Generative AI in Business Operations at A-CX

At A-CX, we have explored how to use Generative AI in our business operations. This is natural as we work on innovative solutions for software development, cloud computing, and digital transformation daily. In this blog post, we will share some ways we use Generative AI to enhance our work and deliver value to our clients.

Case 1 – Code Analysis with Generative AI

Challenge

When onboarding a new client, we audit or assess the client’s code. Depending on the work objectives, we may want to understand how well structured and reusable the current code is, understand its quality, or find risks and vulnerabilities. The code can be tens or even hundreds of thousands of lines long. Processing such an amount manually, even when using sampling, would take a long time.

Solution

Our first example of how Generative AI Helps A-CX in business operations is a customer case. We built a tool using a publicly available AI model to analyze the health of the code. The tool identified potential threats in the code and classified vulnerabilities as severe, high, medium, and low. We can define the tool to provide an executive summary, recommendations for the next steps, and even rate the code per section. Azure processes this in mere hours, and once the logic is defined, it runs a new report with a reasonable cost.

Benefits

This tool’s purpose is not 100% accurate, but it gives a good heat map of the codebase and suggests where to focus. It helps us speed up the value creation for our clients, whether that is an improved level of security or new features. It also saves us time and resources that we can use for other tasks.

Case 2 – Knowledge Extraction with Generative AI

Challenge

We often work with up-and-coming technologies, such as Microsoft Entra Private Access. As a non-GA product, limited public information was available online. We created an AI-assisted solution that could distill information from hundreds of pages of documents provided for us.

Solution

We built a tool leveraging a RAG LLM to query a knowledge base. We feed all publicly available Entra PA material from Microsoft or other reliable sources. It does not need to be documents only. In another instance, we digested video material so the LLM-powered tool could answer questions based on it. This is our second example of Generative AI in business operations.

Benefits

This helps us to stay updated and informed about the latest developments and features of the technology. It also enables us to create high-quality and accurate content for our clients and stakeholders.

Case 3 – Job Application Screening with Generative AI

Challenge

Some of our clients have very detailed requirements for people in their projects. It’s our expertise to have the right people available for projects, often on short notice. We have built an impressive roster of qualified specialists. To keep the roster up to date, we process hundreds of job applications monthly.

Anyone who has recruited for a highly sought-after company knows the challenges this brings. You struggle to find the time to go through the applicants. You also struggle to find people who both can and are willing to spend time helping you do that. Let’s be frank: going through tens of applications is mind-numbingly dull. This is regardless of how suitable the applicants are. With three minutes per application, you can cover 20 in an hour – if you manage to stay focused. 

Many of the applications can be disqualified on the spot. The applicants may be located in the wrong country, for example. Spending even a minute on them is time wasted. On the other end of the spectrum, you may have a superstar that you miss. Maybe your focus isn’t there anymore. The resume or CV may not be well written, even though the applicant is a perfect match for the role. There are many ways you could miss a diamond in the rough.

Solution

We built an LLM-based tool that compares the job description with the application. It provides a short summary, a bulleted list of the applicant’s pros and cons, and a star rating. More importantly, as Ilpo explained in a previous blog post, this all comes out in a structured format. 

This sounds simple, but there were a few catches. Uber, Google reviews, and Yelp have twisted the star ratings scale. 5 stars are expected; sometimes, giving anything but 5 stars is an insult. This super-compressed scale doesn’t work for situations where you want to create meaningful differentiation, but this is what an LLM has learned as a standard. Thus, the star rating had to be calibrated so that 3 stars are good, and you must stand out to reach 5 stars.

A keyword provided under actual work experience is a more meaningful indicator of the candidate’s fit than the same text provided under skills or another generic list. Stating you accomplished your latest project success with Java sets a higher bar than listing Java as one of the languages you know. However, the LLM needs to be guided to make this distinction. 

Once again, we were reminded that building an AI solution is largely creating a solution that leverages AI, but AI isn’t everything. Much of the work was getting the resume systematically and reliably exported from the HR tool, despite the file format. Saving the AI-generated summary to the tool was another necessary step. This naturally depends on the maturity of the APIs your toolset offers.

Finally, it’s crucial to remember that AI only guides us; we’ll make the final decisions as human beings. AI is developing fast, and the underlying technologies are often evolving overnight. A system that delivers wanted outcomes today may behave differently tomorrow.

Benefits

This tool speeds up our process a lot. It allows us to respond quickly to the best candidates and save everyone’s time when dealing with inadequate candidates. It also reduces the risk of human bias and error in the screening process. It’s worth noting that some legislations and plans prevent fully automated processing of people and their data. There are also obligations to inform people if fully automated AI is used. These rules are in place to provide fairness and transparency. We can configure the tool to follow the legislation and policies in place from time to time.

Conclusion

One of the biggest challenges companies face today is leveraging the power of artificial intelligence (AI) to create value and gain a competitive edge. AI offers tremendous opportunities for improving efficiency, quality, and customer satisfaction. Still, it poses significant risks and challenges, such as ethical, legal, and social implications, data privacy and security issues, technical complexity, and cost.

Many companies struggle to use AI in daily life. Some companies focus so much on the risks that they miss out on opportunities to build tools that balance the risks with benefits. Others need help moving from PowerPoints and plans to trials and prototypes. Some have the right ideas but don’t have the execution capability to design or develop. At A-CX, we encounter these challenges all the time.

That’s why we decided to explore the potential of Generative AI, a branch of AI that can create novel and diverse content, such as text, images, audio, or code, based on some input or data. Generative AI can help us solve problems that require creativity, personalization, or adaptation and can also enable us to discover new insights and possibilities that we might not have thought of before.

Generative AI is a powerful and exciting technology that can help us solve many problems and create new possibilities. At A-CX, we constantly explore new ways to use Generative AI to enhance our work and deliver value to our clients. We believe that Generative AI is not a threat but an opportunity to augment our human capabilities and creativity.

Please contact us to hear more. We’d be happy to explore how to harness AI to help your needs. Or, if you still need to get ready, we’d be glad to brag more or chat more about the AI-enhanced tools we’ve built.