Long Lake, a General Catalyst backed AI roll-up firm, made a $6.3 billion bet on AmEx Global Business Travel. By my estimates, this only works if AI takes out half a billion dollars of annual cost (or grows top line commensurately).
If you are a CEO running a business in any services-heavy category, this transaction matters more than the latest model launch from OpenAI, more than the next round of AI conferences, and more than any of the cocktail-party speculation about which incumbent is next.
Here is why.
The strategy is now visible
Long Lake was founded in December 2023. Its proprietary platform, Nexus AI, is a horizontal AI operations layer that gets dropped onto acquired services businesses — automating inbound customer service, accounts payable and receivable, vendor coordination, document generation, compliance reporting, and the sales motion itself.
The results in Long Lake's flagship vertical — HOA and community association management — are now public, courtesy of General Catalyst's own write-up: 25–30% productivity gain per team member, a 10x increase in new-customer pipeline, and free cash flow doubled without reducing headcount. Tasks that took ten hours now take less than one. As General Catalyst put it: "AI-enabled roll-ups can eliminate the Rule-of-40 tradeoff… they are setting their sights on a new Rule of 60 standard."
That is the playbook Long Lake just bought a ticket to run inside the world's largest corporate travel platform.
What that actually looks like at AmEx GBT
AmEx GBT runs ~$2.4B of revenue, ~18,000 employees, and roughly $970M of "cost of revenue" — the line where the services labor lives. Two functions inside that line do almost all the heavy lifting, and both have public benchmarks for what AI does to them.
1. Travel counselors. The biggest pool of automatable labor in the building.
Twelve to fourteen thousand human counselors — across AmEx GBT, Egencia, Ovation, and the just-acquired CWT — answer the phone 24/7 in 30+ languages for booking changes, irregular operations, hotel modifications, VIP requests, and policy exceptions. Call it $600M to $700M of annual labor.
The benchmark for what AI does to this pool is no longer theoretical. Klarna's OpenAI-built assistant did the work of 700 full-time agents in its first month, cut average resolution from 11 minutes to under two, and drove $40M of profit improvement on a single product line. Navan's AI agent already resolves 54% of support requests autonomously at customer satisfaction above its human agents. General Catalyst's sister rollup Crescendo automates 80%+ of customer service interactions and is taking gross margins from the low-30s to 60–65%.
A 50–70% deflection rate on routine traveler interactions across AmEx GBT's counselor pool implies 6,000 to 8,000 FTE of labor displacement, or $400–550M of annualized cost reduction. That alone is enough to roughly double AmEx GBT's $478M Adjusted EBITDA.
2. Mid-office and reconciliation. The same idea, applied to documents instead of calls.
Behind every booking sits a mid-office: fare exchanges, refund processing, queue management, schedule-change reaccommodation, debit-memo handling, supplier invoice matching, expense reconciliation. Rule-based, document-heavy, repetitive — precisely the work agentic AI was built for.
Inside the HOA portfolio, Nexus already runs autonomous AP/AR, vendor coordination, and AI-generated board meeting packets. The structural problem at AmEx GBT is identical: defined input, known rules, defined output. Conservative industry benchmarks put AI automation in transactional finance and back-office at 40–60% labor reduction. Apply that to ~3,000–4,000 mid-office and finance-ops FTEs and you get another $100–180M of annualized savings on top of Example 1.
Stack the two: $500M to $700M of annualized cost reduction against a $478M EBITDA base. That is the math behind the 65% premium Long Lake just paid.
The math problem from the other side of the table
Now run the same math from your side of the table.
If you compete against AmEx GBT — or any large incumbent now in the crosshairs of an AI roll-up — that incumbent already had a 10% to 30% cost advantage over you a year ago. Project forward 24 months, with Nexus or a Nexus-equivalent fully deployed, and the labor cost differential is well over 60%.
A competitor with a 60% labor cost advantage is not a competitor. It is a category killer. They can cut price in half and still expand margins. They can offer service levels you cannot match because their agents are running 24/7 in every language across every account. They can absorb your customers on contract renewal at a price you cannot survive.
This is the asymmetry that mid-market CEOs are not yet pricing into their three-year plans.
Two responses (and a second-order question)
1. Re-platform. This is not a copilot rollout. This is a rebuild of the operating model — workflow by workflow, function by function — to capture the same labor leverage Long Lake is capturing in HOA management. Many do not have the AI fluency, the organizational metabolism, or the data foundation to do this on their own timeline. Most will need outside help. Few will get there in 12 months.
Once you have the platform, a second-order question opens up: stay solo and pocket the margin, or use your new cost structure to roll up your sector the way Long Lake is rolling up theirs? That decision is real and lucrative, but it is a year-two question, not a year-one question. The platform is the precondition for both paths.
2. Sell. If you do not have the capital, the AI talent, or the conviction to re-platform, the rational move may be to take the exit while mid-market multiples are still set by pre-AI assumptions. The arbitrage window closes the moment public market analysts price the AI roll-up threat into incumbent valuations. That window is open today. It will not be open in 18 months.
There is no third option. "Wait and see" is sell — just on worse terms.
So .. how does AI change your strategy? And what are you doing about it?