What is Pionier AI?

Pionier AI is the intelligent assistant of the Nexus ERP ecosystem, a web chatbot platform based on artificial intelligence, developed by Nexus Media and specialized for the Romanian business environment. Its purpose is to reduce the response time for users' personalized requests: from days or weeks to a few minutes. The assistant provides technical support for Nexus applications based on official documentation, answers on economic legislation and accounting topics, as well as support for analysis and file creation.

This case study focuses on 4 concrete scenarios from the daily activity of a Nexus ERP administrator: users constantly come with requests such as "I want a new column in report x" or "notify me by email when stock decreases". Each of these requests is solved through a custom function, and writing it requires time, SQL knowledge, and familiarity with the Nexus database structure. We will follow step by step how the administrator uses Pionier AI as a technical partner in this process: from the request written in ordinary language by the user, to the custom function implemented in the system, calculated fields, blocking approvals, automated alerts, and personalized reports.

How we work with Pionier AI

Pionier AI is built on a language model (LLM) and, as such, is nondeterministic. For the same request, phrased identically, it can generate different answers from one session to another (equivalent formulas written differently, another code style, other details), unlike a classic program that always gives exactly the same result. Consequently, the examples in this case study are illustrative; they show the working method and the type of result we can expect, but they do not guarantee identical results.

The working method is iterative. Pionier AI is not, as a rule, a single exchange of one question and one answer. The best way of working is iterative, like a discussion:

• If the answer is not correct or the function returns errors, we must state exactly what does not work. We can indicate the exact error message, attach a screenshot with the error or the wrong result, and add further details to obtain the desired result.

• The clearer, more complete, and more exact the request is, the better the result. A vague prompt produces a generic answer; a prompt that specifies the module, report/document, fields, conditions, and expected format produces a much more correct result.

Methodological note applicable to all scenarios:

• Specify the exact context in the prompt: the module you are in, the targeted report/document, and the desired function type (calculated field / warning / blocking approval / report).

• Ask for field verification: add to the prompt "check the existing fields, with their exact names". Pionier AI can confirm the real field names before generating the formula. An optional addition that substantially increases response speed and accuracy is to provide the required fields in the prompt, either extracted as text or as screenshots

• Test before production: validate the function on a sample document, not directly on real documents.

• Pionier AI can make mistakes: always verify the received answers before putting them into use.

Introducing SC EXEMPLU SRL

SC EXEMPLU SRL is a distributor of electrical materials and household appliances. The company operates a central warehouse, two owned stores (receipt-based sales), and a B2B channel through which it delivers to certified electricians and installation companies.

SC EXEMPLU SRL uses Nexus ERP for its entire operational and accounting activity. As the activity grew, users started requesting adaptations: new columns in reports, validation rules, alerts, reports. Instead of relying on external consulting every time, the company's system administrator uses Pionier AI to generate these custom functions.

Scenario 1 - Calculated field: product name + product code on invoice

The invoicing manager's request: "Several customers asked us for the product code to also appear on the invoice, because they receive the goods based on SKU. Can it be displayed under each product name?"

Prompt for Pionier AI: I am in the Outputs - Customer invoices module and I want to modify the customer invoice so that the SKU code also appears under the product name. I entered the Reports - Configure reports menu and selected the "Customer invoice forms" report; I entered the menu for adding a calculated field. Generate the formula for the calculated field.

Scenario 1 conversation

Then, based on the formula provided by Pionier AI, we can create the required calculated field.

Scenario 2 - Calculated fields in BI: sales indicators

The sales manager's request: "In the sold products statistics, I want to see the agents' commission calculated based on the attached Excel file comisioane_agenti.xlsx, calculated from the net value after discount."

Detailed sales statistics report provided by Nexus ERP by default

Prompt for Pionier AI: I am in the B.I - Sold products statistics module and I need the "Detailed sales statistics" report to show a column with the sales agent name and a column with the related commission calculated based on the attached Excel file. I want only invoices to appear in the report. Tell me which fields I need to add to the report or generate the formulas for the calculated fields I need to create. Check the existing fields for this report, with their exact names for the formulas.

Scenario 2 conversation

Detailed sales statistics report obtained by following the instructions provided by Pionier AI:

Scenario 3 - Blocking approval for validating large invoices only by responsible users

The general manager's request: "Agents can issue and validate small invoices without problems, but I do not want invoices over 10,000 RON including VAT to be validated by just anyone; those invoices must be validated only by the sales manager or the director before they leave for the customer. (Vlad or Andrei)"

Prompt for Pionier AI: I am in the Outputs - Customer invoices module and I want to create a blocking approval so that only users Vlad and Andrei can validate invoices greater than 10,000 RON including VAT. Check the existing fields and generate the custom function for this blocking approval.

Scenario 3 conversation

Warning resulting from applying the blocking approval generated by Pionier AI

Scenario 4 - Scheduled email alert for a customer order with a short delivery deadline

The logistics department manager's request: "When an order comes in with a delivery date less than 3 days away, I want an alert on the warehouse email so we can prioritize it."

Prompt for Pionier AI: I am in the Outputs - Sales management - Customer orders module. I want a custom warning-type function that, when saving an order, sends an email if the delivery date is less than 3 days from the order validation day. The email, in an aesthetic HTML format (use black and gold colors) and easy to read, using the company logo https://i.imgur.com/Zunenpl.png, should contain: the order number, customer name, delivery date, total value with VAT and without VAT, and the list of products together with their price with VAT and without VAT. Recipient: depozit@exemplu.ro. Check the field structure in customer orders and generate the function

Scenario 4 conversation

After we follow the instructions provided by Pionier AI to create the custom alert, from the Customer Orders module we add an order with a delivery deadline one day away.

After validating the order, we will receive the order summary by email.

Throughout the 4 scenarios, the SC EXEMPLU SRL administrator transformed requests written in natural language into working custom functions, calculated fields, a blocking approval, and an automated email alert, without using external consulting.