We pick the cases worth taking over
We analyse the most common questions, the current workflow, your systems and the moments customers give up or need manual help.
Tool we made
An AI chatbot that doesn't stop at answering
We build chatbots that answer from your company's own knowledge and walk each visitor to a concrete next step: a booking, an enquiry, a payment or a conversation with the right person. On your website, around the clock.
01 — First contact on your website
A customer lands on your website with a question. The chatbot walks them to a resolved case.
A website shows everyone the same content, and a form expects the customer to already know what they need. So your team answers the same questions over and over before the real issue even surfaces. The chatbot adapts to each answer: it explains the available options, collects only what's needed and moves the case to the next step.
A
An answer without the wait
Customers get essential information even when the team is offline — evenings and weekends included.
B
Fewer repetitive conversations
The most common questions stop interrupting your team's day.
C
Complete enquiries
The team receives the purpose of contact, the key details and a summary of the conversation.
D
A concrete next step
A conversation can end in a booking, a request, a payment or a handover to a person.
02 — One system, different industries
We don't ship the same bot to every company.
The core system is shared, but the knowledge, tone of voice, next steps and boundaries come from each company's processes. Pick an industry and watch the same conversation change.
I'd like to book a visit, but I'm not sure which consultation to choose.
I can help with that. Is this your first visit with us, or a follow-up?
First visit.
Then I'd suggest an intro consultation. The nearest openings are Thursday 12:00 and Friday 9:40 — which works for you?
Conversation outcome
Intro consultation — appointment booked
- Visit typeIntro consultation · 30 min
- LocationCity centre, room 4
- TimeThursday, 12:00
- PreparationInstructions sent by e-mail
Confirm visit
Ask reception to call me
Where's my order #48213?
Checking… Order #48213 shipped yesterday and is out for delivery. The courier expects to deliver it tomorrow.
And if the size doesn't fit?
You have 30 days for a free return. I can prepare a return label right away, or show you the size chart.
Conversation outcome
Order status — checked in your system
- Order#48213 · out for delivery
- Deliverytomorrow, courier
- Returns30 days, free
- Next stepReturn flow ready
Track parcel
Start a return
I'm looking for a three-bedroom flat, ideally with a balcony.
Got it. Which area should I search, and what's your budget?
Near the riverside, up to 450k.
I've found 3 matching listings. Two of them can be viewed this week — shall I book a viewing?
Conversation outcome
Matching listings — ready to view
- Preferences3 bedrooms · balcony · riverside
- Budgetup to 450k
- Matches3 listings
- Viewings2 slots this week
Book a viewing
E-mail me the listings
We're looking for help automating our support inbox.
Sure. Roughly how many requests do you handle each month, and in which tools?
Around 600 — mostly e-mail and spreadsheets.
Thanks. I'll route this to our implementation team — I'd suggest a short call with a consultant. When suits you?
Conversation outcome
Enquiry routed to the right team
- ScopeSupport automation
- Volume~600 requests / month
- Current processE-mail + spreadsheets
- Next stepCall with a consultant
Book a call
Send me the summary
In healthcare, legal and finance the chatbot handles the organisational side — it doesn't diagnose, interpret results or replace a licensed professional.
03 — Your company's knowledge
The chatbot answers only from information your company approved. When it doesn't know, it says so.
Answers are built from the materials you designate as binding: service descriptions, price lists, procedures, locations, specialist profiles, instructions and FAQs. When the answer isn't in those materials, the chatbot asks a clarifying question, says the information isn't available or passes the case to the right person.
- Website answers stay consistent with your current materials.
- Your team controls what the chatbot can say.
- New information goes live by updating the sources.
- Unusual questions are never hidden behind a confident-sounding guess.
04 — Understanding the need
Instead of a long form — questions that follow the conversation.
The chatbot doesn't ask every customer the same list of questions. Each question builds on the previous answer and leads to the right service, department, specialist or contact channel. The customer doesn't have to decode your entire offer, and your team receives an enquiry with the context it needs to pick up the case.
- Purpose of contact and case category — recognised in conversation.
- Key preferences, urgency and expected timing.
- Preferred way of getting in touch.
- A short summary of the whole conversation for the team.
05 — Bookings and simple cases
Customers can settle simple matters on the spot, without waiting for staff.
Connected to your calendar or booking system, the chatbot checks availability, suggests a time and creates the booking. It can also send a payment link, deliver a confirmation or log a request. A booking is only confirmed once the system says so — and if the system is briefly unavailable, the customer gets a clear message and another way to reach you.
- No hopping between chat, calendar and forms.
- Available options appear right in the conversation.
- Confirmation arrives in the same thread — plus e-mail or SMS.
- Simple matters get resolved outside business hours too.
06 — Connected to your tools
What's said in the conversation lands where your team already works.
The chatbot doesn't create yet another inbox someone has to check. Collected details are saved to your CRM, ticketing system, calendar or whichever tool the company uses — no retyping. And the chatbot only gets access to the specific actions agreed during implementation, never the whole system.
Highlighted modules show the path of a sample "book a consultation" action. The integration scope is agreed individually for every deployment.
07 — Handover to your team
Automation ends where a human decision or responsibility begins.
Not every conversation should be closed by a chatbot. A customer may ask for a person, a question may fall outside the approved knowledge, or the case may need a specialist. The chatbot then hands your team the whole picture: the goal of the conversation, the details provided, the earlier answers and the reason for the handover.
The customer doesn't repeat their story, and your employee starts from a concrete case.
08 — Control and boundaries
Your company decides what the chatbot knows and what it's allowed to do.
The scope is set before launch: knowledge sources, permitted operations, the data collected and the situations that go straight to a person. In healthcare, legal and finance the chatbot handles the organisational side — it doesn't replace someone with the right qualifications.
-
Approved sources
The chatbot draws only on knowledge your company designates.
-
Limited actions
Every scenario has a clearly defined set of operations the system may perform.
-
Exceptions go to people
Unusual and sensitive cases reach your team.
-
Easy to update
Knowledge and scenarios evolve with new needs and real conversations.
- Service descriptions & pricingactive
- Policies & proceduresactive
- Customer FAQ & guidesactive
- The model's general knowledgeoff
- Checking availabilityallowed
- Creating bookings & leadsallowed
- Paymentssecure link only
- Editing customer datarequires a person
- Customer asks for a personimmediately
- Question outside approved knowledgewith context
- Case needs a specialistwith context
- System error or urgent issuewith an alert
09 — What the conversations tell you
Conversations show what customers can't find on your website.
The panel shows what customers ask about, which conversations ended in a booking or a request, when a person was needed and which questions had no approved answer. We use it to fill knowledge gaps and refine the conversation scenarios — and you learn what your customers actually ask about.
10 — The difference
A standard chatbot answers questions. This one closes the loop.
A standard website chatbot
- Shows everyone the same answers.
- Searches a list of canned FAQ replies.
- Gives up when a question is unusual.
- Leaves data in a separate panel someone has to check.
- Mostly measures message counts.
A chatbot designed by DGBA
- Adapts its questions to the flow of the conversation.
- Answers from knowledge your company approved.
- Ends conversations with a booking, a request or a contact.
- Saves data in the systems your team already uses.
- Hands your team the conversation together with its history.
- Shows which conversations produced results and where knowledge is missing.
11 — How implementation works
We start with how your company serves customers — not with the technology.
We design the conversations and the knowledge
We define the questions, possible answers, next steps, knowledge sources and the situations that must reach a person.
We build the chat window and the connections
We prepare the on-site chat window, the integrations and the actions performed in your systems.
We pilot, then expand
We test real scenarios, review answer quality and widen the scope only once the core paths run reliably.
Technical notes for IT teams
Language model & conversation design
We match the model to complexity, cost, speed and data requirements.
Retrieval from approved knowledge
Answers are assembled from fragments of your approved materials (RAG).
API integrations
Narrow, authorised operations in your systems — no keys in the browser.
Operation validation & event logging
An operation is confirmed only after the source system responds; actions are logged.
Data storage & access
Data minimisation, retention policy and access procedures agreed separately for each deployment.
Answer & failure-path testing
Happy paths, incomplete input, API errors and out-of-scope attempts — all tested before launch.
Live deployment
Open MIA Preserve ↗
See the assistant at work on a healthcare website.
MIA Preserve is a working deployment: the assistant answers patients' questions from the clinic's approved information. Every company gets its own knowledge, conversation scenarios, look and integrations.
Build with DGBA
Let's start with the one process that eats most of your team's time.
Show us how you handle enquiries, bookings or support requests today. Together we'll pick the cases the chatbot should take over first — and map out what's needed to launch them.