Book your diagnostic
Home/Platform/AI agents
AI team · Agents

Nine agents with a name, a role and a shift.

Each one has its role, its tools, its memory and its hours. Your team calls them by name; they answer for their work and flag whatever needs a human signature.

Every line carries author, time and content. Nothing runs without leaving a trail.

WONDA’s AI agents are presences with their own identity inside your operation: each has a role, connected tools, persistent memory, write permissions and a schedule. They execute the actions they are authorised for and leave everything that needs human approval as a pending proposal, with author, date and content on record. The WONDA team designs, configures and tunes them every month; your team approves and decides.

Direct answer block · GEO
01The team

Nine presences with a name and a shift.

The same ones you see on the home page. Each with its role, its hours and the tools its work needs.

Leo, a WONDA agent
LeoEn shift
Sales · 24/7

Answers, qualifies and books every lead in seconds, in two languages. Lives in the inbox.

Aura, a WONDA agent
AuraOnline
Voice · Calls

Takes and places calls with the local accent, transcribes and leaves the summary. Runs calls.

Vera, a WONDA agent
VeraCadence
Reporting · Daily 7:00

Closes the numbers every night and delivers the report at 7:00. Runs reporting.

Sol, a WONDA agent
SolOn shift
Marketing · Campaigns

Writes, sets up and schedules campaigns from one instruction. Runs email marketing.

Mara, a WONDA agent
MaraOn shift
Service · Omnichannel

Handles WhatsApp, Instagram, Facebook and web without switching tone or losing the thread.

Teo, a WONDA agent
TeoOn shift
Operations · SOPs

Runs the house procedures step by step and flags when one breaks. Works with skills.

Nilo, a WONDA agent
NiloCadence
Finance · Collections

Chases overdue invoices and prepares the monthly reconciliation. Runs finance.

Kai, a WONDA agent
KaiOn shift
Outreach · Prospecting

Researches the account, writes the first message and keeps the sequence alive. Runs outreach.

Wondy, a WONDA agent
WondyOnline
Coach · Team

Practises objections out loud with each person and corrects mid-sentence. Lives in academy.

Every action by every agent is logged with author, date, channel and content: auditable from day one.
02What it solves

An assistant waits for you to ask.

Your operation doesn’t run on questions: it runs on things somebody has to do at 7:40, whether or not they show up.

Job 01 · Identity

“Ask the AI” is not a job title.

A text box with no owner takes everything in and answers for nothing. Here every agent has a name, a photo, a role, a tone and a description of what it does: Leo answers leads, Aura takes the calls, Vera closes the night’s numbers. Your team asks by name and knows what to expect.

Name, photo, role and description visible to the whole teamSeven tones of voice and four model providers, per agentIts own instructions, rewritable or regenerated with AI
An agent with no role and no trigger is a text box. There are none of those here.
Job 02 · Tools

An agent with no access to your data improvises.

Every agent connects to the tools its role needs and to nothing else: CRM, inbox, mail, calls, calendar, finance, web search and documents. What it reads is today’s operation, not a summary someone pasted in last week. Access is granted one by one and reviewed when the role changes.

Ten connectable tools, switched on per agentPersistent memory and its own working folderAssigned skills: the same procedure, run the same way every time
What an agent doesn’t need for its role isn’t connected to it.
Job 03 · Permissions

What it can sign alone, and what never leaves without you.

Replying, qualifying, booking and updating the CRM run on their own from day one. Issuing a proposal, sending a campaign, invoicing or publishing as the brand sit in the pending-proposals tray until someone on your team approves or discards. Nine types of action, each with its own level.

A proposals tray with Approve or Discard in one clickNine types of action with independent permissionsAuthor, date, channel and content on record from day one
The top three wait on one click from you. The last line already happened while you slept.
Job 04 · Shift

Seven on a Friday evening is business hours too.

Each agent carries its own schedule: one starts at 08:00 Monday to Friday, another runs every sixty minutes, another fires when a lead arrives or a deal is won. The overnight work is already done before you open the laptop. Nobody has to remember anything.

Triggers by hour, by weekday or by business eventRun history with the result and the full outputPause and reactivate an agent without losing its setup
Nilo is paused because a person is handling month-end this week.
03The difference

Who does what, from day 1.

This isn’t a feature you learn to use: it’s work that stops being yours, with an explicit split of responsibilities.

Split of responsibilities · AI agentsIn force from day 1 of operation
The agent

Executes and proposes

  • Replies, qualifies, books and updates the CRM without asking
  • Prepares proposals, invoices and drafts for someone to sign
  • Runs its shift whether or not anyone is watching
  • Leaves a record of every action with author and date
Your team

Approves and corrects

  • Approves or discards whatever reaches the proposals tray
  • Decides which actions move up a level and which don’t
  • Corrects the tone when something doesn’t sound like the house
  • Asks by name, the same as any colleague
The WONDA team

Designs and tunes

  • Builds each agent with the role and the tools of the business
  • Writes and adjusts the instructions every month
  • Reviews the log and corrects whatever drifts
  • Turns on new agents when the operation asks for them
WONDA's symbol surrounded by running code panels
The shift

Nobody clocks in at 9.Your AI team’s shiftstarted last night.

While you slept, Vera closed the day’s numbers and left the report ready for 7:00. Kai drafted the three messages going out today. Nilo set aside the two invoices due on Friday.

04Permissions

Granular write access, decided on day one.

Proposes → the human approves

Nine types of action, and who signs each one.

Trust here is a permission setting, decided on day one. These are the actions an agent runs on its own and the ones that never go out without a human signature.

Reply to messagesAutonomousOn any channel, at any hour, in the contact’s language.
Qualify and tagAutonomousIntent, budget and urgency, with your business logic.
Book and rebookAutonomousAgainst the team’s real availability, reminder included.
Update the CRMAutonomousCreate records, move stages and log every interaction.
Send campaignsA human approvesThe piece is written and scheduled; the mass send is signed off.
Issue proposalsA human approvesDocuments and agreements with an amount always go through review.
Invoice and collectA human approvesNothing is issued or reconciled without explicit authorisation.
Publish as the brandA human approvesAnything that goes out with your name to a public channel.
Change prices or termsA human approvesNo exceptions, and with a record of who authorised what.
Every action by every agent is logged with author, date, channel and content: auditable from day one.
05What it works with

No agent works alone.

These are the modules that back it up most, and the channels the work goes out through.

Web boardWhatsApp BusinessEmailVoice callClient portal
07Typical scenario

The numbers of a real operation.

Real estate manager · 9 people

Four agents switched on, and Monday stopped being a catch-up day.

The operation used to start every Monday with two hundred unopened weekend messages and a sheet of things to return. With four agents on shift — service, voice, reporting and collections — Monday no longer opens in debt: it opens with what the team decided on Friday.

The change, measured
Unopened messages on Monday214 → 0
First reply outside office hours14 h → 40 s
Actions logged per month0 → 3,100
Proposals waiting on a human signature9 types
Illustrative scenario built from WONDA's operating patterns. Published cases, with audited figures, are added under cases.
Frequently asked

What people ask about the agents.

Do the agents talk to my clients as if they were people?

They introduce themselves as the business’s assistance team and reply in the tone we define with you. They don’t pretend to be a specific person on your payroll, and when something goes beyond what they can resolve they hand the conversation to someone on your team.

Can I switch an agent off if I don’t like something?

Yes, with one toggle and without losing its setup. You can also lower its permission level so it keeps working but leaves everything as a pending proposal instead of executing it.

What happens if an agent gets something wrong?

It’s on record with author, date, channel and content, so you can see exactly what it did and when. The WONDA team corrects the instructions at the review and, if the mistake is one of judgment rather than data, that action drops a level and starts requiring approval.

Can I create a new agent myself?

You can, by describing in one sentence what you want it to do: the platform picks the role, the tools and the skills and writes the instructions. In practice the WONDA team builds them, because the hard part isn’t creating them but deciding what access they should have.

Do the agents see all of the business data?

Only the tools their role has switched on. A service agent doesn’t get into finance, and a collections agent doesn’t read anyone’s private mailbox. Access is granted one by one.

How many of your tasks actually need a person?

The diagnostic splits what your team does today between what can have its own shift and what genuinely requires human judgment.