Integrating ChatGPT with Bitrix24: Sales and Support Automation

We see managers spending up to 15 minutes on a single commercial proposal, while team leaders manually sort through dozens of leads daily. Support teams answer the same questions in chats and open lines. Integration experience shows: connecting ChatGPT with B24 automates up to 70% of these routine operations, yielding savings of up to $5,000/month for a 10-person team.

How It Works: ChatGPT + Bitrix24

The integration chain: Bitrix24 REST API ↔ your server (middleware) ↔ OpenAI API. B24 does not call OpenAI directly — a server-side application sits between them, handling logic, formatting prompts, and controlling token consumption. We implement the middleware on PHP 8.1+ using Bitrix component architecture, ensuring high load tolerance and fault tolerance.

Step-by-Step: Lead Assessment Process

1. A new lead is created in B24 from a web form, email, or call.
2. A CRM robot triggers a webhook to the middleware with lead data (via crm.lead.get).
3. The middleware anonymizes personal data and builds a prompt: "You are a B2B sales manager. Rate the lead 1–10 based on company size, request specificity, budget."
4. The prompt is sent to OpenAI API (POST /v1/chat/completions) with max_tokens: 200.
5. GPT returns a JSON score and recommendation.
6. The middleware writes the score to a custom field (crm.lead.update) and the recommendation to a comment.
7. If score > 7, a robot converts the lead to a deal and assigns a manager.

This whole process takes 1–3 seconds. Managers see a sorted list with scores, no manual sorting needed.

How GPT Helps Score Leads

One of the most effective scenarios. A lead enters B24 from a website form, email, or call. The middleware retrieves lead data (crm.lead.get): name, company, comment, source. It creates a prompt:

You are a B2B sales manager. Rate the lead from 1 to 10. Consider: company size, specificity of request, budget availability. Lead data: {name}, {comment}, {source}. Answer in JSON format: {"score": N, "reason": "...", "recommended_action": "..."} 

GPT returns a score. The middleware writes the score into a custom lead field via crm.lead.update and the recommendation into a comment. If the score > 7, a robot automatically converts the lead to a deal and assigns a responsible person. The manager no longer sorts leads manually: they see a sorted list with scores and recommendations. On average, projects recoup the investment in 2–3 months. For a business handling 1,000 leads monthly, total token cost is under $10, making it highly affordable. Contact us to estimate potential savings for your business.

How AI Assists in Open Channels

ChatGPT works as the first line in open channels (Telegram, VK, website widget). Technically: a client writes to an open line, the event ONIMCONNECTORMESSAGEADD is sent to the middleware. The middleware forms a prompt with context: conversation history, company knowledge base (FAQ, product descriptions). GPT generates a response and sends it via imconnector.send.messages. A key nuance: we set boundaries through a system prompt — if the question goes beyond competence, the bot transfers the dialogue to a live operator via imopenlines.session.transfer.

What Texts Can Be Generated with GPT

A manager clicks a button in the deal card — a custom action via placement CRM_DEAL_DETAIL_ACTIVITY. The middleware collects deal data, products, contact and sends it to GPT with the prompt "Create a proposal for...". The result — a text draft in the deal activity. Other scenarios: a post-meeting email (manager writes brief bullet points, GPT structures them into a follow-up) and task description (manager dictates the gist via voice into chat → Whisper transcribes → GPT structures into a task with a checklist → the bot creates a task via tasks.task.add).

How to Manage OpenAI Costs

OpenAI charges per token. One lead evaluation request consumes a few hundred tokens, costing less than $0.01. But without control, costs can rise. What we do at the middleware level: limit max_tokens in the API call (200 output tokens are enough), choose the model per task (scoring — gpt-4o-mini at $0.15/1M input tokens, generation — gpt-4o at $2.50/1M input tokens), cache identical requests, and set a daily budget — once reached, automatic requests stop and the administrator gets a notification.

Is Data Transfer to OpenAI Secure?

Before sending, the middleware anonymizes personal data: removes names, phones, emails from the prompt. API requests are not used to train OpenAI models by default, but we state this in the agreement. The middleware is hosted on your server, so you fully control which data leaves the perimeter. No direct OpenAI access to the B24 portal.

Comparison of GPT Models

GPT-4o-mini is 20x cheaper than GPT-4 for classification tasks and offers comparable accuracy for lead scoring.

Implementation Timeline

Achievements with 5+ Years on Market

We have completed over 200 Bitrix24 integration projects, with 98% client satisfaction. Our team includes certified Bitrix24 developers and OpenAI API specialists.

What's Included in the Work

• Middleware server for handling requests between B24 and OpenAI API
• GPT chatbot in Bitrix24 interface
• Automatic lead scoring with score writing to CRM
• AI assistant in open channels with transfer to operator
• Text generation (proposals, emails, task descriptions) from deal card
• Prompt tuning to your business specifics
• Token usage control: limits, model selection, caching
• Documentation and employee training

According to OpenAI documentation, for most business tasks the gpt-4o-mini model is sufficient. Order a turnkey integration of ChatGPT with Bitrix24 with a guaranteed result — we'll prepare a custom solution in 1–2 days.