> ## Documentation Index
> Fetch the complete documentation index at: https://docs.blubash.io/llms.txt
> Use this file to discover all available pages before exploring further.
# Hybrid service: AI agent and team working together
> See how the BluBash AI agent answers first on WhatsApp, Instagram and Webchat, qualifies, records and transfers to the team, which continues in Pipelines.
In BluBash, the AI agent and the team serve customers together. The agent answers right away, at any hour, resolves routine requests and organizes the conversation. The team takes over when the conversation requires negotiation, analysis or a decision, with the context already recorded.
Hybrid service works on every channel the agent is linked to: WhatsApp, Instagram and Webchat.
## How it works, step by step
The message reaches a channel (a WhatsApp number, the Instagram Direct inbox or the website Webchat). The AI agent linked to that channel answers in seconds, using the content in **Trainings**.
Frequent questions (deadlines, documents, hours, procedures) are resolved by the agent itself. For a new customer, it asks the qualification questions defined in **Behavior**.
With **Skills**, the agent adds a note summarizing what it gathered and applies tags such as **Hot lead** or **Active customer**. If **Allow updating contact data** is enabled, it also updates the contact's name, email and fields.
When the conversation needs a person, the agent runs the **Transfer to Team** skill configured for that situation (for example, **Sales** or **Support**). The transfer reason is recorded in the conversation.
The conversation enters the team according to its distribution setting (Round Robin, Available or Queue). The team member reads the note, creates the deal in the Pipeline from the conversation itself and continues the negotiation.
When the human part is done, the team member uses **Handoff to AI** to return the conversation to the agent, which collects documents, confirms details or closes the conversation.
## Real example: accounting firm
See how Lia, the AI agent of an accounting firm, serves a new customer on WhatsApp:
| Time | Who acts | What happens |
| - | - | - |
| 2:02 PM | Customer | "I run a distribution company and want to switch accountants. Do you handle presumed profit taxation?" |
| 2:02 PM | AI agent | Confirms the service, explains the switch and asks for revenue, number of employees and best time to call |
| 2:04 PM | Customer | Reports R\$ 6 million a year, 18 employees, call tomorrow after 2 PM |
| 2:04 PM | AI agent | Records the summary in a note, applies the **Hot lead** tag and transfers to **Sales** |
| 2:05 PM | Team member | Reads the summary, creates the deal in the Sales Pipeline and schedules the call as a task |
The transfer note is recorded in the conversation, for example:
> Conversation transferred from AI Lia to the Sales team for the reason: Qualified lead. Food distributor, presumed profit taxation, approximate revenue of R\$ 6 million, 18 employees, best time tomorrow after 2 PM. Interest: switching accountants.
The same flow applies to a visitor who chats through the website Webchat or a follower who writes in the Instagram Direct inbox.
## What to leave to AI and what to leave to the team
| AI agent | Team |
| - | - |
| Answer frequent questions at any hour | Negotiate prices and terms |
| Collect qualification data | Analyze complex cases and exceptions |
| Record summaries and apply tags | Decide deadlines, discounts and priorities |
| Route to the right team | Build relationships with strategic customers |
| Collect documents after the negotiation | Create and track deals in Pipelines |
| Close resolved conversations | Handle complaints and unhappy customers |
## Set up hybrid service
Define purpose, tone and behavior, and train it with frequent questions, files and your company website.
[Create an AI agent →](/ai-agents/create-ai-agent) · [Train the agent →](/ai-agents/train-ai-agent)
Each area that receives conversations from AI needs a team with members, a distribution rule and the channel in **Team Channels**.
[Create and configure teams →](/account-settings/create-configure-teams)
At a minimum: **Add Note** for the summary, **Add Tag** to classify and one **Transfer to Team** for each area.
[Configure skills →](/ai-agents/configure-skills)
Use **Test Agent** to simulate real conversations and see which skills the agent runs.
[Test the agent →](/ai-agents/test-ai-agent)
In **Channels**, open each channel (WhatsApp, Instagram or Webchat) and choose the agent in **AI Agent (Optional)**. From then on, new conversations on that channel start with the agent.
[Activate the agent and link it to channels →](/ai-agents/activate-deactivate-agent)
Add the channel to **Team Channels** for every team that receives conversations from AI. In the **Transfer to AI Agent** skill with **Target Team (Optional)**, this access is required for the team change to happen.
## Track the results
In the **Conversations report**, the **Total conversations over time** chart splits conversations into **Unassigned**, **Human only**, **AI agents only** and **Hybrid (human + AI)**. Use it to track how much volume AI resolves on its own and how much goes through the team. The **Conversations by channel over time** chart shows how volume is split across WhatsApp, Instagram and Webchat.
In the **360° View**, the **AI Analysis** of each conversation evaluates the work of the agent and the team and provides coaching suggestions, including for the agent's training.
## Next steps
* [Hand off conversations to AI and take over from AI →](/ai-agents/handoff-to-ai)
* [Conversations report →](/reports/conversations-report)
* [360° View and AI analysis →](/reports/ai-audit)
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.