TechCrunch has compiled a running list of failed AI products and startups, from workflow tool Relay to chatbot platform Character. ai, and paired it with a stark figure: about 42% of corporate AI initiatives are ultimately abandoned, according to S&P Global Market Intelligence. The reasons include weak funding, technical problems, competition and low demand. For business, the list matters because it shows which AI bets survive inside large platforms and which do not.
Which products shut down and when
Relay, a five-year-old AI workflow automation alternative to Zapier for email and task workflows, shut down entirely in August after OpenAI, Google and other platforms added similar automation directly into their tools. Character. ai, a chatbot platform launched five years ago, more recently filed for bankruptcy. Notion Mail, launched in April 2025 as an AI inbox organizer, is scheduled to shut down on September 22 after users preferred separate AI agents for email. The pattern is consistent: stand-alone assistants lose when the same capability becomes a built-in feature.
OpenAI illustrates the consolidation logic from the other side. On July 9 it tried to turn ChatGPT into a broader "super app" by combining Chat, Codex and Work modes and renaming the familiar version ChatGPT Classic, then rolled the change back after users called the interface confusing. ChatGPT Atlas, a stand-alone AI browser, lasted less than a year before discontinuation on August 9, with features absorbed into ChatGPT. The same happened to Operator, a web-browsing agent, while image generation moved inside ChatGPT and reduced DALL-E to a smaller separate role. Video platform Sora shut down in March 2026 amid high operating costs and retention problems.
Large platform bets outside OpenAI followed a slower failure path. Apple unveiled Apple Intelligence in 2024 with a context-aware Siri that would work across apps, then repeatedly delayed the release over engineering issues and bugs, a delay linked to a $250 million settlement over marketing of iPhone 16 AI capabilities. The new Siri appeared in the iOS 27 beta this past July, with English rollout this month and other languages to follow. Microsoft announced Recall at Build 2024 as photographic memory for Windows PCs through periodic screenshots, delayed it for nearly a year after privacy backlash, and still faces questions after a researcher built a tool extracting captured Recall data.
What failed AI bets mean for business
For companies buying or building AI, the main lesson is to prefer capabilities anchored in platforms with distribution over narrow stand-alone tools. Relay, Huxe and Notion Mail lost when OpenAI, Google and Spotify offered similar automation and audio inside products customers already used. Huxe, an audio app from former Google NotebookLM developers that turned text into podcast-style audio, shut down in May 2026 for that reason. Small firms gain faster access and lower integration costs by waiting for built-in features, while large firms can standardize on fewer vendors instead of supporting fragile point products.
The limits are financial sustainability and security review, not only product quality. Humane raised $230 million for the AI Pin wearable, then warned about a charging-case fire risk and shut the business in February 2025, with HP acquiring most assets for $116 million. Rabbit said it sold 100,000 R1 units after CES in January 2024, yet reviews cited unfinished performance and few integrations; it now positions R1 as an agentic computer controller and plans Project Cyberdeck for vibe-coding. Yupp, which let users compare over 800 models from OpenAI, Google and Anthropic, shut in March 2026 for lack of product-market fit, while Figgs AI closed after serving over 1 million users because a free role-play service became too costly.
The marker to watch is whether delayed platform features reach stable multilingual release without new legal or security setbacks. Siri expansion beyond English after the iOS 27 beta and Microsoft handling of Recall data extraction will show if the giants can convert controversial AI bets into trusted defaults. If both proceed on schedule, reliance on separate AI assistants and recorders will keep shrinking.
