Exclusive contribution by Shairil Yahya, Legal Compliance Technology and Solution Director at Philips

The GenAI Dilemma: To Do or Not To Do (or how to do)

The answer is yes! We should jump onto the GenAI bandwagon. Some would say that it’s the biggest thing since sliced bread. There are a few things that I personally experienced that totally revolutionised the digital world: the invention of personal computers and the internet (and as someone who grew up in the 90s), The Matrix’s CGI. And now we have the GenAI dawn, or mid-morning, kicked off by ChatGPT.  

With the realisation that GenAI is here to stay and the riches it promises, every company dreams of the pot of gold at the end of the GenAI rainbow. But as with any adventures that promise immortality, they are fraught with dangers and boobytraps that lurk at every corner, dragons and trolls waiting to tear you apart right after every small victory.

As the reward is greater than the risk, companies will invest in GenAI. But what do you need to consider before embarking on the GenAI journey?

Based on my sleepless nights developing GenAI applications, below are three key factors you may want to consider. Let’s use one GenAI development example – building a GenAI bot that is able to query a monitoring platform using natural language – to discuss the key contributing factors.  They are:

1. Determine your starting point for the GenAI journey.
2. GenAI now sits atop the digital development tree. Therefore, developing lower-tier activities is necessary to reach the GenAI level. Without this, GenAI will not generate the correct result. This is because it relies on the information from the lower-level activities to feed its engine. Otherwise, Garbage in, Garbage out.
3. Each of the tiers in the diagram below deserves its discussion topic, but for simplicity:
   1. Data Requirements and Connection – Data comes in various shapes and forms. So, understand where your data sits, who owns the data and how to extract them.
   1. Building the logical coding – once the data is extracted, it needs to be cleaned and coded to meet your monitoring objectives.
   1. Visualisation & AI – you will then need to define how the results will be visualised and whether you want to include machine learning to enhance the result further.

It goes without saying that before diving into GenAI development, one must have a solid grasp of the lower-tier activities. Therefore, ensure that you allocate sufficient time to developing them.

• End product visualisation at the scoping stage.
• This may sound basic and un-sexy. However, even the sexiest system design starts with a well-defined scope. Be ambitious, but be realistic about those ambitions.
• Once the scope has been established, be as clear and descriptive as possible on how the end product should look, including high-level design of key activities to support the GenAI. To build a monitoring system to answer your questions in a natural language, design the conceptual visual user interface at the scoping stage. This would allow the consultant/technical team to understand what you want to achieve. Yes, in this instance, a picture does represent a thousand words or a thousand lines of code.
• The visualisation also helps the development team determine the scope’s complexity and set their time/resources/cost accordingly.

I have seen projects delayed or even collapsed, not because the objective was vague but because the developer team produced a totally different platform interaction from what the project lead had in mind, or the key team members were not able to agree on how the scope should be presented in the platform at the User Acceptance Testing (“UAT”) stage. Hence, start early with end-product visualisation.

• Continuous support from management.
• For most of us, GenAi is not a single-step journey, and it is likely not something that can be completed in a single financial budget year. Therefore, a budget has to be requested every year. Hence, continuous management support is essential to ensure project continuity.
• It is also likely that the management is not keen to commit a substantial investment in the project in one go. Therefore, the project has to show something tangible before embarking on the full desired scope. In this instance, the project lead should start with a smaller scope, reducing the variability of the data required and focusing on the most important aspect of the monitoring (e.g. if you want to monitor sales and enquire about the result using  GenAI, focus on a region or a product that you know has a solid financial process and clean data, rather than a global scope) to convince the management of the GenAI worth. It is always powerful to show a real result, even on a smaller scale, to generate the proper support.

While the list above is not exhaustive, I hope it provides readers with ideas for approaching any GenAI project. One last thing I would like to mention: not all GenAI solutions offered in the market are created equal. So, choose wisely.