01 · Support
answers to common questions around the clock
Delivery, payment, returns, warranty and availability. Answers are generated from your documents and catalogue, not from general model knowledge.
Clients are not annoyed by bots. They are annoyed by bots that do not know the answers or invite a human.
We build bots that answer from your documents and systems, show order status, book meetings and pass complex enquiries to a manager with the conversation context.
The arithmetic
Bot value is simple to calculate: number of enquiries × share of common questions × minutes per answer
With 300 enquiries a day, where 60{85c32a404beb3a17fad9859395e98171d5844d669a872c9f6728ab61e397d515} concern order status, delivery terms and availability, and each answer takes 4 minutes, this is about 12 hours of work daily. The second economic factor is out-of-hours enquiries: they either wait until morning or are lost.
If you have 15 enquiries a day and each is unique, we will say a bot is unnecessary and propose organising answers and templates.
- 01Number of enquiries and their channels
- 02Share of questions repeated each week
- 03First-response time during and outside business hours
- 04Are the documents, terms and rules up to date
- 05Where the bot gets statuses, prices and availability
- 06Who takes over when the bot hands off the conversation
Scenarios
Four challenges where a bot delivers value fastest
02 · order
Status, changes and repeat orders
The bot sees orders in accounting, shows status and tracking, accepts change requests and processes repeat purchases without a manager.
03 · Enquiries
Qualification and meeting booking
It clarifies the challenge, budget and timeline by your criteria, offers available calendar times and creates a CRM deal with the full conversation.
04 · Team
Internal knowledge-base bot
answers team members about procedures, customer terms and internal processes with links to the source document.
A good bot is measured not by the number of dialogues. and the share of questions closed without a person or complaint.
Pilot
One channel and one set of topics — and an honest measurement
We take a limited scope: the highest-volume channel and recurring topics. We compare it with how the process worked before.
- 01Channel and topic selection
- 02Baseline time measurement
- 03answers on your data
- 04Rules for handing off to a person
- 05Scaling decision
We measure the share of dialogues closed without a manager, first-response time, handoffs to people and unresolved enquiries. If the figures do not improve, we do not propose scaling.
Honest boundary
What bot does not do independently
-
01 · money
Discounts, compensation and rule exceptions
The bot states the current terms but does not negotiate prices or promise refunds. Such conversations go straight to a manager.
Answer rules · Escalation
-
02 · Complaints
Conflicts and sensitive topics
Dissatisfaction, a complaint, a mention of health or legal consequences is a signal to hand the conversation to a person, not answer with a template.
Handover signals · Manager summary
-
03 · Data customers
Personal data and enquiry history
We transfer the minimum data needed for answers, anonymized where possible. The model provider, processg region and dialogue-retention period are agreed at launch.
Data minimisation · Retention policy
-
01
meeting 30 minutes
Enquiry flow, channels, current ownership and the most frequently repeated questions.
-
02
Enquiry review and knowledge base
We classify dialogue history, collect answer sources and find topics with the greatest scope and lowest risk.
-
03
Pilot in one channel
Launch on a limited set of topics, control answer quality and keep a log of handoffs and errors.
-
04
Scaling and integration
Remaining channels, CRM and accounting connections, knowledge-base update rules, and quality and cost monitoring.
Cost
We calculate the cost per thousand conversations
Payment for models and third-party services goes directly to providers for actual usage.
Enquiry reviewOne-off
from $1 200
Dialogue-history classification and an estimate of the share of enquiries the bot can close without a person.
- Enquiry-flow analysis by channel
- List of topics suitable for automation
- Savings estimate in hours and money
- What must be organised before launch
A bot for the key process4–6 weeks
from $3 500
One channel, answers based on your data, handoff to a person and integration with your systems.
- Knowledge base and answer rules
- Integration with CRM, accounting and calendar
- Escalation scenarios for a manager
- Report on actual savings
Questions
What people usually ask about chatbots
A scripted bot works only when the client follows a predefined path and fails at the first free-form question. A model-based bot understands wording in people’s own words but is responsible only for your materials; you define its rules and boundaries.
Answers are based on your documents and systems: price lists, delivery terms, statuses and the knowledge base. If information is missing, the bot says so and hands the enquiry to a person; this is built into the rules.
when the client asks directly, in sensitive topics, when a question is repeatedly misunderstood and in anything involving money, complaints or commitments. The manager receives the conversation history and a short summary, not an empty chat.
A widget on the website, Telegram, Viber, WhatsApp, Instagram Direct and Facebook Messenger. The logic is one; channels are connected by priority, starting where there are the most enquiries.
A history of typical enquiries, current documents with terms, access to systems that provide statuses and prices, and a person who approves answer wording. Without this, the bot will respond with generic, unhelpful phrases.
Payment for the enquiry scope plus support scenarios and knowledge-base updates. We calculate the cost per 1,000 dialogues so the economics can be compared with current handling.
Start
Let us look on your requests – and we will say what share the bot can close without a person
A 30-minute online meeting with the founder. We will review the channel, typical questions and where the bot should hand off complex conversations, then send a written summary.