Knowledge base chatbot with TechnologyOne: approved documents and cited answers
How a knowledge base chatbot can answer staff questions from approved policy documents alongside TechnologyOne, with sources and human review.
Updated 2026-03-07
Knowledge base chatbot with TechnologyOne
Local government offices, independent schools and charities on TechnologyOne tend to keep a second body of knowledge: delegations schedules, procurement procedures, grant conditions and the questions families, residents or donors ask each week. That material usually sits in policy folders and intranet pages, while the platform holds the transactions. A knowledge base chatbot joins the two for staff who need an answer quickly. It reads only the documents a business approves, answers in plain language and shows the section it drew on so the reader can check the wording. It does not read the live ledger and should not be relied on for figures. A finance officer checking an approval threshold gets the answer from the approved delegations document, while TechnologyOne remains the reference for any amount. MILabs, a Sydney agency, starts with a short list of documents owned by one named person, tests answers against them, and widens the set only after that owner is satisfied.
How it works
Name an owner for each document set, such as delegations, procurement procedures or grant conditions, and confirm which versions count as current.
Where the installed product and the business's permissions allow it, export reference tables or reports from TechnologyOne as files, and check a sample against the live screen before use.
Load the approved documents into a vector index in Pinecone, tagging each passage with its owner, document date and location so answers can point to the section used.
Use n8n to run ingestion on a set schedule and to hold any changed document for owner approval before the index is updated.
Send each question and its matching passages to the Anthropic API, which drafts an answer from those passages only, and show the contact form instead when no passage matches.
Review a sample of answers against the source documents on a regular schedule, and use the unanswered question log to decide where the document set needs work.
Frequently asked questions
- What does a knowledge base chatbot write to TechnologyOne?
- Nothing, in the standard build. The chatbot reads approved documents and answers questions, and it has no write path into TechnologyOne. Its question log is kept outside the ERP, in storage the business controls. A later handoff to a records or finance team would be a separate integration, planned only after the business has confirmed its API access and permissions with the vendor.
- What does a knowledge base chatbot with TechnologyOne need from me?
- A named owner for each document set, such as delegations, procurement procedures or grant conditions, and a decision on which versions count as current. You also need someone who can confirm what your installed TechnologyOne product and permissions allow for exports, since that determines whether any reference data can be included. Expect to review sample answers against the source documents before staff rely on the tool.
- What does a knowledge base chatbot not do with TechnologyOne?
- It does not calculate, reconcile or correct figures in the ERP, and it does not treat a written policy as a live transaction record. It cannot change a payment, a journal or a service request. Where a question depends on a current balance or a status in TechnologyOne, the chatbot should point to the system of record rather than estimate an answer. Staff should keep the platform as the reference for any amount.
- Which tools sit between a knowledge base chatbot and TechnologyOne?
- Pinecone holds the indexed passages from approved documents, n8n schedules the ingestion and routes changes through review, and the Anthropic API generates answers from the passages it is given. TechnologyOne sits at the edge of this chain. Only an export or a permitted API connection would feed it, and that depends on the product and agreement in place. No tool is placed between the chatbot and a live write to the platform.
- Is a knowledge base chatbot right for a small TechnologyOne team?
- Yes, when the team has a defined set of questions and a small body of approved documents. Small teams often gain most from answering routine questions about leave or grant conditions. It is less useful if the documents are scattered, out of date or owned by no one. Then the first task is tidying the source material, which MILabs can help scope before any chatbot is built.
Related
read these nextUp a level
More with TechnologyOne
- Council service request app with TechnologyOne
- Donor and grant report using TechnologyOne data
- Supplier spend report for TechnologyOne: spend by supplier, category and month
- AI adoption for TechnologyOne businesses
- AI strategy for TechnologyOne councils, schools and charities
- Custom development for TechnologyOne businesses
- Product idea for TechnologyOne: test and shape a first version
- Workflow automation for TechnologyOne businesses
