Build AI applications yourself: how Onagero works
Three parts make up Onagero: a data layer that knows your company, an environment in which your people build by describing, and an operation with roles, releases and a log. All on your server.
- Build by describing
- Reviewed data layer
- Release by people
- On your server
How you work with Onagero
Pick an example or describe a task of your own.
With new employees we constantly lose track of who already has which access rights, documents and briefings. I don't want to buy yet another tool for it.
This is your application. The briefing now depends on the site, and only what still needs someone to confirm it is open.
| Step | Comes from | Confirmed by |
|---|---|---|
| Contract and documents | HR system | automatic |
| Account and email | IT directory | automatic |
| Occupational safety instruction | Training list | automatic |
| Briefing at the machine | Foreman | the foreman |
| Key and badge | Reception | reception |
Then first we define what really needs to be tracked and who is responsible for which step. After that we can see whether a small application is enough.
That is enough for a small internal application. Three things come automatically from existing data, two need someone to confirm them.
This is your application. The briefing now depends on the site, and only what still needs someone to confirm it is open.
Before we automate anything, I want to understand where the work really comes from. Show me the process and which systems you use.
Onagero can match the order, maintain the list and gather the documents. After that, the responsible person assesses the complaint.
Then I stop right there. The preparation runs automatically, and you get a complete case to decide on.
This is what a case looks like now. Order, photos and delivery note are already there. The assessment is yours.
FUNCTION_BLOCK FB_AuxSensor
VAR_INPUT Sensor : BOOL; END_VAR
VAR_OUTPUT Report : BOOL; END_VAR
VAR Filter : TON; END_VAR
Filter(IN := Sensor, PT := T#200MS);
Report := Filter.Q; (* display only, no intervention in the safety circuit *)
END_FUNCTION_BLOCK| Case | Expected | Result |
|---|---|---|
| Sensor reports | Display on the panel | passed |
| Signal flickers | no report | passed |
| Sensor fails | Line keeps running safely | passed |
Then let's not start with code. Show me the program too, so we first understand the inputs and outputs and the safety logic.
The sensor itself is straightforward. What is critical is that its signal in the program would affect the guard door release today. I would not build it that way.
If the sensor fails, the line must not become unsafe. So I route its signal into a separate block that only reports. The safety circuit stays as it is.
This is the block. It filters the signal and only reports, it is not allowed to do more.
In the simulation it passes all three cases, including the failure. It only goes to the line after your approval.
The three parts
The data layer
A data map describes your tables, columns, terms and relationships. For every column, the origin and review status are shown.
Build by describing
Your people say what they need. Onagero builds the analysis or application and shows which data each figure comes from.
Operation with rules
Roles per person and application, release by people, masked data in development and an exportable log.
One layer above all systems
Onagero replaces no source. It organises what is in them so that people and AI can work with it.
Known and inferred are kept apart. What your experts have confirmed counts. What was detected automatically is marked as such until someone reviews it.
Confirmed by your experts: this column is the authoritative one, this is how your company calculates.
Detected automatically and marked as such until someone reviews it.
- Relationships
- Terms and meanings
- Origin of every column
- Rights
- Reviewed knowledge
- Corrections
From request to running application
Five steps. The AI builds, people review and release.
- You
Describe
Someone from the department describes in chat what is missing, in the language of their own work.
- Onagero
Find data
Onagero finds the data that belongs to the question and says how reliable it is.
- Test stage
Build and check
The application is created in a test stage, on masked data. Your experts check, correct and describe changes in their own words.
- Human
Release
The application goes into production only when a responsible person releases it. Roles determine who sees and operates it.
- Production
Work
The application runs on your server with the released data. Users’ feedback flows back and makes the next application better.
More precise with every use
Corrections are not lost. They become the knowledge with which the next application starts.
- 1
Use
Someone builds an analysis or application.
- 2
Correct
Experts say what is wrong or meant differently.
- 3
Retain
Onagero stores the reviewed knowledge.
- 4
Build better
The next task starts with this knowledge.
- 5
Use more
More people build, and the cycle begins anew.
and around again, more precise with every pass
Questions about the platform
What applications can you build with Onagero?
Analyses, reports, dashboards and internal applications with input forms and roles, such as a quote costing tool, a scrap analysis or a delivery date lookup. The yardstick is a process that today consists of exporting, reconciling and retyping.
How does our data get into Onagero?
Through read-only access to ERP, BDE, CAQ, databases and files, which your IT releases. This produces an analysis copy on your server, described in a data map with meanings and relationships.
How does the AI know which figure is right?
From the data map: for every column there is a meaning, an origin and a review status. Knowledge reviewed by people is kept separate from what was detected automatically, and you can look this up for every column an application reads.
What happens to our people’s corrections?
They are not lost. Onagero takes into account what your experts correct in later tasks. That way the next person starts with the knowledge that is already confirmed.
Does the AI see personal data?
Not while building. During development the AI works on a copy that holds no personal data. Only a released application reads the data its roles allow.
Which AI is behind it?
Onagero uses the language models of an external model provider for understanding and building. Platform, data and applications stay on your server; the model receives only masked excerpts of individual requests.
Do we need a data warehouse first?
No. Onagero creates the analysis copy and the data map itself, starting with the sources your first problem needs.
“Like Excel without formulas.
Except you could build Excel with it, too.”
Onagero learns with you.
Whoever works with Onagero explains, along the way, how the work really runs. From that grows a map of the processes that stays current, because no one has to maintain it.
See it on your own process
In the demo we show what a process from your company looks like with Onagero. Thirty minutes, free of charge.
Schedule a demo →