Skip to content

Cases

What an AI agent looks like inside a real operation

8 industries, 8 operations. What used to be manual, what the Met does now and which integrations make it work.

How do businesses actually use an AI agent day to day?

Not as a question answering bot, but wired into the systems where the work already lives. A property manager routes three departments arriving on one WhatsApp number and books the showing against real open slots. A food distributor answers price and stock by reading its accounting system in the moment, and leaves the route order built. A restaurant accepts a delivery order with the real prep time and switches off what the kitchen ran out of. A vehicle rental operator answers availability at two in the morning. An agency turns a meeting transcript into a brief. And a food platform picks its content topic from search data before publishing. In health, a medical equipment store reads its catalog before handing off to an advisor, while a veterinary retailer keeps WooCommerce and Mercado Libre aligned with ICA compliance checks. All 8 are below, with the integrations behind them and what each one does not do.

To move from evidence to process design, explore AI process automation for business, where every procedure separates its trigger, tools, requirements and limits.

Industries
8
Integrations
12
Tools
178

Real tools from the connectors that appear in these cases.

Editorial method

How we verify a customer case before publication

A customer story becomes useful evidence when its source, attribution and limits remain visible. We do not turn it into a general Meteor promise.

1. Start with a primary source

Identity, role and experience come from customer-supplied, authorized video or source material. If the video does not yet have a public URL, we say so.

2. Keep every claim attributed

A number or result stated by a customer remains that customer’s account. It does not become a Meteor KPI, rating, guarantee or platform measurement.

3. Check the product capability

General capabilities linked from a case are checked against published product, integration and tool documentation. A customer story does not authorize us to invent a connector or action.

4. Publish the limit too

Each page separates what the source documents from what it cannot support. If a source or capability changes, we correct the page under our editorial principles.

See which sources anyone can open, which ones remain pending and which operations are published anonymously in the public customer evidence register.

Named customer case with evidence

Dakeva: confirmed bookings after hours

Kelly Cardona, CEO of Dakeva, shares her experience in an authorized video. The case separates her statement, external corroboration and what Meteor does not attribute to this customer story.

Read the Dakeva case

6 documented customer stories

Named customers, first-person evidence

Each page keeps the statement attributed and separates what it documents from what Meteor does not promise. The video appears once it has a public URL.

Health and beauty

Sanity

Ana, CEO of Sanity, says Meteor lets the company offer 24/7 service for booking, questions and responses. She also attributes a more organized schedule, clearer payments and automated follow-up to the implementation. Meteor does not publish a conversion rate, revenue figure or clinical outcome for this case.

Watch the video and read the evidence

Marketing and sales

Agencia de marketing digital

Julián Chavarro says his agency replaced rigid bots with Meteor to handle sales and customer service across the journey. He describes conversations involving text, audio and images, and says he later brought the implementation to agency clients. His result statements remain personal testimony, not a general Meteor KPI.

Watch the video and read the evidence

Financial services

CrayOne

CrayOne says it uses Meteor as a tool that brings together ChatGPT and Meta to support collections, sales and customer service. The company attributes lower operational workload to the implementation; Meteor does not claim the agent approves credit, calculates risk or makes financial decisions.

Read the case and its evidence

Logistics and transport

Mega Alfa Trans

Mega Alfa Trans says its agents work over WhatsApp to send notifications, receive worker documents and gather preventive maintenance information. The company highlights having information in one place and making decisions faster. The case does not claim Meteor drives vehicles, dispatches freight or makes operational safety decisions.

Read the case and its evidence

E-commerce and fragrance retail

Moon Parfum

Darwin Hernández, founder of Moon Parfum, says the company’s main bottleneck was responding to message volume on time without losing opportunities. He chose Meteor to configure the service and sales journey step by step through closing. The message volume mentioned in the video belongs to his account and is not a Meteor-audited metric.

Read the case and its evidence

E-commerce and mentoring

Negocios Conscientes

Jenny González, an e-commerce mentor and operator, says she uses Meteor to sell and resolve customer requests while staying focused on scaling and production. The video also shows booking inside a conversation. Her experience does not establish a sales rate or prove results for every store she advises.

Read the case and its evidence

8 operations

Cases by industry

Each one reads top to bottom: the operation, what it cost by hand, what the Met does and where it stops.

Property management Colombia

Three departments arriving on one WhatsApp number

A property management firm fields roughly thirty requests a day across WhatsApp, the website form and email. They come in mixed together: someone looking to rent, someone reporting a leak, someone asking about a lease renewal, all on the same line.

What used to be manual

  • A person read every message and decided which department owned it. Leasing, maintenance and administration shared one inbox.
  • Listing details were looked up in the system and pasted into the chat, one photo at a time.
  • Booking a showing took three rounds of messages before anyone found a slot the agent could actually make.
  • A maintenance report showed up as a loose voice note: no address, no photo, nobody assigned.

What the Met does

  • The Met searches the Wasi portfolio by city, property type, budget and area, answers only with listings that are live and published, and sends the full photo gallery when the prospect asks to see more.
  • The moment the prospect gives a name and a phone number, the client record is created in Wasi, where the sales team already works, instead of in a separate notebook.
  • Showings come from real open slots on the calendar, get confirmed inside the conversation, and get moved right there when plans change.
  • A maintenance report becomes a contact with its tag, its internal note and its reminder, and the conversation moves to the technical team with the reason written down.

Where it stops

  • It does not publish or edit listings. It reads the whole portfolio, but inventory is still loaded in Wasi.
  • It does not sign contracts or handle leases and rent payments.
  • The Meteor calendar does not sync with Google Calendar or Outlook. If the corporate calendar is the source of truth, Google Workspace is a separate connection.
Read the canonical case page
Food distribution Colombia

Route orders stopped living on personal phones

A food distributor sells to restaurants, hotels and neighborhood stores along fixed delivery routes. Orders arrive on WhatsApp at any hour, and usually the night before the route runs.

What used to be manual

  • Every rep worked from a personal line. When someone left the company, their customer history left with them.
  • Prices and stock were checked by hand in the accounting system, or answered from memory.
  • Orders were retyped from the chat into the order form, with data entry errors as a standing risk.
  • Collecting on a past due invoice depended on somebody remembering to call.

What the Met does

  • The Met identifies the customer by phone number even when they write from a new one, then confirms list price and warehouse stock by reading Alegra in the moment.
  • It builds the quote with a link to the formal PDF inside the same conversation, and when the customer accepts, records it as an order after showing the line items and the total for someone to confirm.
  • It applies payments against the right invoices without anyone opening the accounting system.
  • Past due invoices get listed, and the collection reminder goes out on WhatsApp on the agreed date.
  • The whole conversation stays in the CRM with its history, its tags and its assigned rep.

Where it stops

  • It does not replace your accountant. This is an operating window into Alegra, not accounting judgment.
  • It does not issue or void electronic invoices on its own. Those tools sit behind an explicit human confirmation, and a Met can have them turned off entirely.
  • It does not send bulk messages. Reminders go out one at a time as approved templates, because the channel has its own rules.
Read the canonical case page
Restaurants and delivery Colombia

The counter tablet stopped being the bottleneck

A casual dining restaurant with several locations takes delivery orders through Rappi and table reservations through WhatsApp. Both land at the same time, during the same rush.

What used to be manual

  • Orders were accepted from the tablet at the counter. During the rush, any order nobody got to canceled itself.
  • When the kitchen ran out of an ingredient, the item stayed listed until someone logged into the portal to switch it off.
  • With the kitchen backed up, orders kept coming in that were not going to be filled on time.
  • Table reservations were confirmed by message and recorded nowhere, so on the day itself nobody knew how many tables were committed.

What the Met does

  • The Met sees incoming orders, accepts them with the real prep time, and later marks them ready for the courier to pick up.
  • It switches off sold out items by SKU without touching their price, and switches them back on when the ingredient arrives.
  • It closes the storefront when the queue backs up and reopens it when the kitchen catches up, instead of stacking up orders nobody can fill.
  • Table reservations come from real open slots, get confirmed in the conversation, and get moved or canceled right there.

Where it stops

  • It does not change prices or edit the menu. It only turns availability on and off for items that are already published.
  • It does not manage couriers or delivery times. That side belongs to Rappi.
  • It does not resolve marketplace disputes or refunds.
Read the canonical case page
Vehicle rental

Availability gets answered at two in the morning

A vehicle rental operator serving international travelers gets inquiries around the clock and in several languages. The booking is decided on the spot: anyone who does not hear back writes to the next company on the list.

What used to be manual

  • Every availability question meant someone had to open the reservation platform and check dates.
  • Outside office hours there was no answer, and a good share of traveler inquiries land precisely outside office hours.
  • Add ons were offered from memory, and sometimes not offered at all.
  • The morning started with someone assembling the day list of pickups and returns by hand.

What the Met does

  • The Met searches HQ Rental for availability by date range and location, and answers with the classes on hand and the price of each.
  • It offers add-ons read from the account’s actual charge list, such as insurance, GPS, a child seat or an additional driver, so it only offers what the company actually sells.
  • It creates the full reservation: dates, pickup and return locations, customer details and additional charges.
  • It answers in the language the customer wrote in, with nobody on shift.
  • The daily manifest gives pickups, returns and fleet status, which is what organizes the morning.

Where it stops

  • It does not charge or process payments. It creates the reservation with its charges; the payment runs through whatever gateway the business already uses.
  • It does not cancel or modify an existing reservation. It reads and creates; changes are made in the platform.
  • It does not sign the rental agreement or handle the legal paperwork.
  • It does not manage fleet maintenance. It reads vehicle status and mileage, nothing more.
Read the canonical case page
Marketing and design agencies Colombia

The brief stopped living in one person’s head

A digital agency with several years behind it has mature processes and a concentration problem: the person who runs production is the only node connecting the brief, the plan and the delivery. Everything goes through them.

What used to be manual

  • The brief was filled in during the meeting and came out half done. Whatever was missing got reconstructed later from memory.
  • The meeting recording, the contract and the client files ended up in different folders, when anyone remembered to file them at all.
  • Account health was known by feel. Nobody could say how many revision rounds a deliverable cost.
  • Pipeline follow up lived in email, and a stalled opportunity announced nothing.

What the Met does

  • The Meet transcript feeds the brief, and client files get found in Drive without opening Drive.
  • The Met appends tracking rows to the spreadsheet without overwriting what was already there, and searches and sends email through Gmail from the authorized account.
  • The client record, its tags, its internal notes and its pipeline stage live in the CRM, and the pipeline summary answers how many opportunities and how much money sit in each stage.
  • The monthly client dashboard is built as a page inside Meteor using the operation’s own data.

Where it stops

  • The transcript exists only if the meeting was recorded with transcription on. That depends on how the meeting was set up, not on the integration.
  • It does not delete email or files, and it does not administer Google Workspace users, licenses or policies.
  • Meteor pages do not version changes: an update overwrites the previous content.
  • It does not sign documents. Electronic signature is not part of the platform.
Read the canonical case page
Agrifood commerce Colombia

Producing the content stopped being the launch bottleneck

A platform connecting food markets and wholesalers to end consumers reached launch with a finished product and no traction. The problem was not the technology: it was making noise with a small team and a commercial window that does not come back around until next year.

What used to be manual

  • Every social post was designed one at a time and published by hand, account by account.
  • The topic came from the team’s instinct, not from what people were actually searching for.
  • Nobody looked at which queries were already bringing traffic, so new content kept repeating what was not working.

What the Met does

  • The Met checks Google Trends to see which terms are rising and in which region, and picks the topic from data instead of a hunch.
  • It generates the image or the video for the post from a description, without opening a design tool.
  • It publishes the post to the Facebook page and the image, carousel, story or reel to Instagram, straight from the content queue.
  • Search Console reports which queries already bring traffic and which pages receive it, and that report comes back to the team every week.

Where it stops

  • It does not buy or manage paid advertising, and it does not send conversion events. This is organic publishing, nothing else.
  • It does not reply to public comments or mentions. Instagram and Messenger direct messages are handled, but through the Meteor channels.
  • Google Trends returns a relative index from zero to one hundred, not how many people searched for something.
  • It does not measure reach or engagement. It returns the post identifier, not its metrics.
Read the canonical case page
Medical equipment e-commerce Colombia

The product question reaches a human advisor with its full context

A medical equipment retailer combines a broad WooCommerce catalog, support from several locations and products that cannot always be handled with a generic answer. The customer needs to know which item exists and, for consultative cases, speak with a person.

What used to be manual

  • Every product question required someone to search the store before answering with price, characteristics or availability.
  • The sales team received mixed conversations, with no clean separation between a catalog question and a case that needed advice.
  • When a conversation changed hands, the next advisor had to reconstruct which product the customer had seen and what remained unresolved.

What the Met does

  • The Met searches WooCommerce by name or SKU and reads the live product record with price, stock, attributes and images before answering.
  • The contact, history, tags and notes remain in Meteor CRM instead of depending on the phone a person used to respond.
  • When the question requires human judgment, it hands off with a summary and reason, pauses itself and assigns the right advisor.

Where it stops

  • It does not provide clinical judgment or replace advice from a health professional.
  • It does not charge or verify the store payment method. Payment still happens through the gateway configured in WooCommerce.
  • This case does not claim an Odoo quoting workflow: that scope needs operational evidence before it can be published.
Read the canonical case page
Veterinary and agricultural commerce Colombia

The store becomes the marketplace source of truth

A veterinary retailer sells through its own store and Mercado Libre. The challenge is not merely copying a catalog: price, stock, category and compliance have to stay aligned without touching listings the company already managed elsewhere.

What used to be manual

  • Listing one SKU required choosing a category, filling required attributes and checking again whether the product was eligible.
  • Price and stock could drift between the store and the marketplace, creating a risk of selling the same unit twice.
  • A marketplace sale had to be reflected back in store inventory.

What the Met does

  • WooCommerce keeps the product record, variations and stock that act as the source of truth.
  • The Met prepares the Mercado Libre listing, reads required attributes and, for veterinary products, checks the ICA registry before listing.
  • The sync recalculates price with fees, taxes and shipping, mirrors stock and only operates listings marked with a managed SKU.
  • When a Mercado Libre sale arrives, it deducts the sold SKU and quantity in WooCommerce.

Where it stops

  • It does not guarantee Mercado Libre will approve a listing; the marketplace makes the final decision.
  • It does not answer buyer messages inside Mercado Libre or manage marketplace fulfillment.
  • This case does not attribute Rappi operations to Meteor until a channel-specific trace is verified.
Read the canonical case page

The next leg

When the order has to ship with a carrier

None of these cases ends in a shipping label. Plenty of operations do, and that leg is handled from the chat too.

Aveonline

11 tools

Recommended partner

It handles the order end to end on AveCRM: reads the catalog with prices and variants, builds the order and shows you items, totals and customer details before creating it. Confirming the order is what triggers the shipping label with the carrier, and the response comes back with the carrier and the freight cost. It also moves the internal status and voids the order, returning items to inventory.

Where it stops: an order that already has a shipment number cannot be voided, and it does not quote or track on its own.

See the Aveonline page

TCC

23 tools

It quotes against your commercial agreement and returns the real amount with the charges itemized, tracks one shipment or many at once, translates the status into plain language and, when there is an exception, says whether it falls on the carrier, the customer or a third party. Before promising cash on delivery, it confirms the destination accepts it.

Where it stops: these are read only tools. It does not create shipments, request pickups or void labels.

See the TCC page

The full catalog of connectors is on the integrations page.

Benefits

What AI agents are for: less manual work, more sales closed

24/7

Support that never pauses

Your business answers and sells outside business hours too.

Seconds

Instant first reply

Every customer gets a first answer right away, at any hour.

0

Leads without follow-up

Reminders and automated sequences so no prospect goes cold.

100%

Conversations on record

Every chat, order and quote stored and measurable in the CRM.

1

One operation, one place

Customers, orders and quotes in a single place, not across five tools.

42

Integrations ready to go

Connect your store, your logistics and your tools with no custom build.

Testimonials

Companies already using AI agents: real businesses selling with Meteor

Do not take our word for it: hear it from the ones who already automated.

Kelly Cardona

CEO · Dakeva

“Since we got Meteor, I often go to sleep and the next morning we have confirmed bookings.”

Public first-person source

Open the original video on YouTube (opens in a new tab)

Published on YouTube

Julián

Community lead · E-commerce

“With Meteor I automated sales and customer service 100%.”

Public first-person source

Open the original video on YouTube (opens in a new tab)

Published on YouTube

Ana

CEO · Sanity

“With Meteor we give our customers 24/7 support.”

Public first-person source

Open the original video on YouTube (opens in a new tab)

Published on YouTube

Public customer reviews

What customers say about Meteor on Google

Each quotation links to its public source. These are attributed customer accounts, not audited metrics, guaranteed results or a score created by Meteor.

“Una compañía de tecnología que se preocupa realmente por sus clientes, esto es de mucho valor, tomen sus 5 estrellas cracks!”

Camilo Peinado

“Nuestro aliado estratégico en i.a para clientes. Hemos crecido muchísimo este último año, con el apoyo de quienes nos permiten mirar de dentro sus empresas para ofrecerles soluciones”

Andrés García Carpizo

“Excelente plataforma para Gestionar Agentes de IA, fácil de usar, soporte 24/7 y actualización constante. Muy recomendados”

Ximena Villalobos

“Como empresa implementamos Meteor y es lo mejor que pudimos hacer, nos facilitó muchos procesos que teniamos manuales y nos ayudo a gestionar mejor nuestros tiempos. Súper recomendados”

Cindy Rivas

“Responsive, etic and quickly company”

ALEJANDRO STEEVEN GALEANO MORENO

“Meteor es una plataforma maravillosa que está realmente pensada en el cliente final, que tiene un nivel de respuesta de otro nivel y que está en constante evolución, siempre en pro de un servicio de excelencia y un desarrollo de producto increiblemente innovador.”

Fredt Bogotá

FAQ

Questions about these cases

To the extent your tool is in the catalog, yes. Every capability on this page traces back to an integration page, and each one states which tools the connector actually exposes, which credential it needs and what it does not do. If your system is not in the catalog, a Met does not cover that part until the connector exists.

It hands the conversation to a human rep, leaves a note with the reason and steps out of that conversation. Before assigning, it lists the people in the workspace so the name resolves to the right person, which is what keeps the conversation out of the wrong inbox.

It depends on how many distinct jobs it has. A Met is a role with its own tools: the one that answers and qualifies is not the one that takes orders or the one that publishes content. They work together and hand the conversation to each other when the topic changes.

Not in these cases. It builds the quote, the reservation or the order, and payment happens in the gateway or the system the business already uses. Meteor does not store card data. Where a gateway is connected, the Met can generate the payment link and later check whether it cleared, but the money never moves through the conversation.

Wasi and Rappi are connected for Colombia, and Aveonline and TCC are Colombian platforms. Alegra runs in Colombia, Mexico and Argentina. The calendar, the CRM, image and video generation, Google Workspace, Google Trends and publishing to Facebook and Instagram are not country bound. Neither is HQ Rental: you specify the server region of your account.

Want to see this on your own operation?

We show it with your integrations connected, not with a canned demo.