Software & AI · From strategy to production

Our product · Bobby

Business AI Assistant: Bobby, ERP Data Through Conversation

Bobby is our business AI assistant. We designed it, built it and put it into production at Etixio: our teams query our ERP in plain language instead of clicking through screens. It is the AI we deliver to our clients, applied to ourselves first.

Status

In production at Etixio

What we built

Conversational assistant connected to the ERP

Users

Our teams

Technologies

Large language model (LLM), ERP integration, conversational UX

Where we started

Our data was there, but not always easy to find.

Active projects, logged time, time off, sales follow-up: the information our teams need lives in our ERP. Finding it still meant several searches, filters or exports.

Rather than adding another screen, we built an assistant: you ask the question, it fetches the answer from the ERP.

Bobby: our AI assistant, in production in-house

The challenge

An assistant that is useful every day, not a demo.

Simplify access to business data

Ask a question instead of navigating between screens and filters.

Answer from the real data

The answer must come from the ERP, not from the model’s general knowledge.

Respect everyone’s access rights

Each employee must only see the information they are entitled to, especially for HR data.

Get teams to actually use it

A simple interface built around the user’s question, with no change to how people work.

How a request flows
  1. 01Business question

    The user asks for the information they need in plain language.

  2. 02Bobby assistant

    The request is turned into a lookup of the ERP data.

  3. 03Answer

    The answer is shown in the conversational interface.

What we built

Business AI assistant: a model connected to an application.

We ran Bobby like a client project, from scoping to production. Each step produced a concrete deliverable, from selecting the use cases to an assistant connected to the ERP and used by our teams.

01

Use case scoping

Identifying the most frequent questions, the information to make accessible and the access rules.

02

Conversational design

An interface built around the user’s question and the readability of the answer.

03

Development & integration

Building the assistant and connecting it to the ERP’s business data.

Delivery

“Which projects are active?”

Access project tracking information.

Human resources

“How many days off do I have left?”

Check your own HR data available in the ERP.

Sales

“What do we know about this prospect?”

Find the data needed for sales follow-up.

Examples of questions from the areas Bobby covers.

Team and organization

Our team, our rules, our environment.

Bobby was designed and built by our own engineers, with the same practices we use on client projects: use case scoping, code review, testing, secrets management and separate environments.

Because we use it ourselves, we see its strengths and limits every day. That hands-on feedback directly shapes the choices we make on the AI projects we deliver for our clients.

Technologies used

Generative AI, ERP integration and conversational UX.

Bobby pairs a large language model (LLM), which understands the question and phrases the answer, with a connection to the ERP data, which supplies the actual information. The assistant’s quality depends as much on this integration as on the model: it is what lets the answer rely on company data rather than on what the model “knows.”

The ERP’s access rights also apply to the assistant: each employee only sees the information they are already entitled to, such as their own time off. The conversational interface was designed in its own right, so that both question and answer stay readable in everyday use.

Outcome and current status

In production in-house, used day to day.

  • Our teams query the ERP in plain language, without switching tools or changing habits.
  • Answers are based on ERP data and respect each person’s access rights.
  • Fewer searches, filters and exports to find everyday information.
  • Real-world use that feeds our generative AI practice on client projects.

We apply the same approach to our clients’ projects, such as the medical AI agent built for a hospital, and to taking AI POCs to production.

Frequently asked questions

Business AI assistant: your questions.

What is Bobby?

Bobby is our conversational AI assistant. We designed it, built it and put it into production at Etixio. Connected to our ERP, it lets our teams get business information by asking a question in plain language, without clicking through screens, filters or exports.

Is Bobby a client project?

No. It is our own product, used by our teams. We ran it like a client project, with the same engineering practices, and it serves as a day-to-day testing ground for the AI we deliver to our clients.

What data can Bobby access?

Business information in our ERP, including project tracking, HR data such as time off, and information used for sales follow-up. Each employee only sees the data they are already entitled to in the ERP.

How does a business AI assistant connect to an ERP?

The language model interprets the question, then the assistant queries the ERP data and returns the result in the conversational interface. The answer is therefore based on company data, not just on the model’s general knowledge.

Can you build an AI assistant connected to our tools?

Yes. That is what our enterprise AI solutions and AI agents and chatbots are for, from use case scoping through to production release.

Your business data can become more accessible.

Tell us about your users, your software and what you want to build or improve.

Book a 30-min call with a tech lead

What are you looking for?