← All case studies Demo project

AI knowledge assistant

Your team asks a question in plain words, and the assistant answers from your own documents: policies, price lists, manuals and past proposals. Every answer shows where it came from.

This is a demo, not a client project. It plays a fixed script in your browser, for an invented company with four invented documents. The real version uses an AI model that searches the business's own files.

Try the demo

Pick a question to watch the assistant find the answer.

Cedar Home · company documents

Leave policy

2. Annual leave. Staff get 15 working days in their first year. This rises to 18 days after three years.

4. Sick leave. Up to 10 days a year, with a doctor's note from the third day.

Price list · November

Ceramic mug set. $12 per set. For 500 sets or more: $9.50 per set.

Delivery. Free inside Beirut. Outside Beirut: $60 flat per order.

Returns and refunds manual

3. Opened items. Can be returned within 14 days if not damaged, for store credit only.

5. Faulty items. Full refund within 30 days, with the receipt.

Proposal · Al Noor Café · March 2026

Scope and price. 300 branded mugs and 40 gift boxes, $3,450 in total, delivered in three weeks.

How much leave do I get?

A new hire asks, and gets the policy, not a guess.

  1. Rana, new in the team

    How many days of annual leave do I get in my first year?

  2. Searching the company documents

    Best match: Leave policy.

  3. Reading the passage

    Section 2, Annual leave.

  4. Knowledge assistant

    You get 15 working days in your first year. It rises to 18 days after three years. Source: Leave policy, section 2.

  5. Answered in seconds with the source, without waiting for HR.

A price for a big order

Two parts of the price list, added up.

  1. Karim, in sales, on a call

    What is our price for 500 mug sets delivered to Tripoli?

  2. Searching the company documents

    Best match: Price list, November.

  3. Reading the passages

    The bulk price, and delivery outside Beirut.

  4. Knowledge assistant

    500 sets at the bulk price of $9.50 is $4,750. Delivery outside Beirut is a flat $60, so $4,810 in total. Source: Price list, November.

  5. Karim gives the quote on the call, not the next day.

Can a customer return this?

The rule, with its conditions.

  1. Maya, at the shop counter

    A customer wants to return an opened item after 10 days. Can they?

  2. Searching the company documents

    Best match: Returns and refunds manual.

  3. Reading the passage

    Section 3, Opened items.

  4. Knowledge assistant

    Yes, if the item is not damaged. Opened items can be returned within 14 days, for store credit only, not cash. Source: Returns manual, section 3.

  5. Every shop gives the same answer, the one in the manual.

What did we charge last time?

Past work, found again.

  1. Karim, in sales

    What did we charge the last café for branded mugs?

  2. Searching the company documents

    Best match: Proposal, Al Noor Café.

  3. Reading the passage

    Scope and price.

  4. Knowledge assistant

    Al Noor Café, March 2026: 300 branded mugs and 40 gift boxes for $3,450 in total, delivered in three weeks. Source: Proposal, Al Noor Café.

  5. The new quote starts from the last one, not from a blank page.

A question it can’t answer

It says so, and never invents an answer.

  1. Maya, at the shop counter

    What is the Wi-Fi password for the Saida shop?

  2. Searching the company documents

    No document mentions it.

  3. Knowledge assistant

    I couldn’t find that in the company documents, so I won’t guess. Hadi in IT can help. I have noted the question so the answer can be added.

  4. No invented answer. The gap is logged for whoever looks after the documents.

01 · The problem

The answer exists, but nobody can find it

Every company has the same questions asked again and again: how much leave do I get, what is the price for a large order, can this be returned. The answers sit in PDFs, spreadsheets and old emails that only a few people know how to find.

So staff interrupt the one person who knows, or they guess. New hires take weeks to stop asking, and two branches give two different answers to the same customer.

02 · The approach

How a real one is built, in four steps

  • Collect the documents that matter. We choose which files the assistant may read: policies, price lists, manuals, proposals. Anything private stays out.
  • Answer only from those documents. The assistant finds the passage first, then writes the answer from it, and names the document and section.
  • Say “I don't know”. If nothing matches, it says so and points to a person. It is told never to fill the gap with a guess.
  • Put it where the team already works. WhatsApp, Slack, Telegram or a simple web page. Unanswered questions are logged, so you can see which documents are missing.

03 · What it could improve

The goals for a business that uses one

  • Fewer interruptions for the people who know the answers.
  • The same answer everywhere, because it comes from one document, not from memory.
  • New hires find their feet sooner, since they can ask anything, at any hour.
  • Old knowledge stays useful: past proposals and prices are found in seconds.

These are goals for a real setup, not measured results from this demo.

04 · Where else it fits

Teams it could be adapted to

  • Retail and distribution: prices, stock rules, returns and delivery terms.
  • Clinics: procedures, preparation instructions and insurance rules, for the front desk.
  • Real-estate offices: listing details, commission rules and contract steps.
  • Agencies and consultancies: past proposals, rate cards and how-we-work guides.
  • Any growing team: HR policies and onboarding answers.

Want one for your team?

In a free 30-minute call we look at the questions your team asks most, and which documents hold the answers.

Book a free 1:1 call →