AI for execs: How to cut through the noise and deliver results

AI has dominated discussions, not only in the global tech scene but in the business world at large. Such has been the impact of applications like ChatGPT and DALL·E that consumers are now fully aware of the wealth of possibilities large language models (LLMs) and generative AI offer. Indeed, according to research by AppRadar, new AI apps have been downloaded 23.6 million times by Android users since November. More than 700 AI startups have received a combined $7.1 billion in funding in the last three months alone. Very few tech innovations have managed to capture the imagination of the tech, investor, business and consumer worlds simultaneously.

Given this wide-ranging interest and appetite, there are unprecedented opportunities for businesses to experiment with and adopt new AI-driven solutions. However, such is the breadth of potential applications available — everything from customer service to supply chain financing — that decision-makers and investors alike are presented with the challenge of deciding which horses to back and when. After all, those that may have recently committed resources to metaverse-adjacent tech or blockchain only to find that real business value is a long way down the road may be reluctant to follow the latest hype.

Of course, the reality is that although ChatGPT may have brought AI to the mainstream, generative AI is actually just the latest advance in a plethora of data-science-driven applications. The insurtech industry, for example, has been transformed over the past ten years by data solutions that have automated processes, helped to digitally process risks, increased volumes and ultimately improved the customer experience.

I would imagine that, for many people, insurance companies would not be the first legacy business vertical that you would associate with embracing cutting-edge tech. However, the key for these institutions is that they can immediately see the logic and business value of AI solutions. For a relatively small outlay and minimum risk, they can quickly and tangibly transform large aspects of their business. And that is the fundamental rule when we consider the best opportunities for LLMs to make a serious impact on businesses: What can they use that will give them good ROI with minimum risk?

Tried and tested vs. bleeding edge

For decision-makers at large enterprises, LLMs (and AI in general) present a head-scratching number of options. Every single business function can get the AI treatment. The first thing to consider is the differing maturity and development levels of each solution. It can be attractive to experiment with the latest innovation or create your own unique use cases, but this naturally carries some risk. Often, out-of-the-box gen AI solutions (e.g., ChatGPT) present risks that make them unusable for certain enterprise use cases. Decision-makers should think of these capabilities as a toolkit available to accelerate their vision while ensuring that the correct technology is used depending on the nature of each application.

For example, fintech startups have a long track record of using data science to create sophisticated solutions that reduce the burden on finance departments and equip business leaders with real-time insights. Some of the latest advancements have concentrated on AI-enabled cash flow analysis and forecasting. Given the experience of many of these service providers, their products are likely to be more tried and tested — further reducing the risk of AI running amok.

Where are your key business pain points and inefficiencies?

Ultimately, the best approach is to start with the problem rather than the exciting new AI solution. We recommend taking new technologies as building blocks to create enterprise-ready solutions that address real tangible pain points.

Businesses can always further increase their efficiency, improve customer experiences and reduce pain points. Identifying where these are most needed will enable you to deliver the best ROI on your new AI solution. To do that, you need to look at your internal data as well as team and customer feedback. From there, you will be able to narrow your search for AI solutions.

Start small and get the AI infrastructure right for your business

Any new technology carries question marks around exactly how it will integrate with your existing business processes and infrastructure. The rush to get on board the AI train will inevitably lead to some companies getting derailed because they simply do not have the tech stack or internal expertise to effectively use their new solution.

AI systems will work effectively only if the data they use is free-flowing, complete and clean. In many organizations, this is simply not the case. Data management infrastructure can too often be overlooked. Often, information is siloed within departments, platforms are unable to easily share or analyze data, and data collection and management policies are inconsistent. Bad data will lead to bad AI.

Starting small using AI in a contained setting or use case will enable you to feel confident that your infrastructure, policies and processes are capable of more widespread adoption. It also has the virtue of more easily enabling team and management buy-in by reducing initial expense and potential disruption. There are many specialized third parties you can use in a targeted manner to quickly kick off these initiatives.

Don’t forget human oversight

There is a serious data skills shortage that will impact the ability of businesses to effectively adopt AI tools. Basic data education throughout a company is required to identify the most applicable solutions, properly monitor and verify their outputs and use these systems in the most effective ways. Businesses should not blindly trust what AI tells them; they need skilled human oversight. This expertise can not be held solely in the data team — it needs to be from the top down and right across every department.

This model is what is often referred to as the “human on the loop” model, where systems do not rely on human input to perform their activity (as traditional “human in the loop” systems did) but instead push human control farther from the center of the automated decision-making, playing a review role in ensuring the output is accurate and reliable.

Which solutions should I bet on?

Currently, the most talked-about new use cases for generative AI are within marketing — particularly copy and imagery generation. It’s natural that many enterprises will look at applying gen AI here first.

However, as we have discussed, any new tech attracts businesses dreaming about new use cases, which often results in existing use cases not making significant progress. Our recommendation is to think about how AI can accelerate progress in resolving existing pain points, which often do not require the generative component (with its challenges of hallucination) but instead rely on the foundational understanding of unstructured data.

Remember, identifying the best AI solution for your business is only the first step. You need to have the infrastructure, buy-in, internal expertise, and checks and balances to ensure you get the most out of it.

Juan de Castro is COO of Cytora.