/about any SOP
Your team types a question and gets a short answer drawn from your own files, with a link to the exact page it came from. People only see results from folders they are allowed to open.
↓It finds the right passages in your files first, skips what this person may not see, then answers with page references.
Knowledge search (RAG),explained without jargon.
What is retrieval-augmented generation (RAG)?
One idea at a time. Each sheet lays over the last, so the picture builds up. Tap a number to jump.
It works like a librarian who fetches the three right pages from the shelf before answering, and shows you the pages.
RAG is a way of making an AI model answer from a specific set of documents instead of from general internet knowledge.
When someone asks a question, the system first searches your files for the most relevant passages.
It then gives only those passages to the AI and asks it to answer using them, with a reference to each source.
If nothing relevant is found, it says so.
Picture a business like yours,in three frames.
A typical situation, not a client story. The line in each frame goes from tangled to straight as the work is done.
A Pune pump manufacturer helping field engineers find the right fix
Heard yourself say this?Count how many fit.
Five things owners tell us before a project like this starts. Every one you recognise adds a tally mark.
Tap the lines you have said this year.
One fits. Worth a short call to see if Knowledge search (RAG) is the right fix.
Two fit. A short call will tell you whether to do this now or later.
Three fit. This is probably worth planning soon.
Four fit. You are losing time to this already.
All five. This page was written for you.
Six parts,all handed over to you.
Your Knowledge search (RAG) work is made of these six parts. The comb lights each one in turn; tap a cell to read it.
Document connectors
Scheduled syncs from Google Drive, SharePoint, Confluence, Notion or a file server, so new and edited files are picked up without manual uploads.
Cleaned and chunked index
Your files are converted to text, split into sensible sections with their headings, and stored in a vector database such as pgvector or Qdrant.
Answers with citations
Each answer lists the document, section and page it used, and a click opens the original so staff can verify it.
Permission-aware results
Search respects the access rules of each source folder, so an HR policy draft never shows up for a sales executive.
Evaluation question set
A list of real questions with correct answers, which we run after every change to catch drops in accuracy.
Usage and gap report
A dashboard showing what people search for, which questions found no good source, and which documents are never used.
- Document connectors
- Cleaned and chunked index
- Answers with citations
- Permission-aware results
- Evaluation question set
- Usage and gap report
The question we hear most:Off-the-shelf workplace AI or a custom RAG system?
The board turns through each option. Tap one to hold it and read the trade-off.
Your files already live in Google Workspace or Microsoft 365 and the assistant included in your licence searches them well enough.
You have limited control over how answers are formed, which sources are included, or how it handles scanned PDFs and tables.
You need citations to exact pages, mixed sources, regional languages, or the search embedded in your own portal or app.
It takes several weeks to build and needs someone to keep the index and test questions current.
Four legs of one relay,about 6 weeks, start to finish.
Each leg is sized by how long it usually takes. Watch the baton go round, or tap a leg to jump there.
Four things that trip a project,and the fix we fit instead.
Each fuse is a mistake we see often. Watch it blow and get replaced, or tap a fuse to read the fix.
Indexing every version of every file
We mark one version as current and exclude drafts, so answers do not mix old and new rules.
Skipping tables and scanned pages
Specs often live in tables and scans, so we use layout-aware extraction and OCR for those.
Judging accuracy by a few demo questions
A written test set with known answers gives a fair measure every time something changes.
Ignoring who is allowed to see what
Permissions are designed in from the first week, because adding them later means rebuilding the index.
Four factors,four corners of your project.
Each corner grows when that factor is on the larger side. The outline is your project's footprint, and the quote follows it.
A compact project. The smallest footprint for your quote.
Mostly compact, with one area that needs more work.
A balanced build. Half the factors on each side.
A larger build. Most areas need extra work and time.
The full footprint. Every factor on the larger side.
Knowledge search (RAG) questions,answered plainly.
Search the answers. If yours is missing, ask us on the call.
Cost & timeHow long does it take and what affects the price?
A pilot over one well-organised folder can be ready in about three weeks. Large or messy collections with scans, many sources and strict permissions take longer. The quote depends mainly on document volume, number of sources and access rules.
Getting startedWill our documents be used to train public AI models?
We use business API terms from providers that state API data is not used for training by default. For sensitive material, we can also run open models on a server you control. We show you the relevant provider terms before you decide.
Getting startedCan it search scanned PDFs and handwritten notes?
Printed scans work well after OCR. Handwriting is less reliable, so we test a sample first and tell you honestly how usable it is. Poor scans may need to be rescanned.
In IndiaDoes it work with Hindi or mixed-language documents?
Yes. We choose embedding models that handle Indian languages, and a question in English can find a passage written in Hindi. We test both directions with your own documents.
Getting startedWhere is the data stored?
The index can sit in a database in an Indian cloud region such as AWS Mumbai or in your own server. Original files stay where they are, and the system only keeps the text it needs to search.
OwnershipWho owns the system?
You own the code, the index and the cloud accounts. The setup is documented so another developer can maintain it if you choose.
Getting startedWhat happens when a policy changes?
Connectors pick up edited files on the next sync and the old text is replaced. If the old version is kept in the same folder, we help you archive it so it stops appearing in answers.
Getting startedHow do we know answers stay accurate over time?
The test question set is run after every update to the index, model or prompts. On a monthly plan we also review flagged answers and searches that found nothing, and report what we fixed.
Works well alongside,from the same practice.
Stop searching folders and start asking
Share a sample folder and we will show you answers with citations from your own files.