We make AI do the work.
We carry companies into the new era, by a method never seen before.
Read our manifestoIt does not let the AI
make things up.
When it answers a question about your data, it checks every sentence against your records. A sentence with support shows its source beside it, and one without is marked or never said at all.
How do you stop the AI from lying?Measured in our end to end test, September 2026
- 92%Right answersQuestions asked in plain language, on a fixed set of 25, answered right.
- 80%Right refusalsQuestions whose answer is not in the data were refused rather than invented. A right refusal scores as much as a right answer.
It prepares reports
and forecasts what comes next.
It prepares the reports you need as PDF or Word files, on the schedule you set, and they wait for you in the console. To forecast what comes next it races several methods against each other, keeps the one that fits best and shows you the result on a chart.
Measured in our end to end test, September 2026
- 16.4 sLoading a million rowsA sales file of 1,067,371 rows was loaded, and the totals matched the source file exactly.
Its agents connect
to the internet securely.
They scan your competitors' prices, the news in your sector and the complaints written about you, for you. They summarise what they found with the sources beside it, so you can see where every piece came from.
You talk
with your data.
An invoice issued four years ago and long forgotten is on the system's map too. So you can talk with all of your data, old or new, in plain language.
Measured in our end to end test, September 2026
- 98.7%Matching duplicate recordsA customer's scattered records were merged. The 500 matches it was unsure of went to a person, and no customer was lost.
You do not have to learn
how to use it.
You say what you want in plain language and the system writes its own configuration. Ask for an agent, a data flow or a report: the operating system configures itself.
- 34Ready templatesReady setups for sales, support, human resources, content and data work.
- 61Ready step typesSteps such as reading data, reshaping it, writing a report and waiting for an approval come ready.
An example day
- 07:30The overnight rule check finished. 2 exceptions are waiting for you.
- 08:00The weekly report is ready, as a PDF under Reports.
- 09:15A payment amount looks unusual. Your decision will be remembered for it.
- 11:40Competitor prices were scanned. 3 changes are summarised with their sources.
- 14:20Your meetings were summarised and the action items recorded.
- 16:05One sentence had no support in the records and was left out of the answer.
It starts work
before you arrive.
It runs the work you scheduled on its own. It prepares the report, checks your rules and puts an unusual record in front of you. When you open it in the morning, it is all ready.
- 11Kinds of triggerAt a set time, at intervals, on a date, when data changes, when a message arrives or when another system calls, it starts the work itself.
It connects to the systems you already use.
- Excel and CSV
- PostgreSQL
- MySQL
- SQL Server
- Oracle
- MongoDB
- Google Sheets
- SAP
- Logo
- Mikro
- Netsis
- Nebim
- Shopify
- REST API
It connects to your databases, your files and your business applications, and reads Turkish ERPs such as Logo, Mikro, Netsis and Nebim straight from their own databases.
Through its API it works with
every system and screen you have.
Every agent and every flow you set up is served as an API of its own the moment it is deployed. Your website, your mobile app or your intranet calls it directly, and DEHA Lab connects to the APIs of the systems you use in turn.
- 40+Ready connectionsFor databases, files, ERPs and business applications.
- 4Message channelsWhatsApp, Telegram, Slack and mail.
- Website
- Mobile app
- Intranet
- ERP
- CRM
{
"message": "What were last week's sales in Central Anatolia?"
}{
"status": "completed",
"result": {
"content": "Last week's sales in Central Anatolia came to 2.1 mn TL."
}
}Intelligence is not in the model,
it is in the frame
A language model is powerful, and it does not see meaning. It cannot tell what it does not know, so no one can put it at the center of a company. We do the opposite: the model stays at the edge, and we build the discipline around it. Four principles carry the whole operating system.
The model is not the center
The heavy lifting runs on engines that give the same answer to the same input. The model is left with translation and synthesis, the work it is genuinely good at.
Your language lives in the system
When your fiscal year starts, what an active customer means, which names describe the same thing. All of it is written into the map, so the model does not guess it. It knows it.
No sentence without a source
Every answer shows the tables and records it was built from, and a question outside the map is refused rather than invented. A system that says it does not know is worth more than one that guesses.
The machine suggests, you approve
Every relationship, label and anomaly the system finds comes to you first, and what you confirm is never overwritten by the machine. The machine's guess and your decision are kept apart.
Try it now, see the result.
Sign in and ask your own question.