A three-layer architecture built on top of your existing systems.
The platform reads data from source systems, interprets it across applications and turns the result into a decision with a named owner and a trackable action.
Data → Intelligence → Decision
- 01 Source records
Data
Collects and normalises data from source systems.
- 02 Derived indicators
Intelligence
Surfaces deviations, risks and opportunities.
- 03 Human decision and action
Decision
Prioritises, assigns an owner and measures the impact.
Data rises from below, decisions form in KERNEL, actions go down to the applications
Combines sources, surfaces deviations and produces reasoned recommendations; a person decides.
Shop-floor records, reconciliation and mapping with the ERP. In a company without an ERP, the first record of the work.
ERP · MES / SCADA · WMS · MS SQL, PostgreSQL · Excel / CSV · REST API · accounting software
-
A company with an ERP
The problem is not missing data but data split across the ERP, MES, warehouse software and Excel. The connection is built in four steps: a separate adapter per source (read-based; existing records are not changed, and IT is asked for a user with access to the agreed tables), code and identity mapping, shop-floor recording and reconciliation with OPAS, then KERNEL and the management screens. If the ERP changes, reports and business rules are preserved; only the adapter changes.
-
A company without an ERP
Orders, production, stock and shipping run on Excel, messaging and paper. OPAS keeps the first record of the work on the shop floor, and the data feeding KERNEL becomes orderly and time-stamped for the first time. OPAS is not an ERP: invoices, ledgers, payroll and tax stay in the company’s accounting software and with its accountant.
Scope: we tell a company plainly when all its data sits in one ERP and the ERP’s own analytics are enough. The platform’s real field is fragmented landscapes, such as several ERPs, a separate MES and a budget run in Excel.
What does “AI” mean here?
-
Documented functions
The intelligence layer surfaces deviation and anomaly signals and gives early warning. KERNEL interprets data from the applications in one pool, shows its recommendations with their supporting facts and answers questions asked in the user’s own words with answers grounded in the data.
-
What is not documented
The product materials do not state which of these functions run on rules, statistical methods or models, and they make no forecasting promise. Which functions are enabled in your installation, and the methods used, are shared with your organisation during discovery.
-
The boundary that does not move
The system recommends and explains; the recommendation goes to the responsible person as a decision card awaiting approval. People own the decision and its outcome. Source systems are connected on a read principle; any write-back is scoped during installation.
ERP-independent integration
The platform is not tied to a particular ERP; it is installed on top of the existing infrastructure through data adapters.
-
Adapter-based connections
Each source system has its own adapter. If the ERP changes, reports and business rules are preserved.
-
Read access
Connections are built on reading from source systems. If a flow needs to write back to the ERP, its scope is agreed during installation.
-
Mapping layer
Account, warehouse, customer and stock codes are mapped to the organisation’s own rules.
-
A pilot within weeks
The aim is for the pilot to run on live data within a few weeks of discovery.
ERPs named in the product materials
Other source types
For each installation, the systems to connect and the adapter for each are confirmed during discovery.
Security and access
-
Single sign-on (SSO)
Central authentication with OpenID Connect; one account for every application.
-
Role-based access
Permissions at module and record level; user activity is logged.
-
Encrypted transport
All traffic travels over TLS; sensitive settings are stored encrypted.
Deployment options
-
Cloud
A managed installation on Kubernetes, with automatic updates and backups.
-
On-premises
Runs on the organisation’s own server, on Windows/IIS or in containers; platform data is held on the organisation’s infrastructure. Email, messaging and external data connections are agreed during installation.
-
Licensed and branded
A licence specific to the installation; the organisation’s own logo and company name are shown.
Standard features: Excel export, automatic reports (scheduled email and instant-messaging group notifications), Turkish and English interfaces, mobile-friendly use.
How do we start?
The planned roadmap has three phases. Durations are agreed together according to the scope and the data sources to be connected.
- Target: 2–3 weeks
Discovery and preparation
Business priorities, data sources and the KPI set are clarified together, and the pilot scope is defined.
- Target: 4–6 weeks
Pilot
Deployed in the target unit; live data flows and the first decisions are taken on the platform.
- Ongoing
Roll-out
The flow that worked in the pilot is extended to other units.
Let’s map your systems and decisions together.
In a first conversation we discuss your business priorities, your data sources and a suitable unit for a pilot.