“We hired our CTO, Ahmad Al-Dahle. He was the leader of Meta Llama models. He came in, I think we went from a company that was a middle-of-the-pack company for AI to a leader in AI, at least amongst the companies that are not frontier labs or hyperscalers.”
Brian Chesky, co-founder and CEO of Airbnb, said this during Airbnb’s Q2 2026 earnings call. This is actually an interesting change to watch. We are starting to see human capital move from AI-first companies to more traditional companies. People who spent years building frontier AI are now moving into companies whose core business was built before AI. Airbnb is just one example.
For artificial intelligence, there is a main discussion we are stuck in: Is AI creating a productivity boom or an investment bubble? We still do not know.
The IMF estimates that if AI delivers the productivity gains expected by markets, global economic activity could be 0.3% higher in 2026. If AI valuations correct and financial conditions tighten, global growth could be 0.4% lower than the baseline (World Economic Outlook, October 2025).
And then there is the other side of the question. What if AI really creates a massive productivity gain?
If companies can produce more with the same number of people, what happens to jobs?
We seem to be caught between two risks: Too little productivity and too much productivity.
What a dilemma.
While the world is discussing this, maybe we should look not at the future, but at what is already going on inside companies.
Now, you may think that Airbnb was in financial crisis and turned around with the help of AI.
No, it wasn’t. Airbnb was already a well-performing company. In Q1 2026, its revenue grew 18% year over year to $2.7 billion and adjusted EBITDA grew 24% to $519 million. What happened was something different. A well-working company increased its execution capacity with AI.
Let’s check what happened.
I see this execution capacity increase in six areas: Speed, volume, reach, learning, leverage and operating model.
Airbnb is using AI across software development and product creation.
In Q1, the company said nearly 60% of the code its engineers produce is now coauthored with AI, roughly twice the industry average by its estimate. The obvious benefit is faster coding.
The hidden one is speed of learning. There is a shorter distance between an idea and a product. If an engineer can build, test and modify something faster, the organisation can run more experiments. And if you can run more experiments, you increase the possibility of finding something that works. Maybe this is more important than efficiency.
Chesky says AI is transforming how Airbnb executes and builds products.
For years, companies have talked about being agile. Maybe AI is finally giving agility an operating mechanism.
Airbnb’s engineers are not simply producing code faster. They can produce more product improvements with the same organisational capacity.
In Q2, Airbnb said the number of features and improvements it shipped increased by nearly 80% compared with the same period last year.
The question of how many people we need to produce X is slowly being replaced by how much X can the same people produce with AI.For a better understanding, the company that can test 100 ideas while its competitor can test 50 has twice as many opportunities to learn. AI increases that experimental surface. And this is where productivity starts to become something more than efficiency.
Customer support is one of the biggest challenges of the service industry. More customers usually mean more support requests. More support requests mean more people. Airbnb’s AI Assistant is now available in more than 50 languages. It knows the context of the user’s trip and can resolve issues directly inside the conversation. In Q2, nearly 45% of issues that started with the AI Assistant were resolved without a human agent. In Q1, the figure was already over 40%.
Think about what this means operationally.
Without AI formula:
More customers → more support requests → more people.
With AI formula:
More customers → more automated resolution → human capacity concentrated on the difficult issues.
And this is not only a theoretical efficiency.
In Q2, Airbnb’s customer-support-related cost per booking declined approximately 16% year over year, driven partly by improvements to its AI Assistant. The company can therefore reach more customers without scaling every layer at the same rate.
That is a real example of execution capacity.
This may be the more interesting part. Airbnb has more than one billion guest and host reviews.That is an enormous amount of information. But information is not what we call as insight.
AI can turn this information into usable signals. Airbnb’s 2026 product now uses AI to synthesize listing descriptions and guest reviews and highlight things users actually care about, such as location, amenities and whether a home is family-friendly. It is also introducing AI-generated comparisons between homes.
Shortly, the role of data changes with AI.
New formula with AI:
Data → insight → decision → action → new data.
The company starts learning at a different speed.
This is where the employment debate becomes more complicated. AI can certainly replace parts of jobs. But Airbnb shows another possibility. AI can multiply the output of the people who remain.
An engineer becomes not only a coder, but an orchestrator. A support agent does not have to answer every question, but can handle the exceptions. A product team can run more experiments without adding people at the same rate.
There is more space for qualified work. The scarce resource starts to move. From execution itself to judgement, prioritisation and orchestration are placed.
The labour-market discussion is usually framed as AI or human.
Airbnb opens another point:
Human with AI or human without AI.
That is a very different labour-market dynamic.
And this is where the human-capital transfer we started with this article becomes important.
When someone who has spent years building frontier AI moves into a company like Airbnb, the
transfer is not only knowledge. It is also a transfer of how the organisation thinks about AI.
From building models to building products. From technology to execution.
From having AI to using AI wherever it can increase capacity.
Airbnb is not looking at AI simply as a tool. It is looking at what AI makes possible for the organisation.
AI is now being embedded into search, review interpretation, trip planning, customer support and host tools. It is moving inside the workflow.
And once that happens, the question is no longer whether an employee is using AI or not.
It becomes a natural part of how the company operates.
So, what is the competitive advantage?
In competitive strategy, we usually think about advantage through technology and capability.
Technology → Capability → Advantage
AI may be adding something in between.
Technology → Capability → Execution Capacity → Advantage
While capability tells us what a company can do, execution capacity tells us how much, how fast and at what cost it can actually do.
And this may become one of the most important competitive dimensions of the AI era.
Everyone can buy the same models. Everyone can subscribe to the same AI tools.
Everyone can build a chatbot. But not everyone can turn those tools into:
- faster product development,
- more experiments,
- lower support costs,
- better use of data,
- faster learning,
- and more output from the same organisation.
That is the difference between having AI and being AI-enabled.
Airbnb seems to be giving us one of the early answers.
And if more companies start showing the same pattern, the AI story may turn out to be neither a bubble nor a job apocalypse. It may be something more interesting.
A change in the productive capacity of the organisation itself.
