OleonAI
Inbox engineAn agent that answers. And knows when not to.
An agent is only worth having if it is right, and right means answering from your material rather than inventing something plausible. AI answers from what you gave it, in the Inbox where your team can see every word.
Grounded
It answers from what you wrote.
Or it fetches someone who knows.
What it read. Your knowledge base, written like a document. It answers from the passage that matches, and when nothing does, it says so instead of making something up.
What they saw. A straight answer in seconds, or a person who picks up the same conversation with everything already said.
An example: a customer asks about delivery to Jaffna; the agent finds the Delivery passage and answers from it. The customer then asks about instalments; nothing in the knowledge base covers it, so the agent says it will fetch a person, flags the conversation for review, and a teammate answers.
Sales
It turns questions into sales.
Answered fast. Handed over when ready.
- Harvard Business Review study, 2011
- 7×more sales chats when you reply within the hour.Harvard Business Review study, 2011
- MIT study, 2007
- 21×less likely to go anywhere if you reply in 30 minutes, not 5.MIT study, 2007
- Harvard Business Review study, 2011
- 23%of businesses never reply. The rest take about two days.Harvard Business Review study, 2011
- Meta, 2023
- 600Mbusiness chats a day on WhatsApp, Messenger and Instagram.Meta, 2023
From studies of online enquiries, mostly in the US. The lesson for chat: reply fast.
They ask.
Late at night, from your ad. Your team has gone home.
They get an answer.
Straight away: price, stock and delivery, from your own info.
They say they want it.
The agent hands them to a person.
Your team closes it.
The whole chat is there. Nobody asks twice.
Open when you are closed.
Questions do not wait for opening hours, and a buyer who hears nothing asks the next shop. The agent answers at night and at weekends, so the morning starts with conversations to close, not a queue to apologise to.
Every ad click gets a reply.
You paid for that tap. An ad that opens a chat is only worth its price if somebody answers, at the moment the interest is highest.
The same answer, every time.
The price, the returns rule and the delivery time come from one page you control, not from whoever happened to be on shift.
Your team takes the ready ones.
The agent handles the questions. The conversations that want a person, to order, to negotiate, to complain, go to one with the history attached.
Every asker is a lead.
Whoever writes in lands in Contacts, so a question that did not turn into an order today is a name you can come back to with Bulk Messaging.
The people, prices and counts in these pictures are examples. Ads that open a chat are in beta.
AI
Everything an agent should be.
Right, or honest about it.
Grounded answers. From your knowledge base, not from a guess.
A knowledge base. Written like a document, with headings and lists.
Context. The whole thread is part of every answer.
Flags its doubts. Unsure replies go to Reported messages.
Hands over. A person picks up the same conversation.
In the Inbox. Every word it sends, where your team reads.
Your own key. Model usage billed to you, by your provider.
Set up with you. The first knowledge base, written together.
Oleon + Your LLM vs popular AI agents
About 80% less, all in.
Same AI replies. A fraction of the price.
The popular-agent figures are a leading messaging platform’s own published price for its business AI agent from 1 August 2026, US$2 per 1M tokens, about 4–5¢ a message, including delivery. The Oleon + Your LLM figure uses OpenAI’s list price for GPT‑5 mini (US$0.25 per 1M input tokens, US$2.00 per 1M output) and the same 20,000 tokens a reply that price assumes, plus 500 for the answer, and adds the WhatsApp service-message charge for Sri Lanka, US$0.002 a reply, from 1 October 2026. Rates differ by country. Lighter models cost less. Prices as of 28 September 2026.
Bring your own key
Your key. Your model bill.
No markup in between.
The agent runs on your own AI provider account. What it costs to think is billed to you by your provider, at their price, and nothing about the model is added to your Oleon plan. Busy month or quiet one, your provider's own dashboard shows exactly what it used.
Bring your own key from
- OpenAI GPT
- Google Gemini
- OpenRouter
And through OpenRouter
- Anthropic Claude
- Mistral
- DeepSeek
- Qwen
- and hundreds more models
Tech specs
Everything in AI.
- Answers from your knowledge base
- Context across the conversation
- Replies it flags for review
- Handover to a person
- Every answer in the Inbox
- Bring your own AI key
- Set up with our team
Questions
Before you ask.
The ones we hear most about AI. If yours is not here, the answer is a message away.
What stops it saying something wrong?
It answers from your knowledge base, it reports the replies it is unsure of, and every word it sends is in the conversation, so a wrong answer is findable rather than deniable.
Can we set it up ourselves?
Not yet. We set it up with you, because the knowledge base is what makes it right, and the first version of that is worth writing together.
Will it actually help us sell?
It answers the questions that come before a sale, at any hour, and passes the ones ready to buy to your team. Everyone who asks lands in Contacts, so the ones who did not buy today can be followed up with Bulk Messaging.
What does the AI cost to run?
The model runs on your own key, from OpenAI, Google Gemini or OpenRouter (which reaches Claude, Mistral, DeepSeek, Qwen and hundreds more), so its usage is billed by that provider to your account. Nothing about the model's cost is added to your Oleon plan.
Should we use AI or Scout?
Scout for the questions you can predict and want answered word for word, AI for the long tail. They work in the same Inbox, and many teams run both.
Keep exploring.
More of what Oleon does.
All productsIt starts with one conversation.
Bring the channels you already answer on into one place, and see what the rest of the week looks like.
Not ready to sign up? Join the waitlist and we will come to you.
