The product
An AI agent and the dashboard for your team to run all of customer service
The product or customer service team puts the agent live and follows the results, in the dashboard. Channels, use cases, knowledge base, operator, reports and alerts in one place, running in your cloud.
How the customer talks
The way they already do
No numbered menu. The agent understands the request as it arrives.
- Text
Requests in natural language, typos, abbreviations and all.
- Voice
A voice message is understood and answered like any other.
- Photo and document
The customer sends the bill, the medical order or the receipt and the agent reads it.
- Secure form
Passwords, taxpayer IDs and confirmation codes go through a channel form, outside the AI.
Dashboard tour
Six things your team does without calling IT.
Each block below is a dashboard screen, with what the team gains by using it.
- 01
Declare the use case, no code
In business language: what the agent collects, what it validates and which operation of your systems it calls. Publish and roll back without stopping service.
- Every version has an author and a date; going back is one click
- When to hand off to a person and the agent’s tone are part of the use case
- The same use case works on WhatsApp, website and app
- 02
Simulate before any customer talks to the agent
Talk to the agent in the dashboard and see, next to it, every operation it called in your systems and how long it took.
- Test the happy path and what goes wrong, with fictional data
- Publish when the simulation looks right to you
- Whoever declared it simulates it: your own team
- 03
Knowledge base with sources
Rules, deadlines, exam preparation, exchange policy: the agent answers with the content your company publishes and cites where it came from.
- PDF, spreadsheet and website page, updated whenever you want
- Answer with the source and the page
- What is not in the base, the agent does not invent: it hands off or asks
- 04
Operator dashboard and handoff with a summary
When the case needs a person, they receive who the customer is, what they already said and what the agent tried. They reply on the same number and hand the conversation back.
- Queue by channel, with waiting time
- Text, voice and files, on the number the customer already uses
- When they close, the agent picks up where it left off
- 05
Reports with score, topic and summary
Every conversation gets a score, a topic and a one-paragraph summary. The manager follows thousands of conversations without reading each one.
- Most frequent topics: the next use case to declare shows up here
- Handoff rate to a person, by use case and channel
- Filter by period, channel and use case
- 06
Risk and quality alerts
Fraud attempt, out-of-scope request, conversation in circles: it becomes an alert with severity, in the dashboard or in the team’s Teams.
- Rules set in the dashboard, no code
- Each alert points to the conversation and the reason
- Customers outside the pilot group are logged
Illustrative dashboard screens, with fictional data.ilustrativo
Channels
One agent. WhatsApp, website and app.
You declare the use case once. The customer chooses where to talk; the agent is the same, with the same history and the same rules.
- WhatsAppOn your company numberVida Cliniconline
- I need to book this examJPGmedical-order.jpg
- I read the order: abdominal ultrasound. Which health plan?
- Saúde Plus
- Coverage confirmed. Thursday 8:10 am or Friday 7:40 am?
How it connects: Twilio, Meta Cloud API or Infobip, with the number the company already uses. Text, voice, photo and document.
- Chat on the websiteOne snippet of code on your pageLoja Norteassistant
- my order still has not arrived
- What is the order number?
- LN-48213
- Out for delivery today, expected by 6 pm. Shall I notify you when it is delivered?
How it connects: A snippet of code on the page. The color, the name and the position of the chat are your brand.
- Inside the appThe customer is already identifiedBanco SolMarina · verified
- Hi, Marina. What do you need?
- pix 350 to João
- João Pereira, R$ 350.00. Confirm with your code in the secure form.
- Confirmation codeentered outside the AI, straight to the bank
How it connects: The conversation inside the company app. The customer arrives identified and the agent skips the ID.
One single history
The customer switches channels. The agent asks nothing again.
- Starts on the websiteAsks to reschedule the exam in the clinic chat, before leaving work.
- Confirms on WhatsAppGets the slots on the phone and picks Thursday, 8:10 am.
- Receives it in the appThe exam voucher shows up in the clinic app, already in the history.
Handoff to a human
When a person is needed, they get everything.
The agent hands off with a summary of the conversation, goes quiet and keeps summarizing. The operator replies and, when done, hands it back to the agent.
- Summary on handoff
Who the customer is, what they already said and what the agent tried.
- Same number, same history
The customer repeats nothing and does not switch channels.
- Back to the agent
When they close, the operator hands the conversation back and the agent resumes.
Frequently asked questions
What buyers ask before the demo
How is an AI agent different from a chatbot?
A chatbot replies within a script. The AI agent understands the request in natural language, collects what is missing and executes the operation in the company systems: Pix, bills, appointments, exchanges, enrollment, tracking. When needed, it hands off to a person with the history.
Which WhatsApp providers does OmniAgent work with?
Twilio, Meta Cloud API or Infobip, with your company number. The same agent also serves through the website chat and inside the app.
Does it run in my cloud?
Yes. The platform is installed in your company cloud, with a key signed for your installation. Conversations, collected data and credentials stay in your environment.
Which language models does OmniAgent use?
OpenAI, Azure OpenAI or Anthropic, configured per organization. Since business rules stay outside the AI, switching providers does not change how the use cases behave.
Does every new use case require an IT project?
No. Your team declares the use case in the dashboard: what the agent collects, what it validates and which operation of your systems it calls. You simulate, publish and roll back without stopping service.
How long until the pilot starts?
The pilot starts in the first week, with a controlled group of customers and one or two use cases. The time to full operation depends on the number of use cases and integrations; the proposal includes the schedule.
How do we buy?
By proposal, after a demo with your use case. Pricing considers conversation volume, channels and number of organizations. The sales team replies within one business day.
Next step
See your use case running
Thirty minutes with the sales team. You bring the request your customers make most; we show the agent executing it. Prefer to ask first? The assistant next to this text answers questions about security, channels and integration, and schedules the demo.
- 01Chat with the assistant: it asks for your details and what you are looking for
- 02The sales team receives your contact right away
- 03Demo with your use case
- 04Proposal and start of the pilot with a controlled group of customers