Documentation

Introduction

Succinctly is an AI-powered assistant that answers questions in natural language, grounded in your organization's own documents. Instead of digging through PDFs, spreadsheets, and file shares, you simply ask a question and get a concise, accurate answer in seconds — with citations showing exactly which documents and pages the answer came from.

Every deployment is customized for its organization: your branding, your greeting, your document library, and an assistant persona tuned to your users. The same assistant can be reached through the web application, WhatsApp, and SMS.

The application is organized into four main areas, accessible from the navigation bar at the top of every page: Chat (ask questions and get sourced answers), Documents (manage the knowledge base), Analytics (see how the assistant is being used), and Admin (configure the instance).

Chat

Type a question into the input box and press Enter. The assistant responds within seconds with a direct, synthesized answer — not a list of document excerpts.

Answer modes. A Mode selector next to the input box controls where answers come from. Docs Only (default) grounds answers strictly in your uploaded documents — the assistant will not use outside knowledge or invent details, and says so explicitly if the documents don't contain the answer. Docs plus Internet combines your library with information from the web. ChatGPT Mode is general-purpose LLM chat without document grounding.

Source citations. Every grounded answer is accompanied by a Chat Information panel showing the exact source files and page numbers used (for example, "ATHLETE PACKET.pdf — Page 3"), plus any related images from the library. Answers are never a black box.

Targeting specific content. Type @ anywhere in your question to scope it to a specific file, folder, or workspace. Or click + Context to attach a file, folder, workspace, or a fresh upload from your computer.

Feedback and controls. Every answer has copy, thumbs-up, and thumbs-down actions — ratings flow into Analytics so administrators can see which answers helped. A Reset control clears the conversation and starts fresh.

Document Store

The Document Store (Documents → Document Store) is the knowledge base that powers every answer. Upload your materials and go — there is no separate training step.

Uploading. Drag and drop files or folders, or click Browse files. Supported types go well beyond text: PDF, DOCX, PPTX, TXT, XLSX, and HTML documents; JPEG, JPG, and PNG images; MP3 audio and MP4 video. Media files are indexed alongside documents, so the assistant can draw on maps, diagrams, and recorded briefings too.

Cloud import. Built-in connectors pull documents directly from Google Drive and SharePoint — no downloading and re-uploading.

Organization. Search files by name, sort by name, type, date, or size, and create folders to keep material organized. Select files for bulk edit or delete, or use the per-file menu to download or delete individual documents. An Associated Media panel shows images and media linked to each document.

Compare Documents

The Compare tool (Documents → Compare) produces an AI-generated comparison of any two documents — this year's rules against last year's, or two versions of a contract — saving the manual side-by-side review.

Load each of the two panels by either uploading a file from your computer (Choose File) or picking an existing one (Browse Document Library), then click Compare Documents. The assistant analyzes both and reports the differences.

Workspaces

Workspaces (Documents → Workspaces) group files and projects into named collections — one per department, project, or event.

Create one with + New Workspace, then organize files into it. Workspaces become targets in chat: mention one with @ or attach it via + Context, and the assistant answers from that collection only.

Analytics

The Analytics page gives administrators full visibility into how the assistant is being used and how well it is performing.

Filters. Query conversations by date range (with auto-filled common periods), question text, user ID, name, or email, conversation ID, and whether the answer was rated helpful. A Field Definitions panel explains each filter.

Results. Three sections appear after submitting: collapsible Aggregate Statistics; a Data Distribution by Date graph with a Group By selector and downloadable image; and a Conversation Data table listing every conversation with timestamp, first question, question count, and user ID — sortable, expandable, and exportable with Download CSV.

What the data shows. Because the assistant is reachable through multiple channels, the user ID column reveals the mix: web-app usernames alongside WhatsApp IDs and SMS phone numbers. Recurring questions and unhelpful ratings are a direct roadmap for improving the document library.

Admin Console

The Admin console (User → Admin) is where the instance itself is configured — users, the AI model, and the assistant's standing instructions.

User management. Create new users or select an existing user to update their settings.

Model selection. Succinctly is vendor-agnostic: swap the underlying large language model from a dropdown without changing anything else.

Chat context. Control how much conversation context the model uses per chat — Max context by default, with smaller sizes available to trade recall for speed and cost.

System prompt. View, edit, or reset the assistant's instructions. This controls the persona, grounding rules (answer only from your documents), how conflicts between sources are handled, answer length and format, and privacy boundaries such as never revealing personal information. Editing the prompt changes behavior instantly, and Reset Default is always available as a safety net.

Quick-Start Checklist

From empty instance to working assistant in seven steps.

1. Upload your documents (Documents → Document Store) — drag in PDFs, Word files, spreadsheets, images, and media, or import from Google Drive or SharePoint.

2. Organize into folders and workspaces if you have distinct audiences or projects.

3. Set the system prompt (User → Admin) — define the assistant's persona, grounding rules, answer style, and boundaries.

4. Pick your model and context size (User → Admin).

5. Create users and share the URL.

6. Ask test questions in Chat, check the Source Files panel, and rate the answers.

7. Monitor Analytics — review what users ask, export the CSV, and fill gaps in the document library where answers were rated unhelpful.

Support

Succinctly is built by Austin Artificial Intelligence, Inc. We're happy to help.

Website: www.austinai.io

Email: support@austinai.io