How to get talent KPIs, skillset breakdown, and more custom reports using GPT Bot for ERP systems

Find out how you can use GPT Bot to get custom reports in 30 seconds or less.

undefined
7 min read
Business intelligence
Digital Transformation
Machine learning
Tools & techs

Many companies integrate analytics solutions such as MS Power BI, Google Looker, Tableau, etc., on top of their existing systems, and we do the same. This approach is reasonable for generating common reports and dashboards with key metrics that should always be at hand. The game changer in using an AI bot lies in its application for custom reports and analytics that are not needed daily but may be required quite urgently, eliminating the need to build a separate system. 

Hi, I’m Vitali Yurkevich, CEO at Symfa, and today I’m going to share with you the details of using GPT Bot to get custom reporting from our ERP system to see:

  • Talent KPIs
  • Project history and client relationship evolution
  • Detailed company skillset and more – all based on our internal company data. 

ChatGPT empowers regular employees, not just senior project management and business intelligence professionals, to generate actionable insights based on the company’s data. 

Make sure to read the article by Ilya Mokin, Head of R&D at Symfa, on the same project, and stay with me to see why I trust GPT Bot to help me with business reporting.

Table of Contents

  • Why use GPT Bot to do performance analysis?
  • What Symfa's in-house ERP solution was built for?
  • A very short description of the ERP modules and available functionality
  • How does GPT Bot help tackle the reporting format and accessibility issues?
  • The challenges you’re going to face when implementing GPT Bot
  • Let’s draw some conclusions: It’s not all roses with AI.
  • Three last things you should know before building your own reporting AI-driven bot

Why use GPT Bot to do performance analysis?

The right question is – why not?

Similar cases abound, as businesses are increasingly using AI to enhance decision-making and empower their employees. L'Oréal, for example, plans to leverage AI and its vast beauty data to identify new solutions and drive creativity. Similarly, Procter & Gamble has developed an internal AI tool with strong intellectual property safeguards. Mercedes-Benz Group is piloting the use of ChatGPT to streamline its production process.

We have integrated ChatGPT in our ERP system and have achieved great results. You can connect ChatGPT to any other APIs – CRM, Jira, you name it. Moreover, it’s not about the performance charts only. You can get any kind of report with any data that your system does not easily allow you to get, too.

Yet, in order to see how you can replicate our own case, you need to have some context on how our ERP was built.

What Symfa's in-house ERP solution was built for?

Our proprietary ERP solution was developed to provide a holistic picture of our talents performance against the company revenue. To enable this, we’ve developed eight modules, which are as follows:

  • Worklogs
  • Resource Management
  • Project Management
  • Reports
  • Analytics
  • Finance
  • Dashboards
  • Settings

A very short description of the ERP modules and available functionality

We’ve developed this ERP system to fit our unique needs with special focus on performance and financial reporting, so the data availability isn’t an issue for us (but may be for you, if your software wasn’t built with detailed financial and performance reporting in mind). The problem that GPT Bot helps us solve is the reporting format and accessibility.

See a brief overview of the Symfa's ERP functionality below.

Description of the Erp Modules

How does GPT Bot help tackle the reporting format and accessibility issues?

Imagine an HR manager using our ERP's UI to generate reports. Sometimes, the available UI options may not produce the desired report format or the insights format may not be suitable. In such situations, the HR manager can leverage the power of ChatGPT.

ChatGPT can analyze the data from various tables and generate a graph or any other desired visualization. It can also group the data in a meaningful way to provide valuable insights. The manager simply needs to prompt ChatGPT with the specific report requirements and the bot will handle the rest, including calling the necessary APIs, retrieving the data, grouping it, and generating the report.

See the custom-built reports that a user can make with the help of the bot.

Worklog analysis of a specific developer:

Mail.google.com Chat U 0  (3)

Worklog analysis visualization for the same developer:

Mail.google.com Chat U 0  (4)

Revenue forecast for the next half of the year (test data):

Revenue forecast by GPT Bot

Let’s compare this laconic, yet cohesive visuals with the reports that our ERP functionality could provide us with.

Revenue forecast for the next half of the year, ERP-in-built functionality (test data):

Revenue forecast Symfa ERP

Impressive, isn’t it?

However, fine-tuning and testing efforts should be significant to ensure accuracy and reliability. Let’s have a closer look at the AI development challenges in the next section.

The challenges you’re going to face when implementing GPT Bot

Here are the drawbacks to ChatGPT-enabled solutions that you must be wary of:

Code Stability:

  • The bot may not be able to process a large amount of data returned by the API.
  • The bot's behavior in determining the number of API requests is unpredictable.
  • It may make multiple requests, potentially wasting API tokens.

Data Omission:

  • The bot may omit a significant portion of the retrieved data.
  • It may only insert a small portion of the data into the generated code.
  • The resulting graph or report may be inaccurate or incomplete.

To address these issues:

  1. Utilize function calling as suggested by OpenAI for better control over the behavior of the bot.
  2. Identify and block APIs that cause the bot to freeze.
  3. Prepare a dataset for your API and fine-tune the model to achieve more accurate and reliable results.

Let’s draw some conclusions: It’s not all roses with AI.

For example, some reports require up to 20 attempts to build, depending on the complexity of a request. Eventually, we’ve learnt how to put those requests properly so that the bot could understand them correctly. The bot in its turn is learning fast, too, and when we’re done with testing it on our set of data, we’ll present the solution to our clients.

Do I see the AI-enabled bot as an optional tool or as a must? You can see from the graphs above, it outperforms our in-built ERP functionality for reports visualization in certain requests. However, it still needs a lot of training. See this very simple example below:

Report on the engineers on bench going on vacation next month (in-built ERP functionality, the engineers going on vacation this month are highlighted in yellow):

Symfa's ERP reporting functionality

Let’s see how GPT Bot tackled this one:

Reporting with GPT Bot for ERP

As you can see, it’s a great smart assistant IN THE MAKING, but so much more work is lying ahead.

 

Three last things you should know before building your own reporting AI-driven bot

Here are just three small tips for you before you’re gone implementing GPT Bot in your own ERP or CRM.

1. Clean data and persistence are crucial

Clean data and proper training are crucial for unbiased insightful business reporting. To make Gen AI models serve their unique needs, most companies use their own data and build data models behind their firewalls. 

However, using pre-trained models and bringing them into your environment to build a foundation model is a rule of a thumb, as this is too much work if you do it from scratch. 

2. MLOps investments – both time and money – are going to be substantial

Infrastructure is going to pose another challenge – you’re looking at unifying and integrating data at a scale. Here we’re talking about MLOps – how you make sure that those models are trained, deployed into production, maintained and monitored in the long run.

3. Talent and knowledge management strategy are the cornerstones of success

Training your models on complete, up-to-date, accurate data will ensure reliable results. A strong talent pool and proper knowledge management strategy are even more important than having the technology at hand.

Don't miss out! Follow us on LinkedIn and X to be the first to know about our latest updates. We regularly post about the latest tech trends and provide transparent insights into our software development process.

Credits

Vitali Yurkevich

Chief Executive Officer at Symfa

Chief Executive Officer at Symfa

More Like This

BACK TO BLOG

Contact us

Our team will get back to you promptly to discuss the next steps