Refence: Automated business code creation
Company
Lojer Oy, Finland
Time
January 2024
Case
Lojer Oy is an expert in hospital furniture, employing 260 people and with a turnover of 61M euros. Lojer’s products have been delivered to 120 countries.
Neptunux implemented AI Data Engineering (Aidde) for Lojer to free the company’s IT department from routine tasks. The purpose of Aidde is to create code that enables the company’s financial management to access the financial reports they need.
Aidde was implemented using the Microsoft 365 Copilot assistant already available in the company, and Aidde operates on a chat-based system. Aidde was taught the data structure to be used and the special terms used by the financial administration. After this, Aidde understands the language used by financial management and can generate the code for calculating reports. Aidde does not operate independently; the IT department checks the code produced by it and installs the approved code into production between the Data Lakehouse and the data mart.
The idea for Aidde came from Repe Mäensivu, Development Manager at Lojer Oy.
Aidde, Made in Neptunux.
Business case: Automated CV and skill profile updating
Industry
Leadership and financial consulting
Requirements
Company data cannot be visible to others and cannot be used to train LLM.
Time
January 2024
Case
The customer provides leadership and financial consulting to public companies. The length of consulting contracts varies, and at the end of each contract, the consultant’s CV and internal skill profile are updated based on new skills acquired from the concluding contract. This updating is a manual task, despite all the necessary information being available in the existing skill profile, CV, and billing information. The customer asked Neptunux to investigate whether this updating process could be automated with AI.
Solution
irst, we estimated the business value of the case, and it amounted to a few thousand euros per year. This is precisely the right size of a case to start the utilization of AI in the company, especially since AI was not previously in use and was unfamiliar to the company.
Because the business value was small and all the documents were in the Microsoft 365 environment, an off-the-shelf solution should be found. Considering the requirements, Microsoft 365 Copilot, whose tenant architecture keeps the company’s data protected, was selected to create a proof of concept.
Microsoft 365 Copilot was able to identify missing skills from the skill profile and CV and presented them, but automatic updating of those files was not possible.
Conclusion
The AI solution was not adopted, and the main reason for rejection was the nature of generative AI (GenAI) being fully probabilistic. This characteristic leads to the outcome that GenAI’s results are always somewhat random and must be checked, which contradicts the requirement for automatic updating of files.
This business case requires a precise solution: ‘Identify skills that exist in document A but not in documents B and C, and then write them to B and C.’ An exact solution requires logic, which is currently a weakness of GenAI. This business case can be easily solved with traditional software, which excels in logic.
The main outcome from this business case was that the customer initiated its own AI learning journey and did so correctly by taking baby steps. Now, they understand what kind of problems can be solved with the current AI technologies and are ready to utilize AI as it acquires new capabilities.
