Knowledge Base

The Knowledge Base is the layer of contextual understanding that makes the AI Copilot relevant to your specific business. Rather than working from generic knowledge alone, the Copilot draws on patterns, terminology, and facts specific to your operation.

What gets captured

FlexotiumPOS automatically captures:

| Category | Examples | |----------|---------| | Product context | Top sellers, seasonal patterns, typical order quantities | | Supplier patterns | Usual lead times, delivery reliability, typical invoice amounts | | Operational rhythms | Busiest hours/days, shift lengths, average transaction size | | Staff patterns | Shift preferences, typical sales performance | | Business terminology | Your category names, SKU conventions, warehouse names |

How to search the knowledge base

Go to AI & Automation → Knowledge → Browse. Use the search bar to find any captured fact. Knowledge is organised into cards by category (Products, Suppliers, Operations, Staff).

You can pin knowledge cards that you want the Copilot to weight more heavily in its responses.

How knowledge informs the Copilot

When you ask the Copilot a question, it retrieves relevant knowledge cards automatically before formulating the response. This allows it to answer questions like:

  • Which is my best-selling product on Friday evenings? — answered using operational rhythm knowledge
  • When should I reorder from Supplier X? — answered using supplier lead time knowledge + current stock

The Copilot displays which knowledge cards it used in the Sources section of its response, so you can verify the reasoning.

Adding manual knowledge

You can add facts manually that the system might not detect automatically:

  1. Go to Knowledge → Add Entry
  2. Choose a category
  3. Write the fact in plain English: Our main delivery day is Tuesday. Orders must be placed by Sunday 17:00.
  4. Save

Manual entries are prioritised over inferred entries.

Privacy and data isolation

Knowledge is strictly scoped to your tenant workspace. It is never shared between tenants, never used to train the underlying AI model, and is deleted within 30 days of account closure. See the Privacy Policy for full details.