Seven Lessons to Drive a Successful Enterprise AI Transformation
People and change management are as important as the technology under the hood
What does it mean to drive an AI transformation at an enterprise? Can a rollout be flexible enough to handle the messy processes of modern-day organizations filled with specialized knowledge workers? How much change management is required?
I recently had a chance to sit down with two forward-thinking technology leaders at a real-estate brokerage who managed to drive a successful AI transformation effort in their organization.
Brokerage work involves a messy combination of workstreams that are difficult to “automate” (most realtors would argue that wouldn’t even be the goal). Successful real estate agents need to be agile and balance juggling commercial opportunities, providing client service, and closing deals. Brokerage firms have historically not invested significantly in technology. And realtors themselves are notoriously independent (they’re free agents in many respects). So looking at a brokerage that successfully managed an AI transformation is a perfect case study for many other service heavy fields (such as law, accounting, finance, medical practice management) with highly educated - and highly opinionated - knowledge workers.
What are some key lessons learned?
1. Start With the Human Problem, Not the Tech Solution
The most successful AI deployments are rooted in understanding where humans add the most value—and where they’re bogged down by repetitive, low-leverage work. This team focused on helping real estate agents spend less time doing admin and more time selling. It’s a human-first approach: AI should empower, not replace.
“If a salesperson is not selling, everybody’s losing money.”
Early features shipped prioritized identifying tasks that brokers hate doing, which could be offloaded to AI. Solving these pain points early helped drive buy-in for the AI transformation effort early on, culturally “priming” the organization for change.
2. Keep Humans in the Loop—Intentionally
AI wasn't deployed as a standalone system but as part of a human+machine teaming model. From "objection handling" suggestions to psychometric-based personalization, the AI augments the agents rather than trying to mimic or override them.
“We want to give them sharp, shining tools… not automate them out of the loop.”
Erring on the side of keeping humans in the loop, and focusing on augmentation (versus full automation) early only was important - both in terms of setting expectations within the organization, and building trust in AI outputs.
3. Build for Simplicity, Not More Screens
Rather than adding more software layers, the product emphasized a screenless experience—invisible AI that reduces cognitive load. Design choices early on focused on integrations to existing tools, and features where the AI could be called on from those tools (versus making users adopt yet another program).
The most loved feature? A "sync" button for users to automatically update a CRM tool, enabling brokers to avoid manual back and forth, copy / pasting between screens which they previously were forced to do.
“The vision is being the world’s first mostly screenless brokerage.”
4. Don’t Rebuild the Wheel—Be Strategic About Buy vs. Build
The team focused on building only what was unique to their business—like the “Simple” assistant. For everything else (CRMs, integrations), they partnered or used off-the-shelf solutions. The question wasn’t “can we build it?”— it was “should we?”
“If there’s not a capital-N need to reinvent the wheel… we don’t.”
Using off the shelf tools, and avoiding custom builds enabled the team to avoid costly development cycles, proving value sooner.
5. Personalization is a Killer Feature
Another standout area where AI drove significant value were capabilities powered by AI enabling personalization at scale, in areas like customer messaging or internal recruiting.
A notable example involved staffing for the AI team itself. The company had limited success with recruiters, who were finding generic candidates with broadcast type approaches. So the CTO tried a different method, using AI to generate a more tailored JD created from a transcript of a conversation he captured with the leader of the business, then combining that with the output of a team strength assessment to narrow down a more targeted set of candidates, whom he then messaged using AI to generate personalized outreach to each. The result was the CTO found his ideal AI product leader (sitting next to him in the conversation).
“We used OpenAI to write personalized recruiting messages based on candidate personality… and we haven’t used recruiters since.”
Personalization use cases in particular offer a spectrum of benefits — enabling bespoke customer experiences, not just cost savings — that can create “aha” moments in service-heavy businesses.
6. Transparency, Explainability, and Trust Matter
Especially in high-touch industries, people need to understand why the AI is recommending something. That makes agents more confident presenting AI-assisted outputs, for example.
One popular use case for brokers involved AI automating the time-consuming generation of comps for property listings. The team found, however, that it wasn’t enough to generate just the client-facing output itself, but they also had to include a separate report with the backups and all the details that went into the main report (even if this wasn’t client facing).
“They need to be able to speak about it authoritatively with their clients as if they did it themselves.”
Providing the backups created trust and equipped the agents to better answer questions that came up from customers. The result of this approach was greater adoption and usage of the tool, with greater agent confidence in the accuracy of AI-generated customer deliverables.
7. Give Primitives, Not Black Boxes
Lastly the team emphasized the approach of giving users modular, composable tools—like a LEGO set. Not a monolithic agent that takes over, but building blocks they can use and adapt.
“If you give people the primitives and they have a LEGO-like approach, that’s more valuable.”
In the example of the comp reports, providing the backups also enabled agents to combine outputs with their own expertise and understanding of the customer, giving agents the flexibility to modify outputs and use the tools in the service of customers, not the other way around.
Conclusion
Service heavy businesses represent a unique opportunity to deploy AI, creating both opportunities and challenges. Approaches focusing on just the technology alone miss the mark, with success dependent on approaches that empower workers and adapt to messy workflows. But incorporating some of the lessons can drive real transformation.