We built Mihu, an AI contact-centre app for managing conversations, calls and tickets, where AI agents and human teams work the same queues.
Client: MihuIndustry: SaaS & TechnologyService: AI & Automation
Overview
Client
Mihu
AI contact-centre and voice agent platform
Engagement snapshot
The challenge
Contact centres that add AI often end up running two systems. The bot's transcript lives in one tool and the staff member's screen in another, so the customer is asked to start again after a handover.
Staff then move between an AI platform, a softphone and a CRM. Once a call ends, follow-up depends on someone remembering it, and complaints or sales leads can simply go quiet.
When the AI's history is not in front of the person taking over, the customer explains it all twice.
Switching between tools during a live call slows staff down and lets details slip.
A complaint or a warm lead can vanish if nobody turns the conversation into a task.
The solution
We built a shared inbox and queue where AI agents and team members work the same threads, with both sets of replies in one conversation.
Staff take and make calls from the same screen, and can join or take over a live AI call with their own voice.
When a conversation ends, an Observer agent decides whether it needs follow-up and, if so, opens a ticket for the right person or queue.
Key features
Autopilot sends replies itself, Copilot drafts for a person to approve, and Human First keeps AI out of the conversation.
Business impact
Set per queue or channel
Team members follow each AI conversation as it happens and can approve or decline actions the agent asks permission for.
Business impact
Person stays in control
AI and human conversations are analysed against the same rubric, so managers compare them on equal terms.
Business impact
One rubric for both
Before and after
Before
Start from the beginning
With separate bot and agent tools, the person picking up usually has to ask what happened.
After
Same thread
The person inherits the whole conversation, the facts found and the actions already taken.
Before
On or off
Many setups treat AI as a separate front door that either handles a contact or does not.
After
Set per queue
Teams choose how much the AI does for a queue, a channel or a single conversation.
Before
Relies on memory
After a call, someone has to notice the issue and log it by hand.
After
Ticket on the board
A titled, prioritised ticket reaches the chosen person or queue as soon as the conversation closes.
Results
AI-to-human handoff
A person can take a conversation from the AI with the full history, then hand it back when finished.
Before: Two toolsAfter: One thread
Automatic tickets
Complaints and sales opportunities become tickets with a type, priority and link to the original call or thread.
Before: Manual loggingAfter: Auto-created
iOS and Android
The Loop app puts the inbox, calls and tickets on a mobile device, with alerts for new messages and assignments.
Before: Desk onlyAfter: On mobile
Trust and control
The AI passes a conversation to a named person when the rules a team has set say it should.
Team-set rulesIn Copilot mode, a reply waits for a person, and teams can require a tap on anything touching money or a commitment.
Approve firstEvery change to an agent is saved as a version, and any earlier version can be restored.
Versioned agentsSimulator Studio lets teams try an agent out before it speaks to customers.
Pre-launch testsFaster delivery
AI agents and team members work from these same modules in the Mihu contact centre.
Holds calls and messages with queues, assignments and SLA timers.
Provides calling, recording and transfers inside the desk.
Moves follow-up work through five stages, on a board or in a list.
Keeps customer records, tags, custom fields and a calendar beside each conversation.
The Mihu app is used to manage customer conversations, calls and tickets in one place. AI agents and team members share the same inbox and queues across voice, WhatsApp, SMS, email, Instagram and Messenger. Staff can follow AI conversations as they happen, step in when needed, and work from desktop or mobile.
They work the same threads in one queue. Teams choose between Autopilot, where the AI replies itself, Copilot, where it drafts for approval, and Human First, with no AI involved. The setting applies to a queue, a channel or one conversation, and the person taking over sees the full history.
Mihu creates tickets automatically when a conversation ends with something to follow up. An Observer agent reads the conversation against plain-language instructions set by the team, then sets the type and priority. The ticket goes to the person or queue chosen for that type and links back to the conversation. Staff can also open tickets by hand.
If your bots and your support team work in separate tools, we can help you build a contact-centre product that keeps every conversation in one place.