CustomGPT.ai has introduced Document Analyst, a new feature that allows users to upload files directly into conversations with their custom agents and analyze those documents against an organization’s private knowledge base. By toggling on the Document Analyst action, users can attach text and image files in chat and receive responses grounded in their own data, policies, and standards rather than generic internet information. The company positions the tool as a way to turn unstructured documents into business intelligence, with every response including citations so users can verify the exact sources behind each insight.
The feature supports uploads of a wide range of formats including PDF (.pdf), Word documents (.docx, .doc), OpenDocument (.odt), plain text (.txt), and image types such as JPG, JPEG, PNG, and WebP, and it can handle both single-file deep analysis and multi-file comparisons. Premium plan users can upload files up to 5MB, and content beyond 3,000 words per document on Premium plans is automatically truncated, with enterprise customers able to request extended file size and word limits. Premium plans support uploading 1 file per prompt by default, while enterprise customers can enable multi-file uploads – up to 3 files per prompt by default, with a maximum of 10 files available upon request, and the 3,000-word limit (or a custom enterprise limit) applies to all uploaded files combined and later files may be partially or fully truncated when the limit is exceeded.
CustomGPT.ai highlights diverse business use cases already emerging from early access, including financial analysis of earnings call transcripts, compliance pre-checks on marketing materials, contract review against standard templates, resume screening aligned with custom hiring criteria, framework application for consultants and students, and technical support driven by spec sheets and incident reports. Document Analyst also features Advanced Vision Intelligence that interprets complex visuals such as technical schematics, architectural plans, product photos, charts, and real estate images in the context of a customer’s own visual knowledge base, and it can search for similar images when vision-processed assets exist in the knowledge base. The company emphasizes security and privacy, stating that uploaded files are processed in a secure environment, stored only temporarily during analysis, not added to the permanent knowledge base, and visible only to the end user, with neither the agent owner nor CustomGPT.ai administrators able to access them.
Document Analyst must be enabled at the agent level via the Actions settings, after which users see an attachment icon in the chat interface and can upload files from web, mobile, embedded widgets, Slack, or API-based deployments. The feature is available only on Premium and Enterprise plans, is fully supported via the CustomGPT.ai API, and is designed for organizations that want to save time on document-heavy workflows or even monetize document analysis. Each document analysis uses advanced multi-step reasoning and counts as 10 queries (1 base query + 9 agentic action cost), and enterprise customers can negotiate custom pricing, extended limits for file size, document word count, and image resolution, as well as volume-based discounts through custom agentic action modifiers.
