David SenraFrom HOA Management to $4B in Revenue & $6.3B Deal for Amex GBT | Alexander Taubman, Long Lake
At a glance
WHAT IT’S REALLY ABOUT
Long Lake’s playbook: buy service giants and embed AI deeply
- Long Lake, founded about three years ago, aims to deploy AI into “real economy” service businesses where legacy systems, messy data, and workflow change are the primary obstacles to adoption.
- The company announced a $6.3B take-private of American Express Global Business Travel, positioning it as a flagship opportunity to modernize travel operations and personalize traveler experiences using AI at massive scale.
- Long Lake operates across five verticals (HOA management, HR services, specialty tax, infrastructure services such as architecture/engineering, and travel), with roughly 40 acquisitions and projected combined revenue of over $4B next year.
- Its AI platform, Nexus, is model-agnostic and largely reusable across industries, but requires intensive in-field engineering to integrate into each business’s workflows—leading to faster adoption and retention of both employees and customers.
- Taubman frames Long Lake as a permanent-capital, long-duration compounding model influenced by Danaher, Henry Singleton, and John Malone, with a potential future IPO to access deeper, lower-cost capital for even larger ambitions.
IDEAS WORTH REMEMBERING
5 ideasThe big opportunity isn’t AI invention; it’s AI diffusion into legacy workflows.
Long Lake is built to take frontier AI capabilities and make them work inside messy, legacy-heavy service businesses where data, workflows, and change management are the real bottlenecks. The bet is that “AI diffusion” into real operations is harder—and more valuable—than building another standalone software tool.
Long Lake uses AI to grow revenue, not to gut headcount.
Taubman argues productivity gains in service businesses expand capacity, improve retention, and enable faster growth—so you often hire more, not less. Long Lake’s goal is to remove low-value toil (email, data entry, process work) so humans spend time on customers and growth.
Nexus is a reusable horizontal AI platform plus hard-won vertical workflow integration.
Nexus sits between AI models and company systems: model-agnostic routing, security, orchestration on one side; deep integrations into each business’s systems and workflows on the other. Most of the platform is reusable across industries, while the hardest work is embedding it into real operational processes.
Real-economy AI requires engineers in the field, not just in the lab.
Instead of centralizing engineers in HQ, Long Lake places engineers with frontline operators across many states to observe friction, map workflows, and iterate quickly. This “Skunk Works” style approach is presented as essential to making AI usable and adopted.
They choose ownership over selling software to stay accountable for outcomes.
Long Lake doesn’t sell Nexus broadly because software vendors can succeed without ensuring the customer’s operations truly change. Ownership aligns incentives around outcomes—growth, customer experience, employee experience, and long-term compounding.
WORDS WORTH SAVING
5 quotesWe formed the company three years ago with the mission of basically bringing AI into the real economy.
— Alexander Taubman
Deploying technology innovations in the real economy is actually really hard. There's tons of legacy systems, tons of legacy data, workflows, change management.
— Alexander Taubman
We're very positive sum. You know, we really believe in this concept of, like, AI abundance.
— Alexander Taubman
If we have to force someone to use our Nexus platform, then the platform's not good enough.
— Alexander Taubman
We have these amazing quotes now from our team members. They're saying like, "I don't feel like I'm working anymore," because there's so much mental friction that comes from that thirty percent of work that's, you know, this manual road, responding to emails, data entry, da, da, da, da, da.
— Alexander Taubman
High quality AI-generated summary created from speaker-labeled transcript.
