Zero upfront. You pay when it works. Book a free call →

RAG ยท Knowledge assistants

RAG development services

A knowledge assistant that answers from your own documents, shows where each answer came from, and says so when it does not know.

We build retrieval-augmented generation (RAG) assistants on your policies, manuals, contracts, product data and past work. Answers come with sources, access follows your existing permissions, every question is logged, and your documents are never used to train a model.

Reads fromZoho WorkDriveSharePoint or Google DriveZoho Desk help centreZoho CRM recordsPDFs and manualsYour intranet or wiki

Today

Why your team still cannot find the answer

The answer is in a manual, a policy, a past proposal or an old email thread. Finding it means asking the one person who knows, or searching five systems. New hires ask the same questions for months. A general chatbot does not know your business, and guesses when it does not.

With ICIGAI

Sourcedevery answer linked to the document it came from
Samepermissions as your files: people only see what they already can
"I don't know"when the answer is not in your documents

What we commit to in every RAG quote.

Why ICIGAI

What many RAG builds leave out.

01

Permissions that hold

The assistant respects who can see what. A sales rep never gets an answer drawn from an HR file they cannot open.

02

Answers you can check

Every answer shows its sources, so people can click through and confirm before they act on it.

03

Connected to Zoho and your tools

We pull from Zoho WorkDrive, Desk and CRM as well as SharePoint, Google Drive and file shares, and keep the index up to date.

04

Privacy and a free quote

Set up to support PIPEDA and Quebec Law 25, with no training on your data, scope and price in writing up front, and a named contact who replies within one business day.

How we build it

Six steps to an assistant that knows your business.

  1. 01

    Gather the sources

    We agree which documents and systems the assistant should read, which it must not, and who will use it.

  2. 02

    Clean and split

    Old versions, duplicates and drafts are removed or marked. Documents are split into sections the assistant can find and quote.

  3. 03

    Index with permissions

    Each section is indexed with who is allowed to see it, so searches only return what the person asking may read.

  4. 04

    Retrieve and answer

    When someone asks, the assistant finds the most relevant sections and writes an answer using only them.

  5. 05

    Show the sources

    Answers link back to the exact document and section. If nothing relevant is found, it says so instead of guessing.

  6. 06

    Test and improve

    We test on a set of real questions with known answers, measure how many it gets right, and keep tuning after launch.

Where RAG goes next

Knowledge that powers other agents.

The same knowledge base can help agents that answer customers or prepare work, always with a person approving.

Service desk replies

Drafts answers to customer tickets from your approved help content, for an agent to check and send in Zoho Desk.

See the service agent →

Quotes from past work

Uses past quotes and price lists to draft new quotes from an enquiry, ready for a person to review.

See the RFQ agent →

HR questions

Answers staff questions about policies and benefits from your handbook, and passes sensitive ones to HR.

See the HR agent →

RAG guide

What RAG is, in plain words

Retrieval-augmented generation means the AI looks things up before it answers. When someone asks a question, the system searches your documents for the most relevant passages, gives those passages to the AI model, and asks it to answer using only them.

That matters because a general AI model does not know your products, prices or policies. RAG gives it your information at the moment of the question, without training a model on your data.

It also keeps answers current. When a policy changes, the index is refreshed and the next answer uses the new version, with no retraining.

What a knowledge assistant can do

  • Answer staff questions from policies, SOPs and manuals
  • Help support teams find the right fix from past tickets and help articles
  • Help sales find past proposals, case notes and product details
  • Answer questions about contracts and supplier terms, with the clause shown
  • Help new hires get up to speed without interrupting senior staff
  • Summarise a long document and point to the parts that matter

Build custom, or use a built-in connector?

ChatGPT, Claude, Copilot and Gemini business plans now connect to common file stores, and for simple cases that may be enough. We will tell you if it is. A custom build is worth it when you need strict permissions across several systems, sources that the built-in connectors do not reach, a specific answer format, a measured accuracy level, or the assistant inside another tool such as Zoho Desk or your website.

Accuracy you can measure

We do not judge a RAG system by how it feels in a demo. Before launch we build a set of real questions with known answers, and measure how often the assistant finds the right source and answers correctly. You see that number, and we track it after launch as documents change.

  1. Collect real questions from the people who will use it
  2. Write the correct answer and source for each
  3. Run the assistant and score every answer
  4. Fix the weak spots: missing documents, bad splits, unclear instructions
  5. Repeat until it meets the agreed level

Guardrails and privacy

  • Least access: the assistant only searches what the person asking is allowed to see
  • Logs: every question, the sources used and the answer are recorded
  • Approvals: answers that go to customers are drafts until a person sends them
  • PIPEDA and Quebec Law 25: personal information is mapped, sensitive files can be left out, and data stays in the region you choose
  • No training on your data: model providers and settings are chosen so your documents are not used for training
  • Hosting options in Canada where the data needs it

When not to use RAG

  • Your documents are out of date or contradict each other: fix the source first
  • The answers you need are numbers from a database: use reporting or an agent that queries the data
  • A few dozen questions cover everything: a good FAQ page may do
  • A built-in connector in your current AI plan already does the job

We will tell you on the free call if a simpler option fits.

What you get

An assistant your team can trust.

  • A source list agreed with you, including what to leave out
  • Document cleanup and sensible splitting
  • An index that follows your existing permissions
  • Answers with links to their sources
  • A test set of real questions and an accuracy score
  • Logs of every question and answer
  • A data-flow diagram for your privacy records
  • Index updates and tuning after launch

Free custom quote

Know the price before we start.

  1. Free call30 minutes on what you run today and what you need. No sales script.
  2. Free quoteA custom quote with scope, timeline and price in writing.
  3. Build and go liveWe set it up, move the data and test it with your team before switch-over.
  4. AfterOptional care plan: fixes, changes and a monthly check-in.
Get a free RAG quote →

Questions

Straight answers.

What are RAG development services?

They cover building an AI assistant that searches your own documents and answers from them, with sources shown. RAG stands for retrieval-augmented generation.

Is our data used to train the model?

No. RAG looks up your documents at the moment of the question. We choose providers and settings that do not train on your data.

Can it respect who can see which documents?

Yes. Permissions are built into the index, so people only get answers from documents they can already open.

What if the answer is not in our documents?

The assistant says it could not find it, and can pass the question to a person. It does not make up an answer.

Can it work with Zoho?

Yes. It can read from Zoho WorkDrive, Desk and CRM, and the answers can appear inside Zoho Desk or other tools.

Can the data stay in Canada?

Where needed, yes. We use Canadian hosting options and document where each part of the system runs.

How much does RAG development cost?

It depends on your sources, permissions and where the assistant will live. Book a call and you get a free custom quote in writing.

Let’s talk

Answers from your documents,
with the source shown.

Tell us a little about it. A real person from our team will reply within one business day.

  1. 1Send the form. Two minutes, no sales script.
  2. 2We reply within one business day with times for a 30-minute call.
  3. 3You get a free custom quote. Scope, timeline and price in writing, yours to keep either way.
I’m interested in Services Tools