ErucaAI

AI built for your business. Build it, run it, own it.

AI that adapts to your data and runs on your own servers. The local model answers first; when that isn't enough, and only where you allow it, a cloud model steps in. Your data stays in the building and your costs stay predictable.

How does it work?
ErucaAI · Assistant● on-premises
Which departments had purchase requests piling up for approval last month?
[Summary table by department and the 3 longest-pending requests.]
source: satinalma.talepaccess: user's role
Answered by
Structured query
Local model
not needed
Cloud model
not needed
Why your own AI

Not a generic chat tool. A model that knows your business.

Your data stays with you The model runs on your own servers. You decide which requests may leave, and under what conditions.
It knows your business Fed by your tables, packages, documents and approved outputs, it doesn't guess; it cites its sources.
Predictable cost Common questions are answered without ever reaching a language model. Start on ordinary hardware, with just a CPU.
Solutions

Enterprise AI solutions.

From a custom-trained model for your company to an on-premise LLM, from document reading to AI agents: start with the solution you need and scale on the same platform.

Custom-Trained AI for Your Company An enterprise model adapted to your data, fluent in the language of your business, and better every month. Explore On-Premise LLM Run the language model in your own data center; you decide which requests may leave. Explore Enterprise AI Assistant Ask about procedures, documents and ERP data in plain language; get answers that respect permissions and cite their sources. Explore Document Data Extraction Reads fields from invoices, contracts and application documents and fills in the relevant form as a draft. Explore AI Agents Agents that route requests, complete routine checks and leave the exceptions to people. Explore Forecasting and Decision Support Sales and cash forecasts, risk scores and anomaly alerts, each with its reasoning. Explore
Build

AI tailored to your business, in four steps.

  1. 01 · Connect Connect your data ERP tables, documents and the Studio knowledge library are scanned; text and semantic indexes are built on-premises.
  2. 02 · Define Define the task Q&A, document data extraction, work routing or scoring. Every task gets its own rules and validator.
  3. 03 · Adapt Train it on your own work Approved AI outputs become training data; recurring tasks are tied to ready-made recipes.
  4. 04 · Measure Test it in shadow mode Compare a new model side by side with the current one on the same tasks, before it goes live. Let the data decide.
Run

Every question is resolved at the cheapest tier that can answer it.

ErucaAI first tries a request in tiers that need no language model. It moves to the local model only when necessary, and to a cloud model only if that falls short; each escalation carries the local draft and error details.

  1. Tier 0 Structured query Count, list and status questions answered with SQL, straight from authorized data.
  2. Tier 1 Semantic cache Recognizes a previously approved answer, even when the question is phrased differently.
  3. Tier 2 FAQ cards Ready answers written by your experts, with their sources.
  4. Tier 3 Recipes Ready, validated workflows for recurring tasks.
  5. Local model On-premises language model Runs on your own servers; every answer passes through source gates and validators.
  6. Fallback Cloud model Steps in when the local model can't handle a task, only for tasks you allow.
Deployment options

No expensive hardware required. Start small, grow with the data.

Recommended start Your own server Runs on CPU, on an ordinary server with 32–64 GB of memory. Move to GPU only when shadow-run data justifies it.
Hybrid Local first, cloud as backup Routine work stays in-house; harder tasks are routed to a cloud model according to your rules.
Managed Operated by us We handle installation, updates and monitoring in your dedicated environment. Scope and service level (SLA) are defined together in the contract.
Use cases

AI that works inside your daily operations.

In-house assistant

Employees ask about ERP data and procedures in plain language; answers come only from data they're authorized to see.

Document data extraction

Reads fields from invoices, contracts and application documents, and fills in the relevant form as a draft.

Natural-language reporting

“Which products had the most returns this quarter?” Answers as tables and charts, no query writing required.

Smart work routing

Assigns incoming requests, applications and tasks to the right person by topic, urgency, team workload and skills, and flags delay risks in advance.

Decision-support scoring

Generates risk and priority scores for applications, customers or suppliers from your own historical data, and shows the reasoning behind each score. The final decision always rests with a person.

Decision-support summaries

Turns a loan application, proposal or complaint file into a committee-ready summary.

Anomaly and risk alerts

Spots unusual invoices, payments or stock movements and notifies the right person, along with the reason.

Forecasting

Predicts sales, demand and cash flow from your historical data, so planning rests on data, not guesswork.

Process automation

Completes routine checks and approvals automatically by your rules, leaving only the exceptions to people.

Security and governance

You're always in control.

What the AI sees, what it decides and on whose behalf it acts is traceable at every moment.

Permission-based access
The model can't see any data the user can't see.
Personal AI identity
No shared keys; every call is tied to a person.
Full audit trail
Every request, response, source and approval is logged.
Validators
If an output breaks the rules, it never reaches the user; it escalates to the next tier.
Egress rules
You decide which tasks may go to the cloud.
KVKK (Turkish GDPR) compliance
Data stays within your organization; you set the rules for access, retention and masking.
FAQ

Frequently asked questions about ErucaAI

What is ErucaAI?

ErucaAI is an enterprise AI platform that is adapted to your company's own data and runs on your own servers. It offers solutions such as an internal knowledge assistant, document data extraction, AI agents and forecasting on a single platform.

What is the difference between enterprise AI and general-purpose chatbots?

General-purpose chatbots answer with what they learned from the internet, and your data goes to the vendor's servers. ErucaAI is adapted to your data, follows your permission rules, shows its sources, and your data never leaves your organization.

Do I need Eruca ERP to use ErucaAI?

No. ErucaAI can also connect to other ERP, database and document systems. When used with Eruca ERP, setup is much faster because tables, permissions and processes come ready.

How should we start an AI project?

We recommend starting with a single measurable business problem, such as reading supplier invoices or Q&A over procedures. In a 4–8 week pilot we connect the data, measure in shadow mode and review the results together, with numbers.

Is ErucaAI KVKK compliant?

Because ErucaAI runs on-premise, personal data is not transferred abroad; access is permission-based, personal data can be masked, and every request is written to the audit trail. This setup makes it easier to meet your KVKK obligations; we recommend doing the legal assessment with your own advisor.

Let's build your first model together.

Pick a real business problem; in a short pilot we connect your data, measure in shadow mode and review the results together.