Chatbots and Chat Assistants · Ankara

If it doesn't know the answer, it should say so.Chatbots and Chat Assistants

We build chat assistants for websites, WhatsApp and internal systems: assistants that work on the organization's own documents, show their sources and hand over to a person when needed.

WebsiteWhatsAppInternal useDocument-based answersHand-over to peopleConversation log

The most common complaint in chatbot projects is the bot giving wrong answers with confidence. Once a user is misdirected, they stop using it — and the support load stays where it was.

That's why the assistants we build work on the organization's own documents, show where an answer comes from and hand the topic to a person when unsure. When the hand-over point isn't designed from the start, the bot gets switched off after a while.

If this sounds familiar

  • The support team spends a large part of the day answering the same questions.
  • Questions arriving on the website outside office hours go unanswered.
  • The bot you tried either said "I didn't understand" to everything or made answers up.
  • Employees ask each other where to find procedures and documents.

What it covers

Working with company documentsAnswers based on product documentation, procedures and FAQ content — not general knowledge.
Showing sourcesShowing which document an answer came from. Answers that can't be verified don't earn trust.
Hand-over to a personUncertain or sensitive topics are transferred to an agent, together with the conversation history.
Channel supportWebsite widget, WhatsApp and internal use.
Conversation logsWhich questions come in and where they go unanswered. Content gaps emerge from here.
Permission awarenessIn internal use, documents the user isn't allowed to see never make it into an answer.

Example scenarios

Illustrative examples; the specifics are worked out during discovery at your site.

WebsiteFirst answer to product and service questionsThe assistant answers from product documentation and FAQs, and in a sales-relevant conversation takes contact details and passes them to the team.
Customer serviceStatus queries and hand-over to a personOrder or application status is read from the source system; conversations involving complaints or exceptions pass to an agent along with their history.
InternalHR and quality proceduresAn employee asks about leave, expenses or a quality procedure; the assistant answers only from documents the user is authorised to see, showing its source.

What we deliver

  • Question inventory drawn from real questions
  • Content set the answers rely on and how it is kept up to date
  • Assistant working on the chosen channels (web, WhatsApp, internal)
  • Hand-over rules and agent screen
  • Report of unanswered questions

How we proceed

1 · Question inventoryThe real questions coming in today are collected; the bot is built around them.
2 · Content preparationCompiling the documents answers will rely on and verifying they're up to date.
3 · Limited launchStarting with a specific topic or user group.
4 · ImprovementUnanswered questions are closed as content gaps.

When it's the right choice

  • Volume of repeated questions is high and the answers exist in written documents.
  • Someone is responsible for keeping the documents up to date.
  • You accept that the bot hands over to a person where it doesn't know.

When it's not the right choice

  • If the answers aren't written down and live only in a few people's heads — content comes first.
  • If every conversation is expected to be carried through to the end by the bot.
  • If legally or medically binding advice is expected.

Frequently asked

What happens if the bot gives wrong information?

The assistant is set up not to go beyond the organization's documents and shows the source of its answer. When unsure, it hands the topic to a person instead of producing an answer; this behaviour is defined from the start.

Does it work on WhatsApp?

Yes. It can be used through a WhatsApp business line; setup requires an official business account and an approval process.

How are permissions managed in internal use?

Documents the user isn't allowed to see never make it into an answer. The permission model follows the source system's permissions, not the assistant's.

Let's talk about this.

We'll listen to your current situation and work out together where to start.

Request a Call →Other solutions