Tenet Tecnologia · Applied AI for conversation and commerce
Otto · being set up at its first storeTechnology that sustains the conversation.
Tenet builds applied artificial intelligence systems for businesses that sell through conversation. The first is Otto, a WhatsApp customer service hub built for retail.
Fig. 01 — The same clock, on both sides of the conversation
Illustration · fictitious data
Hi! Do you have these in a 40? 14:02
- 14:02:11Message received
- source item on the site, size 40
- stock size 40, 2 in stock
- queue 3rd, by wait
- turn Ana, store round-robin
- transcription ready
- drafts in progress
3 waiting · by wait time
- Marina12:40Can I swap for a 38?
- Júlia04:12Can it arrive today?
- Beatriz00:12Hi! Do you have these in a 40?
Why we started
Every sale that runs through a conversation depends on someone on the other side. Customers feel every minute of waiting. Businesses rarely see that minute on screen.
At a premium fashion store, the average wait for a reply was
77 minutes and 58 seconds.1
Fig. 02 — From message to first reply
0 to 80 min · linear scale
More features don’t fix that. Seeing the wait, and deciding who answers, does.
So we started with the clock, the catalog, and the queue. And we put artificial intelligence to work helping the team, not taking its place.
We measure what we promise,and we promise what we measure.
Product · Otto
Being set up at its first store. Not yet in production.
WhatsApp customer service that sees the wait.
Customers feel every minute of waiting. With Otto, the team sees every minute too.
A WhatsApp customer service hub for multiple stores, each under its own brand, built for retail and starting with premium fashion. It knows the catalog and the stock, shows how long each customer has been waiting, and assigns conversations across the team, once the store turns that on.
Fig. 03 — The queue and the conversation, in Otto
Illustration · fictitious data
- Marina12:40Can I swap for a 38?
- Júlia04:12Can it arrive today?
- Beatriz00:12Hi! Do you have these in a 40?
I bought these in a 40 and they’re too big.13:49
Can I swap for a 38?13:50
-
Queue
The wait, on screen.
The queue is sorted by who has waited longest, not by latest message. Under a minute, nothing lights up. From 1 to 5 minutes, caution. Past 5, urgent. The clock counts minutes up to two hours, with no rounding.
-
Catalog
The catalog inside the conversation.
Search, photo, price, and the sizes actually in stock, without leaving the chat. One tap sends the item with photo, price, and link. The catalog is read from the store and refreshed every hour, and Otto never edits a product.
-
Assignment
Who answers.
Round-robin assignment among the people on shift, and a handoff when a conversation goes quiet. A cap keeps any customer from being passed around all night. It ships switched off: the store’s admin turns it on.
-
Metrics
Values appear here only once there is measured production, with date and source.
First response, measured.
Four numbers per period: the longest wait right now, average first response, how many were answered within a minute, and how many conversations never got a human reply. The summary comes as a ready-to-copy sentence, with the before alongside.
- Product
- WhatsApp customer service hub
- Model
- Multiple stores, each under its own brand
- Sector
- Retail, starting with premium fashion
- Channel
- Meta’s official WhatsApp Business Platform
- Catalog
- Shopify, read every hour · read-only
- Data
- Stored in Brazil, in São Paulo
- AI
- Voice transcription ready · off by default
- In progress
- AI reply suggestions · will ship switched off
- Stage
- Being set up at its first store · not yet in production
Also in Otto
- Where the conversation came from: when a customer arrives from an item on the site or an ad, the conversation opens with it.
- Size holds for a few hours or until the next day, and a flag when a requested size is back in stock.
- A sales board that shows how many days each conversation has been stalled.
- WhatsApp’s 24-hour window, with the deadline on every conversation. When it closes, the reply box gives way to the path that still works.
- Multiple stores, each with its own number, team, and color, adjusted to stay legible in light and dark themes.
- After-hours greeting, ready and switched off until the store turns it on.
Assistive AI.
The design rule stays: prices and stock come only from the catalog, and the AI that suggests replies starts switched off. The store turns it on.
- Ready
- Transcribes customers’ voice messages into text. It ships switched off, and is turned on only at the store’s request.
- In progress
- Draft replies for the team to review, and conversation summaries. Not available yet.
The mark of two Ts.
Tenet’s symbol is a T split by an axis, which makes two Ts. It is the same pair that sits in the middle of OTTO. Tenet and Otto are both palindromes, reading the same in both directions, so one mark serves the company and the product.
Principles
Tenet means principle. These are ours.
-
Measure before promising.
Every number we publish has a date, a source, and a unit. What we haven’t measured, we don’t claim.
-
AI helps the people answering.
It comes in to help the team, not to take its place. The AI that suggests replies starts switched off, and the store turns it on.
-
Nothing made up.
Price, stock, and terms come from the store’s own source. When the data is missing, the line is left out, not filled with a guess.
-
The store decides.
Automatic assignment, after-hours greeting, and reply suggestions all ship switched off. Changing how a team works is up to whoever runs it.
-
Promise only what can be measured.
The yardstick is each customer’s wait. It is the first thing on screen, and it is what we want to be held to.
Read top to bottom or bottom to top, they say the same: the first answers the fifth, the second the fourth, and the third sits on the axis. Like the name.
Craft
What doesn’t show on screen is product too.
-
Sign-inone-time code
No passwords stored.
Sign-in is by a one-time code sent to your email. What isn’t stored can’t leak.
-
Secretsvault
Credentials in a vault.
Each store’s WhatsApp and catalog access lives in a secrets vault, outside regular tables. No screen displays it.
-
Sendingunique key
One message, once.
Every send carries a unique key. A double tap or a flaky connection won’t send the same message twice, and when there’s no confirmation, the screen says so.
-
Isolationdatabase rule
Each store sees only itself.
Isolation between stores is enforced by the database, not the interface. A customer’s phone number shows only to those who need it, and an opt-out is honored on every outbound path.
-
Legibilitytwo themes
Contrast measured, not estimated.
Colors checked in light and dark themes for people who read the screen eight hours a day, often standing up.
-
Originautomated tests
Safeguards learned at the counter.
Otto’s engine was built from a WhatsApp service operation already running in retail. Each safeguard came from a real incident and ships with automated tests.
Log
What got built, month by month.
5 entries
- 01
Otto begins rollout at its first store.
Rollout - 02
Catalog inside the conversation, refreshed hourly.
Catalog - 03
Queue sorted by wait, and automatic assignment off by default.
Queue - 04
First-response metrics.
Metrics - 05
After-hours greeting, ready and switched off.
Greeting
About
A Brazilian technology and artificial intelligence company.
Tenet Tecnologia builds applied AI systems for conversation and commerce. We started with retail, because that is where conversation decides the sale, and where the question is almost always about an item: do you have my size? another color? can it arrive today?
Otto is our first product. It was built from a WhatsApp customer service operation already running in retail, with real customers, and it starts with premium fashion.
The name
Tenet is Latin for “it holds”, from tenere: to hold, to keep. In English the word came to mean a principle: something held to be true. It is also a palindrome that reads the same in both directions, with the N at its center as an axis. A conversation has two sides. We want technology that sustains both, and looks the same from either one.
- Name
- Tenet Tecnologia (being incorporated)
- Run by
- André Luiz Bastos Arruda
- CNPJ
- Being incorporated · Brazilian company registry number
- Address
- Manaus, Amazonas, Brazil
- Domain
- tenet.ia.br
- First product
- Otto · being set up
Contact
Talk to Tenet.
If your business sells over WhatsApp, tell us how customer service works today: how many people answer, how many conversations arrive each day, and where they come from.
Personal data and data subject requests: privacidade@tenet.ia.br
No form, no tracker: it’s an email, and a person reads it.