Around the table were representatives from the following sectors:

• healthcare
• Retail
• Technology/Telecommunications
• Finance
• Producing
• Media
• Marketing and Communications
• Hospitality

The table asks “What can we actually do with artificial intelligence?” Similar challenges such as were also presented. “What are the practical use cases of AI?” The answer to the question is underpinned by concern for the legal and ethical use of relevant data. Artificial intelligence may have been a buzzword for the last two years, and it can be hard to avoid seeing it in the news or from vendors, but the uncertainty of whether to use it to the benefit of business is clearly on the minds of many. organizations.

Perhaps there are some success stories that show that the topic of AI has broken out (or is breaking out) of the hype cycle of any new technology. Examples of how artificial intelligence has been successfully applied in daily business activities were given:

• Using AI to compile free-text survey data and present feedback summaries in the form of a conversation between two people discussing the results. This saved valuable time and resources and summarized feedback in an easy-to-use format.
• Using Generative AI services to translate materials and save significant amounts of money in the process. The quality of the translations was high enough to be an acceptable long-term alternative.
• Using Generative AI to produce marketing and sales images of products rather than bespoke photography for online sales.

But what happens when it’s implemented well and results in real-world digital transformation?? As one guest noted:

“Reducing workload and gaining efficiency is great, but only if that time is spent working on the job and not just drinking a cup of tea.”

For example, just because there are efficiencies that amount to 20% of time savings does not mean that there will be a comparable increase in productivity or profits. This element highlights the importance of governance to ensure the realization of AI projects. Projects that are pursued and invested in are not just vanity projects, they are actually transformative for the business. There was consensus that the business case for an effective AI project needed to have a very high bar.

But the first thrust of the speech was clearly about the democratization of AI versus governance; In other words, should we let our users come up with the work for AI and find use cases through unregulated experiments, or should it be strictly governed? Also, what is the role of government and regulation in this expansion and adoption process? There were concerns that AI models available for free could be located in unsuitable countries, violate privacy laws, or even be unaware of what data the model being used was actually trained on. Were the data collected ethically? Has it been properly disinfected? And is the data collected actually suitable for use in the specific use cases in which it is used?

While there is broad agreement that governance is vital in an emerging technology such as AI, it has become clear that the pace of technology is significantly faster than the pace of governance, with organizations and governments alike struggling to keep up with the pace of change and revolution. With all this, it was noted that “governance done right” actually allows organizations to act quickly without falling foul of laws and regulations, both now and in the future.

The concept of Reinforcement Learning Human Input has been introduced as an approach to implement and support governance efforts that enable input and output from any AI model to be filtered through a human and reinforce the AI ​​model accordingly. Therefore, the human is the primary control that ensures fail-safe human in the loop. It also highlights that the key output of AI is now data, knowledge and intelligence (although this is changing very rapidly, see above); It is emphasized that what really matters is the human interpretation of the output.

Meanwhile, there was a small but vital conversation about how and where AI models are used; It’s fast to use general AI models like Chat-GPTIt’s cheap and easy, but there are privacy and security concerns about how the information organizations upload is used elsewhere. The lack of governance dictates that many use cases are piloted and deployed in public models when they should be deployed in private models. AWS and technology consultants (like SoftServe) have many secure and easy-to-deploy custom models that organizations can use to test and build securely.

Using these external AI models introduces other challenges, such as ensuring that the source of the model is fully understood, that it is not poisoned (maliciously or unintentionally) or harmed by producing unexpected results due to the model being unsuitable. full understanding. (One guest discussed testing an AI model to review and score CVs where test applicants containing English names and US universities consistently score much higher than other identical CVs for selection purposes; this is just a case waiting to happen !

The importance of a data governance-focused approach to AI was a key part of the conversation. Given the low barrier to entry for AI use, it has been recognized that putting the “house” in order is a prerequisite for any adoption to ensure an organization reaps the greatest benefit. data. In practical terms, this means robust records management and data classification of all unstructured data in the business. The challenge, of course, is that in the vast majority of cases, these activities are not prioritized until external auditors, clients, or regulators need to be involved, resulting in leadership reacting to “solve the problem.” It was recognized that good data management was a cultural issue rather than a technological one.

Whatever you do, however you do it, we are only at the beginning of something that is already making huge waves around the world, and what was impossible five years ago is now not only possible, but within the reach of every household in the world. World. Without at least some guidance in the form of effective management regarding its adoption and use, the unintended consequence may give us more than we expect.