AI Assistant
An assistant that knows your shop
A built-in chat that’s aware of your documents, BOMs, change requests, named workflows, and the Thoma user manual. Ask "how do I check out a drawing?" or "show me released changes from last week" and get a real answer in seconds.
What it does
Trains your team faster
New users land in Thoma and immediately have a conversational guide. "How do I create a change request?" "What does this status mean?" "Where is the audit log?" — answered against the live user manual, not generic LLM hallucinations.
Answers from your data
The assistant gets a recent-context snapshot of your top documents, changes, projects, tasks, users, and named workflows on every question. So "what changes are blocking project Aurora?" is a real query with a real answer, not a search-the-UI exercise.
Help-corpus backed
Questions about Thoma itself trigger keyword-matched injection from a curated help-corpus (10+ articles covering common workflows). You get the actual procedure, not a guess.
Configurable tone
Admins pick Conversational (default — friendly, walks the user through) or Reference (terse, command-and- response) per organization. One corpus, two voices, your call.
Off by default, on per org
Disabled out of the box. An admin enables it via System Preferences with a privacy disclosure modal that explains exactly what gets sent to the AI provider. No surprises.
Fully audited
Admins read every chat session in the org via the 🤖 AI Chat Oversight admin tab — paginated session list, click-to-drill into the full message thread, and per-user token / cost rollups over the last 7 / 30 / 90 days. The assistant can be disabled for the whole organization or a specific user at any time.
Common use cases
Onboarding a new employee
New hire sits down with Thoma open. Instead of paging through a 200-page PDF, they ask the assistant: "How do I check out a drawing?" "What's the difference between Released and Obsolete?" "Where do I attach a file to a change request?" The answers route from the actual help corpus + their live data.
Quick lookup mid-task
Mid-flow you need to find something. "What changes is project Aurora blocked on?" "Show me documents owned by Sarah that are still in Draft." "What's the status of Engineering Change Request 178?" — answered against your live data without leaving the screen you’re on.
Process refresher
You haven’t done a particular workflow in months — say, configuring a new document category or setting up an electronic-signature ceremony. The assistant walks you through it, step by step, from the user manual. No "let me find that page" hunting.
Helping vendors and contractors
External users with viewer seats can use the assistant to navigate your documentation conventions without you fielding every question. "Where do I find the latest revision of part 4451?" "How do I download a STEP file?" — handled.
Not available on the Enterprise tier
Enterprise customers run Thoma in environments that need to satisfy CMMC Level 2 requirements for handling Controlled Unclassified Information (CUI). Sending data to a third-party LLM (in this case Anthropic Claude) is incompatible with the data-handling restrictions that CMMC L2 places on CUI environments — even at the per-org opt-in level.
For that reason the AI Assistant is included on the Standard and Professional tiers (where admins choose to enable it), and is not enabled on Enterprise regardless of admin preference. Customers who don’t handle CUI but want enterprise-tier features (MFA-required, AD/SSO, advanced audit, SOC 2 / CMMC documentation) without the AI restriction should contact Sales — we can accommodate hybrid configurations on a per-account basis.
See the tier comparison for the full feature-by-tier breakdown.
Privacy + data handling
Where the data goes
When the assistant is enabled, your question, the recent-context snapshot, and any matched help-corpus articles are sent to Anthropic (Claude API) over TLS. Anthropic’s standard API terms apply — notably no training on customer API data. Responses come back over the same channel and are stored in your per-organization chat history.
What gets sent
Recent-context snapshot is the top N most-recently-touched documents, changes, projects, tasks, and users (by name + key fields, not file contents). Matched help-corpus articles are vendor-authored material. Chat history persists per organization for admin review.
Quotas + cost protection
Vendor-paid model with per-user-daily and per-organization-monthly quotas. Exceeding quota disables further requests with a clear message. Quotas are configurable per-license; defaults are sized for typical engineering-team usage without surprise costs.
How to disable it
Admins toggle the feature off in System Preferences. The chat panel disappears for all users immediately. Pre-existing chat history is preserved for audit but no new sessions can be opened.
Want it on for your team?
The AI Assistant is included on the Standard and Professional tiers and switches on with one toggle in System Preferences after install. Start a trial to set it up for your shop, or talk to Sales if you want to walk through the privacy + quota model first.
Note: the AI Assistant is intentionally disabled in our public demo sandbox to prevent anonymous-user abuse. Trial installations have it available with the standard per-org toggle.