Add OCR capabilities to AI agent
Autopilot
Merged in a post:
AI Skills Cannot Access or Parse Image Attachments — Critical Triage Context Gap
A
Andrew Esau
Thread's AI skills (KB Search, Technician Brief, RIA-KB, SuperMagic) cannot read or parse image attachments submitted by customers, even when attached directly to the Thread conversation and synced to the PSA ticket.
Real example: On a recent ticket submitted, the customer attached a Dell SupportAssist diagnostic screenshot. The KB Search and Brief skills ran with zero awareness of the image — operating only on her one-line text message. SuperMagic confirmed it has no tool to retrieve or parse attachments at all.
Urgently Requested:
Attachment retrieval tool for SuperMagic and AI skills
OCR/image parsing so extracted text (error codes, diagnostic readouts) is injected into AI skill context
Explicit flagging in skill Brief when an unreadable attachment is present — rather than silently omitting it
Diagnostic screenshots are among the most valuable context a customer can provide. This gap silently degrades every AI-assisted triage run where a customer attaches additional context outside of text.
Computer Village
I feel like it's reading some attachments and not others in our PSA (Autotask). I'm going to try and catch that.