Developing a bot for automatic FAQ responses in Bitrix24
Half of incoming support chat messages are the same questions: 'How do I return an item?', 'Where is my order?', 'What are the delivery terms?'. Operators waste time on routine, and business loses money on unhandled complex requests. An online store with 500 orders per day received 200 inquiries, 80% of which were typical. After implementing a bot, response time dropped from 5 minutes to instant, and operators focused on complex cases. Support cost savings reached 40% in the first month — without loss of quality.
We develop a bot that automatically answers common questions and transfers non-standard ones to a live operator. The solution works directly in Bitrix24 Open Lines — no third-party services. The core is either fast keyword search or semantic RAG on embeddings.
How does the bot connect to Open Lines?
Bitrix24 Open Lines are the main entry point. The bot registers as the first operator of the line. When a client writes to online chat, Telegram, or VK, the message goes to the bot. If the bot answers, the operator is not disturbed. If not, it hands over to a human.
Registration of the bot in Open Lines is done via Bitrix24 REST API:
POST /rest/imopenlines.bot.register { "CODE": "faq_bot", "EVENT_HANDLER": "https://your-server.com/openline/event", "OPENLINE": "Y" } — source: REST API documentation
How does the bot determine what to answer? Two approaches to matching questions
Approach 1: Keyword search
Questions and keywords are stored in a database; the bot looks for matches in the message text. Simple to implement, easy to maintain. Ideal for 50–200 common questions.
CREATE TABLE custom_faq_entries ( id SERIAL PRIMARY KEY, question TEXT NOT NULL, answer TEXT NOT NULL, keywords TEXT NOT NULL, category VARCHAR(100), hits INT DEFAULT 0, is_active TINYINT DEFAULT 1 ); function findFaqAnswer(string $userMessage): ?array { $words = array_filter(explode(' ', mb_strtolower($userMessage))); $results = []; $faqs = getFaqEntries(); foreach ($faqs as $faq) { $keywords = explode(',', mb_strtolower($faq['keywords'])); $score = 0; foreach ($keywords as $kw) { $kw = trim($kw); if ($kw && (mb_stripos($userMessage, $kw) !== false)) { $score += mb_strlen($kw); } } if ($score > 0) { $results[] = ['faq' => $faq, 'score' => $score]; } } if (empty($results)) return null; usort($results, fn($a, $b) => $b['score'] <=> $a['score']); return $results[0]['faq']; } Approach 2: Vector search (RAG)
Questions are indexed as embeddings (OpenAI text-embedding-3-small, Yandex GPT Embeddings). When a new question arrives, nearest neighbors are searched in a vector DB (pgvector, Qdrant). Works on semantically similar phrasings without exact word match.
| Criterion | Keyword search | Vector search (RAG) |
|---|---|---|
| Database size | Up to 200 questions | From 200 questions |
| Accuracy on synonyms | Low | High — RAG is 3–4 times better on rephrased queries |
| Implementation complexity | Low | Medium (needs model and vector DB) |
| Speed | Instant | <1 sec for bases up to 10,000 |
| Maintenance | Edit keywords | Add questions without code changes |
Choice: For 50–200 FAQs keyword search suffices. For a large knowledge base with varied phrasings — RAG. RAG handles synonyms and typos better but requires model setup.
Examples of typical questions and bot answers
Real scenarios from an online store:
| Client question | Bot answer | Result |
|---|---|---|
| 'How to return an item?' | 'Return is possible within 14 days. The item must be unused, in original packaging. Fill out the return form in your personal account.' | Auto-closed |
| 'How much is delivery to Moscow?' | 'Standard delivery to Moscow is 300 rubles, delivery time 3–5 business days. Free delivery for orders over 5000 rubles.' | Auto-closed |
| 'Where can I track my order?' | 'Tracking link has been sent to your email. Tracking number: {number}. Estimated delivery date: {date}.' | Auto-closed |
| 'Do you accept cards?' | 'We accept Visa, MasterCard, Maestro, payments via Yandex.Kassa, and alternative payment methods.' | Auto-closed |
The bot successfully closes 60–70% of inquiries without operator involvement.
Requirements for integration and preliminary setup
For successful launch of the FAQ bot:
- Access to Bitrix24 REST API — requires Admin account and OAuth token generation.
- HTTPS server to receive webhooks from Open Lines (critical, B24 only works over secure protocol).
- FAQ storage — SQL database (PostgreSQL, MySQL) or CMS-specific (e.g., module in 1C-Bitrix).
- For RAG — subscription to OpenAI API, Yandex GPT Embeddings, or local model (e.g., nomic-embed-text).
- Monitoring system — logging of message processing errors, tracking missed questions.
Handling an event in Open Line
$event = json_decode(file_get_contents('php://input'), true); if ($event['event'] === 'ONIMBOTMESSAGEADD') { $sessionId = $event['data']['PARAMS']['SESSION_ID']; $dialogId = $event['data']['PARAMS']['DIALOG_ID']; $text = $event['data']['PARAMS']['MESSAGE']; $chatId = $event['data']['PARAMS']['CHAT_ID']; $faq = findFaqAnswer($text); if ($faq) { $b24->callMethod('imopenlines.session.message.send', [ 'SESSION_ID' => $sessionId, 'MESSAGE' => $faq['answer'], ]); incrementFaqHits($faq['id']); $b24->callMethod('imopenlines.session.close', ['SESSION_ID' => $sessionId]); } else { $b24->callMethod('imopenlines.session.transfer', [ 'SESSION_ID' => $sessionId, 'QUEUE_ID' => 1, ]); $b24->callMethod('imopenlines.session.message.send', [ 'SESSION_ID' => $sessionId, 'MESSAGE' => 'Connecting you with a manager, please wait...', ]); } } Escalation logic and transfer to operator
The bot transfers the conversation to an operator based on several criteria:
- Low confidence — match score below threshold (e.g., <0.7 for RAG or <5 for keyword).
- Explicit complaint request — if the client uses words like 'complaint', 'claim', 'didn't work'.
- Complex question — the message contains multiple topics simultaneously.
- Personal data — request for password, bank details (bot should decline and transfer to operator).
- Timeout — if conversation with client lasts >3 turns, bot offers a manager.
Clarifying questions
If multiple matches exist, the bot shows options:
if (count($topResults) > 1 && $topResults[0]['score'] < 10) { $options = array_slice($topResults, 0, 3); $message = "Please clarify what interests you:\n"; foreach ($options as $i => $r) { $message .= ($i + 1) . ". {$r['faq']['question']}\n"; } $message .= "0. Other (connect manager)"; $redis->set("faq_choice:{$sessionId}", json_encode($options), 300); } Administrative interface for FAQ database management
Content managers must manage FAQ without editing code. We develop an admin section in /bitrix/admin/ or as a Bitrix24 app:
- CRUD for questions and answers.
- Categories (Delivery, Payment, Returns, Technical questions).
- Statistics: how many times a question was closed by bot (
hits), how many transferred to operators. - View sessions where the bot didn't find an answer — priority for database expansion.
Want to reduce support load? Get a consultation for your project — we'll assess the architecture and suggest the optimal approach.
Why analytics matters?
Analytics shows the bot's real impact. Without it, you can't know which questions the bot handles well and which need improvement.
To evaluate effectiveness, we calculate the percentage of automatically closed sessions. Target auto-close rate for a mature FAQ database is 60–70%. Our clients achieve this within 2–3 weeks after launch.
Monitoring and continuous improvement
After launch, the bot requires regular maintenance:
- Weekly analysis of missed questions — review sessions transferred to operator, identify gaps in FAQ database.
- Update keywords (for keyword search) or retrain embeddings (for RAG) based on new questions.
- A/B testing of answers — which phrasings lead to fewer follow-up questions from clients.
- Response time monitoring — analyze delays in message forwarding and identify bottlenecks.
The tool automatically suggests entries to add to FAQ if the same unrecognized word appears more than 10 times per week.
Example session with the bot
- Client: 'Where is my order?'
- Bot: 'Your order is in status 'In transit'. Expected delivery date: December 5.'
- Operator not involved.
What's included
- Registration and configuration of the bot in Bitrix24 Open Lines.
- Design of FAQ database structure (keywords or embeddings).
- Development of logic: answer search, transfer to operator, clarifying questions.
- Creation of administrative interface for database management.
- Implementation of analytics and reports (auto-close rate, popular questions).
- Populating database with typical questions (up to 100 entries) and testing.
- Operations documentation and training for content manager.
Timeline and experience
We have over 5 years of hands-on experience and have delivered 50+ projects on Bitrix24. We work with 1C-Bitrix and Bitrix24 from the start.
| Stage | Duration |
|---|---|
| Bot registration in Open Lines, basic integration | 1–2 days |
| FAQ database structure, keyword search | 2–3 days |
| Transfer to operator logic, clarifying questions | 2–3 days |
| Administrative interface (FAQ CRUD) | 2–3 days |
| Analytics and reports | 1–2 days |
| Database population + testing | 2–3 days |
Total: 1.5–2.5 weeks excluding database population. Population is a separate task depending on documentation volume.
Order development now
We'll assess your project in one day. Tell us about your typical questions — we'll choose the optimal approach. Contact us via the form on the website or write to Telegram. Get a consultation right now.