Relay Shutdown Adds to the Expanding List of AI Product Failures
Relay, an AI-powered workflow automation tool positioned as an alternative to Zapier, shut down entirely on Monday after five years. The product used AI agents to automate email and task workflows, but faced intensifying competition as OpenAI, Google and other large platforms added comparable automation capabilities to their own products.
Its closure illustrates a broader pattern of AI launches being shut down, delayed, absorbed into larger services or redirected. S&P Global Market Intelligence estimates that about 42% of AI initiatives are ultimately abandoned by their corporate parents. The reasons cited include insufficient funding, technical obstacles, competition, scaling difficulties and weak user demand.
Standalone products face platform pressure
Relay’s challenge was not simply building automation. Larger vendors increasingly control the applications where work already happens, making it harder for a separate product to retain a distinct role when similar functions are integrated into established platforms.
Several recent examples show how quickly AI products can be consolidated. OpenAI rolled back a July 9 redesign that combined Chat, Codex and Work modes and renamed the familiar experience ChatGPT Classic after users criticized the interface as confusing and cluttered. It also discontinued its standalone ChatGPT Atlas browser on August 9, moving useful features into ChatGPT.
OpenAI similarly folded Operator, an agent designed to browse the web and complete tasks, into ChatGPT. Image generation has also become increasingly integrated into ChatGPT, reducing DALL-E’s role as a separate destination. Its Sora video-sharing platform shut down in March 2026 after high operating costs and user-retention challenges.
Delays, privacy concerns and difficult economics
Not every AI setback is a shutdown. Apple delayed the enhanced Siri promised with Apple Intelligence because of engineering issues and bugs. The delay contributed to a $250 million settlement over claims about iPhone 16 AI marketing. Siri AI appeared in the iOS 27 beta in July, and the English-language rollout began this month.
Microsoft delayed Recall for nearly a year after its screenshot-based search feature for Windows PCs prompted privacy and security concerns. Even following a redesign, a cybersecurity researcher demonstrated a tool that could extract and display data captured by Recall, renewing questions about the protections.
Notion Mail, launched in April 2025 as an AI-focused inbox product, is scheduled to close on September 22 after Notion observed users turning to separate AI agents for email work. Humane shut down its AI Pin business in February 2025; HP acquired most assets for $116 million after the wearable struggled with performance and its charging case raised a potential battery fire risk.
What businesses should test first
These cases do not make AI experimentation futile, but they show why deployment decisions need to examine more than technical novelty. Businesses should validate sustained demand, operating costs, security exposure, integration needs and whether a larger platform can readily reproduce the proposed capability. A practical implication is to treat a standalone AI tool as a durable business option only when it solves a clear user problem that remains valuable beyond the next platform update.