Many teams now work with ChatGPT, Claude or Copilot on a daily basis. A model like that starts from general knowledge. It knows the world in general, but not yet your organisation, your products, your procedures or your clients. As soon as a question is really about your own company, it gets stuck.
For company questions, the reverse order works better: first pull in your documents, then formulate an answer. That's what the abbreviation RAG stands for: retrieval augmented generation. In other words: retrieve first, then answer.
In this article you'll read what such a system solves, where it saves time and how to start small with it.
Why your company knowledge currently delivers too little
Many organisations recognise the same pattern. The knowledge sits in a SharePoint, on a shared drive, in a wiki, in mail archives and in the heads of a few experienced colleagues. Someone with a question first searches themselves, gives up, and then walks over to that one colleague who happens to know.
That costs time in three places. The back office answers the same questions over and over. New employees take a long time to get up to speed. And valuable documents such as policy papers, specifications and terms are barely used, simply because they're hard to search.
So the knowledge is already there. It just isn't used in a way that works.
What is a RAG system?
A RAG system is an AI that answers based on your own company knowledge instead of based on general information from the internet. The way it works is easiest to recognise in a new colleague who gets a question and first pulls in the right document before answering.
At the moment of the question, the system reads the relevant passages from your sources and uses them to give an answer that fits your situation. It also states which document the answer comes from, so you can always check it. Being able to verify an answer is, for many people, decisive in actually trusting and using the system.
How a RAG system gives your team time back
For your team, the work shifts. Less time goes to searching and repeating, and more to the work people were hired for.
Your team spends more time on:
- Giving customers and colleagues an accurate answer more quickly
- Assessing and deciding based on information found
- Work with substance and attention to the situation
Your team spends less time on:
- Digging through documents in search of a single paragraph
- Answering the same question for the tenth time
- Explaining everything to new colleagues step by step
What does a RAG system deliver?
The value becomes concrete fastest with recurring work: processing customer questions, looking up internal knowledge, onboarding new people and checking documents. The table below gives an indication of the time saved per month.
| Situation | Without AI support | With a RAG system | Indicative time saved per month |
|---|---|---|---|
| Frequently asked customer and colleague questions | Looking up and answering manually | Instant answer with source citation | 8 to 16 hours |
| Looking up knowledge in documents | Searching across multiple systems | One question in plain language | 6 to 12 hours |
| Onboarding new employees | Lots of questions to colleagues | Looking up answers themselves | 4 to 10 hours |
| Checking documents and terms | Manually finding the right passage | Targeted reference to the source | 3 to 8 hours |
The exact gain varies per organisation and depends on how many people use the system and how well the sources are prepared. The direction is always the same, though: time that now goes to searching is freed up for work with more value.
Common objections about a RAG system
"We're not ready for AI yet, our documents are scattered everywhere."
That's understandable, and it's rarely a blocker. A RAG system already works well with a few hundred accurate documents. You don't have to get your entire archive in order first. We start with the sources that answer the most questions and expand from there.
"How do I know whether an AI answer is correct?"
That's why the source citation is so important. The system shows which document and which passage the answer comes from, so your employee can check it themselves if in doubt. People remain in charge and the system supports their judgement. This keeps AI in your organisation manageable and explainable as it grows.
How we build a RAG system at DTT
Our background lies in custom software. So we're used to fitting a solution into the way your organisation works, and letting it grow over time. With a RAG system, much of the value lies in the preparation of your sources, in the connection with your existing processes and in the management afterwards.
Having advice and build in the hands of the same people saves coordination and keeps the solution coherent. We think along about the approach and build the system where your people already work: a chat window on the intranet, a button in your CRM or an environment for your customers. In doing so, we look ahead at workability, such as how often it will be used, who manages it and how it gets sharper as more questions come in.
Our approach focuses on:
- Practical applications that immediately save time in recurring work
- A solution that aligns with existing processes and systems
- A setup that stays maintainable and scalable as your organisation grows
Start small and build out sustainably
You don't have to start big. A first working version around one clear question or one important source already shows what it delivers, and gives you something to build on. From that first gain, you expand to more sources and more departments.
The most important thing: you can start where you are now. Even if you feel you're behind on digitisation or AI, that's no reason to wait. The first step is a short exploration in which we look together at what suits your situation.
Have a RAG system explored for your organisation
DTT stands for Doing Things Together, and that's how we approach a RAG system too: together with your people, from the knowledge you already have. Would you like to know what a RAG system can mean for your organisation? In an exploratory conversation, we map out together which knowledge sources you have, which questions your people ask most often and where the first gain lies. After that, you'll know whether it's worthwhile and what a small first step would be. Schedule an exploratory conversation.
Key terms at a glance
A few terms came up in this article. We list them here once more, clearly set out.
- RAG system — an AI that first retrieves information from your own sources and uses it to answer a question.
- Retrieval augmented generation — the full term behind RAG: retrieve first, then answer.
- AI knowledge base — your collected company knowledge, made searchable via a question in plain language. See also AI transformation is organising work more smartly.
- Source citation — the reference to the document an answer comes from, so you can verify it.
- AI readiness — the question of whether you're ready to start with AI; in practice, you can begin where you are now. See Is your organisation ready for Artificial Intelligence?.
- Choosing an AI route — the trade-off between an off-the-shelf tool, integration or custom development. See AI implementation: buy, integrate or develop bespoke?.
