Stop hunting screenshots — surface your visual library inside AI chats

Knowledge workers—estimated at roughly 120 million globally—amass large, fragmented visual libraries of screenshots, slides, and diagrams that sit in folders, note apps, and cloud drives and are difficult to retrieve when needed. That friction becomes acute as more work is happening inside LLM-driven chat UIs: people trying to answer questions, draft documents, or make decisions often hunt for images outside the chat, breaking flow and costing time. You could build an in-chat visual retrieval layer that indexes personal and team visual assets with image embeddings and surfaces semantically relevant images directly inside AI chats as users ask questions or draft content. Core capabilities would include automatic capture/import connectors (desktop/mobile/cloud/Slack), fast vector search tuned for conversational prompts, and privacy-first options such as on-device embeddings or encrypted indexes plus simple sharing controls for teams. The market is attractive now because the TAM is roughly $48B (120M knowledge workers × $400/yr on productivity and visual-asset tooling) and three trends converge: chat-first workflows, multimodal models with accurate image embeddings, and a surge in personal knowledge management making centralized access an expectation. To stand out you must solve hard product and engineering problems: seamless native integrations into major LLM UIs, enterprise-grade privacy defaults, and retrieval relevance optimized for dialog rather than file-finding. Those are non-trivial but defensible advantages; the main challenges are platform fragmentation and building trust around images and privacy, while the upside is a clear distribution pathway and attractive revenue from users who will pay to eliminate repeated context switching.