← sources

Hi everyone, how would you approach building an internal AI assistant for a logistics company?

source: reddit (reddit:AI_Agents)published 2026-10-12 06:53ingested 2026-10-13 02:36

retained text

I’m a developer at a logistics company, and our management is starting to explore integrating AI into our internal systems.

We’re still in the research/planning stage, so I’m trying to understand how other companies are approaching this before we start building anything.

One example our CEO mentioned is an internal AI assistant that employees could ask questions about company data.

For example, someone from Sales might ask:

* “How much are our total sales today?”
* “What are our sales for this month?”
* “How does this month compare to last month?”
* “Which customers generated the most revenue this month?”
* “Can you summarize our sales performance this week?”

The AI would then retrieve the necessary information from our internal systems/database, analyze it, and return a simple summary instead of requiring the employee to go through reports or dashboards manually.

Eventually, I imagine the same concept could be used for other departments as well, such as operations, logistics, finance, HR, or management.

I’m trying to understand what the best architecture would be for something like this.

Some things I’d like advice on:

* Would you let an LLM query the database directly, or create a controlled API/service layer between the AI and the database?
* Would something like text-to-SQL be appropriate for this?
* Would RAG even be necessary for structured company data like sales?
* How would you handle employee permissions? For example, Sales may be allowed to see sales data while another employee shouldn't.
* How do you prevent the AI from running unsafe or expensive database queries?
* Would you use OpenAI, Claude, Gemini, Azure OpenAI, a local model, or something else?
* Should this be a standalone internal AI platform that multiple company systems can connect to?
* How do you make sure the AI does not hallucinate financial numbers?
* Would it be better to connect the AI to an existing BI system rather than directly to the production database?
* How are companies handling sensitive/internal data when using external LLM APIs?

I’m also interested in hearing from anyone who has built something similar.

What architecture did you use?

For example, something like:

Employee  
→ Internal AI Chat  
→ LLM  
→ Backend/API  
→ Company database / ERP / sales system  
→ AI summarizes the result

Or is there a better pattern?

I’m not looking to add AI just for the sake of having AI. I want to understand how to make something that is actually useful, secure, maintainable, and reliable enough for employees to use for business information.

Any advice, architecture examples, lessons learned, recommended technologies, or things we should avoid would be appreciated.

document fetched 2026-10-13 00:19 · canonical https://reddit.com/r/AI_Agents/comments/1x3uwpl/hi_everyone_how_would_you_approach_building_an