How Airbnb’s AI Code Shift Is Changing Engineers and Managers

Airbnb’s use of AI is broader than software development. A Q4 2025 earnings-call transcript published by The Motley Fool said Airbnb’s proprietary AI support agent had resolved one-third of support issues in North America and handled nearly 30% of tickets, with plans to expand globally and into voice ^[4]. The same transcript described a broader AI-native experience intended to help guests plan trips and help the company operate more efficiently at scale ^[4].

What changes for software engineers

When AI can draft a large share of implementation work, the engineer’s job does not become irrelevant. It becomes more supervisory, architectural, and product-focused.

The core shift is from writing every line to creating the conditions for good software to exist. That means engineers need to be better at:

• Framing the problem. AI tools are more useful when the engineer can define the goal, constraints, expected behavior, and relevant context.
• Supplying system context. Generated code has to fit the existing codebase, ownership boundaries, data assumptions, and product requirements.
• Reviewing the output. Engineers still need to read the code, understand the tradeoffs, and decide whether it is safe to ship.
• Testing edge cases. Faster drafts do not remove the need for rigorous tests, debugging, and failure-mode thinking.
• Protecting architecture. More generated code can also mean more code to maintain, so boundaries, dependencies, and long-term simplicity matter more.
• Owning production outcomes. A human team remains accountable for reliability, maintainability, and user impact.

The practical result is a different definition of engineering strength. Speed still matters, but raw code volume becomes a weaker signal of value. Judgment becomes the differentiator: knowing what to ask AI to build, what to reject, what to refactor, and what should not be built at all.

Why review becomes a higher-value skill

AI-assisted coding can make first drafts cheaper. That raises the value of the person who can separate a useful draft from a fragile one.

A strong engineer in this environment is not simply someone who prompts a model and accepts the result. The stronger profile is closer to an editor, systems designer, and operator combined: someone who can turn generated code into reliable software. That includes checking whether the implementation matches the product intent, whether it breaks hidden assumptions, whether it fits the architecture, and whether another team can maintain it later.

This is why AI can make senior engineering judgment more important, not less. If teams can produce code faster, the bottleneck often moves to deciding which code deserves to exist.

What changes for managers

Chesky’s comments about managers are just as significant as the coding number. Airbnb’s reported expectation is that managers stay close enough to the work to code or use AI coding tools themselves ^[3][5]. People Matters’ report on Chesky’s warning about “pure people managers” points in the same direction: leadership roles that are only coordination layers may face pressure as AI changes workflows ^[1][2].

That does not mean every engineering manager must become the strongest individual contributor on the team. It does mean technical fluency becomes harder to avoid.

Managers in AI-heavy teams need to be able to:

• understand what AI tools can and cannot do;
• evaluate technical tradeoffs instead of only tracking status;
• challenge weak assumptions in estimates, designs, and generated code;
• set quality gates for reviews, tests, security, and reliability;
• judge teams by shipped value and system health, not by manual coding effort.

The manager’s role still includes hiring, coaching, prioritization, and team health. But at companies moving in Airbnb’s direction, those responsibilities sit alongside hands-on understanding of the tools and technical work.

The roles under the most pressure

The risky role is not “software engineer” or “manager” as a category. The risky role is one defined too narrowly around routine output.

More pressure falls on people who mainly:

• write code without understanding the system around it;
• accept AI-generated output without careful review;
• measure impact by task volume rather than product value;
• coordinate work without understanding the technical substance;
• avoid learning the tools that are becoming part of the team’s workflow.

The safer profile is a technical operator with strong judgment: someone who can use AI to move faster while still owning quality, architecture, and outcomes.

What engineers and managers should do now

For engineers, the immediate move is not to ignore AI or blindly trust it. It is to become excellent at AI-assisted delivery while keeping responsibility for the result. That means writing clearer specifications, giving better context, reviewing diffs carefully, expanding test coverage, and investing in architecture, reliability, security, and product sense.

For managers, the practical move is to stay close to the craft. Use the tools enough to understand their strengths and limits. Join design and review conversations. Make quality standards explicit. Reward teams for durable product outcomes, not for how much code was typed by hand.

The caveat: Airbnb is a signal, not the whole market

Airbnb’s reported 60% figure is an Airbnb-specific data point, not an industry benchmark. It should not be read as proof that every software organization has reached the same level of AI adoption.

Chesky’s own broader view includes both acceleration and patience. In 2024, he said AI would change the world more than many people realize, while also taking longer than many expect ^[6]. That is the right frame for this shift: AI may deeply reshape software work, but the transition will be uneven.

The bottom line is that AI is not simply replacing engineers or managers. It is changing their unit of contribution. Engineers need to become better directors and reviewers of machine-generated work. Managers need enough technical fluency to lead teams that use AI every day. At Airbnb, the durable advantage is moving from manual production to judgment, ownership, and the ability to turn AI output into reliable products ^[1][3][5].