NX Knowledge Chatbot
NX Knowledge Chatbot
Your best answers already exist. They’re just distributed across three PDFs, one wiki, and two colleagues on vacation.
The NX Knowledge Chatbot answers questions from your own material — design guidelines, works standards, past projects, NX know-how, CAD and Teamcenter metadata — and cites the source of every answer. It is read-only by design: it searches, explains and points, and can never touch a model or a PLM record. Onboarding questions, knowledge search, “I’m stuck in NX” moments and routine CAD-support tickets get answered in seconds, with the citation that makes the answer checkable.
It is the answer tier of our NX AI Assistants & Chatbots family — the grounded, in-production counterpart to the action assistants that execute.
Who this is for
- New designers spend their first months asking the same forty questions — and seniors spend hours answering them.
- Your standards live in documents nobody searches under deadline — so tribal memory decides instead.
- “How do we usually do this?” is answered by whoever happens to be reachable.
- CAD support tickets repeat: settings, licenses, template questions — deflectable, if answers were findable.
- You want AI grounded in your knowledge — not a general chatbot’s best guess about it.
// A new designer’s Tuesday, answered in seconds
designer> Which thread inserts do we use for aluminium housings, and is there a minimum edge distance?
For AlSi9Cu3 housings we use wire thread inserts per WN-311, standard series M4–M10. Minimum edge distance is 2×D from the bore axis; below that, a manufacturing sign-off is required. Two released projects deviate — both documented with a waiver.
- WN-311 §3.1 · rev D
- KR-041 §6.4
- project 4471 · waiver W-88
read-only · nothing touched · answer checkable in one click
Citation-first, by policy. If no source in your material supports an answer, the chatbot says so and points to the closest thing it found — it does not improvise. That single rule is what makes engineers trust it after the second week.
The four jobs it does every day
The first-year question machine
New designers ask freely and get guideline-true answers with sources — seniors get their afternoons back, and the standard gets learned as written, not as remembered.
“How do we usually do this?”
Design decisions, past solutions, comparable projects — found by meaning, not filename, with the project reference attached.
Help at the moment of confusion
NX functions, settings and your house-specific workflows explained in context — the colleague who always knows, minus the interruption.
CAD support, self-served
The repeating template, license and settings tickets answered before they become tickets — support capacity back for the real problems.
Read-only is a guarantee, not a setting
The knowledge chatbot has no write path: it holds no NX session rights and no PLM write permissions — searching and answering is technically all it can do. It runs on your infrastructure, on-premises or in an EU cloud under your control, GDPR-compliant with no data retention. When an answer should become an action, that is a deliberate handover to the act tier — the NX Action Assistant — with its own preview-and-confirm gate.
Your material only
Guidelines, standards, wikis, project docs, CAD & Teamcenter metadata — indexed under your access rules.
Cited or declined
Every claim carries its source; without a supporting source, the chatbot says “not in your material”.
No write path
No model access, no PLM writes — read-only enforced by architecture, not by promise.
Honest limits
Answers are only as good as the material. Where your documentation is wrong or outdated, the chatbot will faithfully cite the wrong document — which, in practice, is how customers find outdated standards faster than any review cycle. The pilot indexes your real corpus and measures answer quality on questions your team actually asks, so you see the gaps before rollout. And it does not replace training or the senior review: it answers what is written; judgement about what should be written stays human.
Fixed price — fixed scope
Knowledge Chatbot Pilot
€14,500 fixed
- Your core corpus indexed — guidelines, standards, one project archive
- Citation-first chatbot live for a pilot group, on your infrastructure
- Answer-quality report on your team’s real questions — measured, not claimed
- Access rules respected — who may read a source, may read its answers
- Rollout blueprint: more sources, more teams, the act-tier handover
4–6 weeks · price creditable against rollout · GDPR, no data retention.
For comparison
Knowledge by interruption
hours weekly of senior time
- The same forty questions, answered live, by your most expensive people
- Standards applied from memory — each memory slightly different
- Knowledge leaving the company one retirement at a time
Not every question is about NX know-how. When the question is which past projects or parts match a new enquiry, across PDM, ERP, wiki and file servers, see our engineering data AI services.
Frequently asked questions
What is an NX knowledge chatbot?
How do you prevent invented answers?
Which sources can it use?
Is it really unable to change anything?
Where does it run, and what about GDPR?
Our documentation is partly outdated. Deal-breaker?
Related in the family: when answers should become actions, the NX Action Assistant takes over with preview and confirm; regulated industries find their audit-trail variants under the industry knowledge assistants. Overview: NX AI Assistants & Chatbots.
Write down the three questions your new designers always ask. We’ll answer them from your own standards.
Send us the three questions plus the documents that should answer them — we’ll show the chatbot answering with citations, on your material, before you commit to anything.
Talk to an EngineerPilot €14,500 fixed · creditable against rollout · on-premises option · GDPR, no data retention