Airbnb CEO says ChatGPT not ready for app integration

Airbnb CEO Brian Chesky said the company wants to add ChatGPT’s AI technology to its travel app, but the software is not ready for integration.

He told that the current software development kit “wasn’t quite robust enough” for the platform’s needs.

Airbnb recently rolled out new social features, including direct messaging, and upgraded its in-app chatbot, which can now change or cancel reservations for North American users.

Chesky said the company may use ChatGPT in the future, but is currently relying more on Alibaba’s Qwen AI model.

He described the industry as being at the start of an AI revolution and predicted increasing use of the technology in consumer apps in the coming years.

🔗 Source: CNBC

🧠 Food for thought

Implications, context, and why it matters.

Airbnb’s chatbot shows AI gaps that SDKs cannot fix

Airbnb runs its chatbot on 13 models with heavy use of Qwen (Alibaba’s family of large language models), and Brian Chesky told CNBC the software development kit (SDK) “wasn’t quite robust enough” while saying ChatGPT will likely be added.
The bot changes or cancels bookings in North America, needing multi-step reasoning plus application programming interface (API) calls. It needs state kept across booking systems (keeping track of context and user progress) that single-model software development kits (SDKs) struggle to run 1.
Big consumer apps face prompt injection, weak context memory and service-level agreement (SLA) reliability that generic Large Language Model (LLM) APIs do not meet 1.
LangChain, LlamaIndex and LangGraph (developer tools that coordinate multiple models plus tools) help fill the gap. They coordinate models, manage flows and add fault tolerance that enterprises need 2 3.

Third-party orchestration vendors can win business from complex AI rollouts

Startups that build Large Language Model (LLM) orchestration such as Orq.ai focus on multi-model control and observability. They offer real-time monitoring plus System and Organization Controls 2 (SOC 2) plus General Data Protection Regulation (GDPR) compliance. That helps teams with challenges like Airbnb’s and without deep Machine Learning (ML) expertise 3.
Investors should favor vendors with model routing and guardrails for policy plus data. Look for latency tuning and a single workflow across tools, which cuts the need for a custom 13 model stack 1 3.
Beyond travel, healthcare, finance and customer service need multi-agent setups (coordinating several specialized AI agents) for compliance heavy, multi-step work that single SDKs cannot support 4 3.