Demystifying Generative AI and its risks

Following the release of Helpdesk, Xelix’s brand new Gen AI tool for P2P, we invited Head of Data Science at Xelix, Paul Solomon, to join us for a chat down The AP Arms. Here, Paul provides clarity on the Gen AI topic, helping decipher Gen AI fact from fiction, whilst giving an inside scoop on the Data Science build of Helpdesk.

Watch the episode ▶️

First and foremost, what is Gen AI?

“Gen AI is a general-purpose model that can take an input and predict the next words to generate an answer.” Explains Paul.

The idea is to take our words and repurpose them to be indistinguishable from how a human would respond.”

Paul finds the ever-evolving nature of Gen AI super intriguing “We are still discovering and learning what it can do which is extremely exciting”

“We are finding that it can expand the repertoire of things you can do really quickly and generate a lot of value.”

However, the masses of Gen AI information in the market are both overwhelming and conflicting. So, it’s hard to decide what to take on board and what to ignore, especially when it comes to risk.

Paul lays out three risks you should address and how to overcome them…

Accuracy

Artificial intelligence, and specifically Generative AI, is still early in its inception. This means it can still be susceptible to mistakes as the models learn. For finance teams especially, slip-ups can be significant.

Therefore, having humans in the loop is still vital. Humans should oversee the system, picking up things only their expert knowledge can identify. Not every process has to be fully automated.

Paul challenges the misconception that AI is here to replace people. Instead, it should be used to take away time-consuming tasks from “the expert” (the person whose task we are trying the emulate). This frees them to focus on more progressive actions.

Data security 

Handling sensitive information requires a thorough understanding of your technology stack and knowledge of where your data is going. To mitigate these risks, it's crucial to keep your data enclosed and avoid taking excessive risks during the early days of adoption.

“A lot of our work on Helpdesk has been speaking with different cloud providers and understanding how our customers data will all contained (within the platform) and can we be confident it is secure”.

Over-reliance vs being left behind: striking a balance

The first step is identifying your low-hanging AI fruit, which can be addressed easily. Starting with small task automation that’s low risk, such as vendor query management, will help build confidence for more ambitious tasks, such as predictive analytics.

Doing nothing puts you behind your competitors and will impact your cost efficiency down the line.

Introducing Helpdesk

Helpdesk from Xelix, utilises Large Language Models (LLMs) to read emails, extract key information and generate intelligent responses that significantly reduce the AP workload.

Despite its advanced capabilities, our Helpdesk tool maintains a low-risk profile by keeping users in control of the process, whilst delivering high value:

• Faster response and resolution times
• Improved SLA compliance
• High vendor satisfaction
• Hundreds of hours back each quarter
• Fewer business disruptions (account on hold, orders not fulfilled)

Want to learn more about what Helpdesk?

Seal of approval

As is tradition, we asked Paul, who elicited the help of Gen AI, to create a meme to hang up in The AP Arms – here is what he produced...  🥁🥁🥁

Huge thanks Paul for joining us at The AP Arms and for sharing his brilliant insight into Generative AI.

Interested in popping in? Let us know – as a new establishment ‘round here, we’re always looking to welcome more guests!