AI assistants
An assistant that answers from your own documents
We build assistants for staff or customers that look up the answer in your documents and systems, show where it came from and pass the question to a person when the sources do not cover it.
Internal assistant
What is our return period for custom-made items?
Custom-made items can be returned only if they arrive damaged or differ from the confirmed order. The customer has to report it within the period stated in the terms of sale, section 6.
What it is
Search that reads the documents and answers the question
A general chatbot answers from what it learned in training. An assistant built for your company first retrieves the relevant passages from your own material and then answers from those, with a link to the source. That is what makes the answer checkable. It is worth building when the same questions are asked again and again and the answers already exist somewhere in writing.
- Answers come with the source, so a person can check them
- When the sources do not cover the question, it says so and hands over
- Access follows your permissions: people see only what they may see
- Works inside the tools you use: team chat, a web page, the helpdesk
What the assistant can read
- Terms of sale and returns procedureDocs
- Product sheets and manualsDrive
- Customer and order statusCRM
- Resolved support ticketsHelp
What we deliver
An assistant, and the plumbing that keeps it right
Knowledge base
Your documents collected, split and indexed so the right passage is found, and refreshed when the originals change.
Live lookups
Read-only connections to the CRM, ERP or order system for questions whose answer changes by the hour.
The place people ask
A chat window on your site, a bot in Slack or Teams, or a drafting panel inside the helpdesk.
Permissions
The assistant respects who is asking. Internal price lists do not appear in a customer chat.
Handover to a person
A clear path to a colleague when the question is out of scope, sensitive or simply not answered by the sources.
Review and improvement
A log of questions, answers and feedback, and a regular look at what it could not answer and why.
How a project runs
Start with one group of questions
- 01
Collect the questions
We gather real questions from the inbox, the helpdesk or the team, and check that written answers exist.
- 02
Prepare the sources
Documents are tidied, duplicates and outdated versions removed, and the index is built.
- 03
Test against real questions
The assistant answers the collected questions and your people mark each answer right, wrong or incomplete.
- 04
Release to a small group
It goes to a few users first. We widen access once the answers hold up and the handover works.
What we build with
- Models
- Claude models, GPT models
- Search and data
- PostgreSQL with pgvector, Document parsing, OCR for scans
- Where it lives
- Website chat, Slack, Microsoft Teams, Helpdesk, Your internal app
We start from what you already have. The list shows what we reach for when something is missing, not tools you must adopt.
Is an assistant worth building?
A good fit when…
- The same questions arrive every week and the answers are written down
- New staff spend their first weeks asking where things are
- Support answers are found by searching old tickets
- Knowledge is spread across folders, wikis and inboxes
Probably not for you when…
- The knowledge exists only in people’s heads
- Documents contradict each other and nobody owns them
- A wrong answer could cause harm and no person would review it
- Only a few questions a month would be asked
Use cases
Where assistants are used
Typical places an assistant earns its keep.
Every enquiry in the CRM, without retyping
Enquiries from forms, email and chat become contacts and deals in your CRM, with the right person notified.
Open use case Finance and adminAI integrationsBusiness automationSupplier invoices read, matched and ready to post
Incoming invoices are read, matched to purchase orders and sent for approval before they reach the books.
Open use case Finance and adminAI integrationsBusiness automationData out of documents, into your systems
Contracts, forms, delivery notes and statements are read and their data placed where it is needed.
Open use case Finance and adminBusiness automationAI integrationsThe monthly report, assembled for you
Figures from accounting, sales and operations pulled into one report on a schedule, with a written summary.
Open use case Customer supportAI integrationsBusiness automationA support inbox that sorts itself
Incoming messages are sorted by topic and urgency, routed to the right person and given a draft reply.
Open use case Customer supportAI integrationsCustom softwareAn assistant that answers from your documents
Staff or customers ask in plain language and get an answer drawn from your documents, with the source shown.
Open use case Customer supportBusiness automationAI integrationsCustomers check their order without writing to you
Order and delivery status available on your site or by reply, pulled live from the shop and the courier.
Open use caseQuestions about AI assistants
Will it make things up?
We build it to answer only from the passages it retrieved and to say when they do not cover the question. Every answer shows its source, so a person can check. That reduces invented answers a great deal, but it does not make them impossible, which is why anything with consequences still goes past a person.
Is our data used to train the model?
We use the model providers’ business API access, under terms that state that data sent through it is not used to train their models. We show you the terms of the provider we propose before any of your data is sent.
Where are our documents stored?
The index is stored in a database in your own cloud account or on your server, in a region you choose. The model receives only the passages needed to answer one question.
What happens when a document changes?
The index is refreshed on a schedule or when a file changes, so the assistant answers from the current version. Removed documents are removed from the index too.
Can it take actions, or only answer?
An assistant looks things up and answers. When it should also do something, such as open a ticket or update a record, that is an agent with defined permissions and approval points.
Does it work in Croatian?
Yes. Current models read and write Croatian well, and an assistant can answer in the language of the question while the sources are in another.
Show us one process. We will show you what it could look like.
The analysis is free, takes about an hour of your time, and ends with a written summary you keep.
Prefer to read first? See examples