discourage internal development. They are the practical questions that come with owning and operating any business technology, AI or otherwise. The difference with AI is that the technology is changing faster, and the consequences of getting it wrong, such as an inaccurate response presented confidently to a seller or used in a commercial decision, can be greater than with traditional software. Internally developed AI does not remove these responsibilities. It shifts them from an external technology provider to the business’s own team.

Real estate companies already operate across a wide range of systems: CRM, email, messaging platforms, property portals, document management, compliance tools, transaction management, marketing, accounting and general collaboration software. Much of the operational friction in the industry today comes from information being scattered across these systems.

As internal AI development accelerates, it is worth asking a question: does this simplify that environment, or does it risk recreating the same fragmentation in a new form? If different parts of the business, from sales and lettings to property management, valuations and marketing, develop their own AI applications to solve specific needs, each may work well in isolation. The risk is that the business ends up with another layer of disconnected tools, with information and activity still fragmented across the organisation.

Building an application is not the same as achieving lasting adoption, and this matters more in real estate than in most industries, given how much of the working day happens away from a desk. An internally built tool competes for attention with every other system an estate agent already juggles. If it saves genuine time, fits naturally into how they work and proves reliable, it earns a place in the daily routine. If it requires extra steps, produces inconsistent results or simply becomes one more login among many, usage quietly falls away, regardless of how capable the technology is.

It is worth drawing the distinction between building individual AI features and building the operating system that connects and coordinates them, with context shared across the business. Drafting a message, summarising a document or querying CRM data are increasingly accessible capabilities, and a capable internal team can build credible versions of each today. What is harder to replicate is the structure that allows those capabilities to work together consistently across the business.

None of this amounts to an argument against building internally. The more useful question for any real estate leadership team is not simply whether something can be built, but whether it should be built internally or bought. What is the business genuinely prepared to own and support for years rather than months? What creates a real competitive advantage rather than simply replicating capabilities that are now widely available? And will the approach taken today still make sense three to five years from now?

AI has handed real estate companies a genuine opportunity, the ability to build technology around their own business in a way that was previously out of reach for all but the largest organisations. That opportunity is real and worth taking seriously. But making technology easier to build does not make it easier to operate over the long term. As more real estate companies develop AI internally, they are also taking on responsibility for security, governance, integration and ongoing ownership, areas that have traditionally been managed by specialist technology providers.

At its core, a real estate company is in the business of real estate, not software development. Building internally may make sense, but the time, budget, resources and attention it requires still have to come from somewhere.

Deciding where AI can create lasting value and how it should work across the business has never been more important.