AI in ERP: What a Local Assistant Can Really Do

Hardly any software today comes without “AI.” In the ERP context, however, the question is more concrete: What does an AI assistant really deliver in day-to-day operations — and where does the data go in the process? This article separates value from marketing.

Where AI Genuinely Helps in ERP

An AI assistant in an ERP system is not an end in itself. It becomes useful for specific, recurring tasks:

• Data queries in natural language: “How many products in category X are below minimum stock?” — instead of building filters yourself.
• Creating product descriptions: SEO-ready texts for many products at once, instead of writing each one manually.
• Summaries: condensing long transactions or documents down to the essentials.

The value emerges where AI takes over routine work, not where it is expected to make decisions.

Cloud AI vs. Local Model

Most AI features send data to external services (OpenAI, Google, and others). For an ERP system holding customer, order, and financial data, this is sensitive: that data leaves the premises and ends up with a third party.

The alternative is a locally operated language model — for example via Ollama, which runs LLMs on your own hardware.

	Cloud AI	Local Model
Data location	External provider	Your own infrastructure
Data protection	DPA/third-country transfer required	Data stays in-house
Ongoing costs	Per request/token	Your own hardware
Model control	Provider-dependent	Freely selectable

For data-sensitive companies, the local model is often the cleaner choice: the ERP data that the assistant uses as context never leaves your own environment.

Realistic Expectations

Staying honest pays off: a local model on standard hardware is not as powerful as the largest cloud models. For typical ERP tasks, however — structured data queries, text generation, summaries — it is more than sufficient. Anyone expecting highly complex reasoning tasks should calibrate their expectations.

How VertooERP Implements This

The AI module of VertooERP uses a local LLM via Ollama. The assistant answers questions about ERP data with context (products, orders, suppliers) and creates SEO-optimized product descriptions — individually or in bulk. Because the model runs locally, the data it uses stays within your infrastructure. The model and endpoint are configurable via the app settings.

Conclusion

AI in ERP is valuable when it takes over concrete routine work — not as a buzzword. For companies that care about data protection, a locally operated model is the logical choice: the benefit of AI without the price of handing over sensitive data. What matters is assessing its performance realistically and deploying AI where it saves time.