Video summary
Как AI-агенты меняют работу команд. Онлайн-митап 02.06.2026
Main summary
Key takeaways
Key topic
The meetup explains how AI agents integrated into Bitrix24 are changing everyday work in companies and teams—moving from “ask questions” AI toward chat-driven task execution, automation, and agent platform customization.
Why companies need agents (analysis)
- The speakers present AI agents as already becoming reality, not hype:
- Over half of office workers already use AI at work (e.g., reusing customer data, analyzing parts of letters/tasks).
- The main challenge shifts to control and governance:
- How companies manage what agents can do
- How they ensure agents follow access rules
- Adoption trend:
- Started with large enterprises, now expanding to medium and small businesses
- Agents are delegated for specific functions such as:
- Marketing
- Legal document handling
- Document analysis/comparison
- Technical support
- Client replies in social/chat channels
Bitrix24 AI Agents: two main types
-
Bitrix GPT agent (chat agent)
- Available via messenger/chat, and also through web, desktop, and mobile apps
- Accessible from the right-side portal panel after rollout/update
- Works by understanding context (where the user is in Bitrix24: email, CRM, tasks, etc.) and offering/performing scenarios
- Respects user permissions:
- Runs on behalf of the requesting user
- If a user lacks access to CRM/tasks/entities, the agent won’t use them
-
Automation agents (auto-execution)
- Run on triggers (schedule, events like overdue tasks, missed calls, employee actions)
- Execute not only inside chat, but within Bitrix24 workflows
- Support customizable scenarios; admins can extend/modify agents
Bitrix GPT: main product capabilities (feature tour)
Bitrix GPT is positioned as a “product management interface”—helping users do more without navigating between sections.
CRM
- Reports and analytics via chat for:
- Managers, deals, leads
- Risk analysis, “hot deals”
- Task/risk guidance:
- Identify stuck tasks
- Summarize daily status and highlight risks
Tasks & Projects
- Determine priorities (“what to work on first”)
- Explain why tasks are stuck
- Create/close tasks and support project work
- Produce day summaries, including from mobile
- Find client-related emails and highlight important messages
- Draft/send replies on the user’s behalf (subject to permissions)
Company structure
- Create/change structure
- Find department managers
- Invite colleagues to the portal (reducing manual onboarding)
BI / reports
- Request report reviews directly from chat
- Permissions apply: users only see what they’re allowed to access
Internet research inside chat
- The agent can search the web and bring up-to-date information directly into chat.
Group chat collaboration
- Call the agent inside group chats so colleagues can view answers and continue discussion.
Third-party integrations via MCP
- Bitrix GPT can connect external services through MCP:
- 1C available; Gmail and GDrive planned
- “30+ apps” mentioned via marketplace MCP
- Benefit: employees stay in Bitrix24 chat, while the agent performs research/actions in connected services.
Security & deployment notes (model/data)
- The Bitrix GPT model is referenced in subtitles as “Bitrix GPT 55” / “BTrik GPT 55”
- Claims made by speakers:
- Hosted on servers inside the Russian Federation
- Requests do not leave Russia
- If personal data is sent, it is encrypted
- Speakers emphasize legal compliance and claim the quality is “on par with world leaders.”
Rollout / availability
- Rollout starts with partners, then expands to all clients:
- “Starting today… partners”
- Within a month: gradual rollout to all clients
- Bitrix GPT chat is expected to be available even on free Bitrix24, supporting team setup and product guidance.
Automation agents: examples of ready-made scenarios
The meetup highlights event- and schedule-driven agents.
Auto-start: speech analytics (converted to agent)
- Previously, AI analyzed calls against a script.
- Now it’s presented as an agent that can be configured:
- Change prompt
- Add actions after analysis
- Example:
- If rating is too low → notify manager / create tasks
Planning & control agent (start of day / end of day)
- At workday start:
- Creates a chatbot message with top priorities and “what needs special attention”
- Includes links to adjust deadlines from chat
- Analyzes comments and overdue/non-overdue items; sends a ready summary
- At day end:
- Summarizes what was done, created tasks, and missed comments
Weekly manager reporting
- Compiles employee summaries for a given period (e.g., week) and sends them in chat
- Framed as useful for managers who don’t want to open BI
Project management agent (weekly Monday summary)
- Sends a weekly status summary to project chat:
- Tasks at risk
- Delays
- Deadlines needing shifts
- Team dependencies
Agent customization: “not a black box”
A key emphasis: agents can be edited and extended by admins.
Two customization levels
- Basic level
- Edit prompt and chatbot names
- Set rights/access for the agent and configuration details
- Advanced level
- Add extra business rules/actions after analysis
- Extend workflows beyond defaults
Agent platform editor (“notes cubes” scripting)
- Uses an editor with trigger + action nodes (“script cubes”).
- Supports:
- Triggers: schedule, employee actions, task overdue, missed client calls, etc.
- AI processing layers:
- AI processor node (simple “analyze input and respond”)
- AI agent node (richer capabilities)
- Knowledge base connectivity (Knowledge Base 2.0 referenced) for AI-grounded responses
- MCP modules for tool interactions
- Example: CRM actions restricted by configured permissions (view list, create deals, create funnel)
Manual and training-related agents
Alongside ready-made auto agents:
- Manual launch agents
- Open line agent: answers client questions using internal documentation; can escalate to a manager if needed
- AI call agent for online booking: checks schedules, suggests/reschedules slots, updates bookings, cancels appointments
- Knowledge base search agent (manual)
- For new employees: answers questions using internal regulations/structure docs
- Employee testing agent (manual)
- Builds quizzes, sends questions/answers in chat, scores results, and reports back
Governance, confirmation, and admin controls (Q&A highlights)
- No “coding” by clients inside the agent; the agent follows predefined scenarios
- In support scenarios:
- Emotional/pattern detection can route complaint-like requests to employees/managers via rules
- Disabling the day-planning agent requires admin rights
- Personal data transparency:
- When asked whether the system can show what client personal data the agent processed/stored, the speaker claimed:
- Bitrix24 agents do not store requests/personal data on agent servers, so the agent can’t provide that
- When asked whether the system can show what client personal data the agent processed/stored, the speaker claimed:
- Agent action confirmation:
- Typical chat actions: no confirmation needed
- Irreversible changes (e.g., deletions): the agent asks for confirmation
- Update naming:
- Agents are described as part of the same platform update as Business Process 2.0, with added agent capabilities
Notable limitations / roadmap items
- Document content access:
- Agents currently may not fully read content of certain document/file types immediately
- Planned support mentioned for “this summer,” including:
- Files/tables and document formatting
- PDFs, tables, reading and generating documents
- Knowledge base 2.0 integration:
- Moving toward connecting with agents
Main speakers / sources
- Regina Olimbaeva
- Katya Obraztsova
- Video/product references: Bitrix24 / Btrix24
- Mentioned in the event schedule within the livestream: Anna Mogutova