Business Leaders: Asking the Right Questions About Generative AI

A majority of business leaders don’t have a PhD in machine learning or artificial intelligence. Even without formal ML or AI credentials, there are many ways to lead in the generative AI movement. If you’re a business leader looking to implement generative AI or “GenAI”, it might seem natural to take the same approach that you have for other technologies. But GenAI is different, for three main reasons.

  • It can do so much. At our own firm, supported by a $1 billion investment, GenAI is transforming entire processes and functions and soon will transform business models too. In some areas, we’re seeing it boost productivity by as much as 40%.
  • It can scale incredibly fast. You can often deploy a single GenAI model with a similar “pattern” of training in multiple functions and lines of business. That’s different from conventional AI, where you often need a new AI model for each new use case.
  • You don’t need to build it. GenAI usually involves adapting models that someone else has built. Increasingly, it’s also becoming embedded in major enterprise applications. That can dramatically speed up deployment and reduce costs.

In light of these and other GenAI differences, many sensible questions that business leaders ask about this technology simply don’t apply. Here are the top seven questions we’re hearing from non-tech leaders and why you may want to reframe how you think about implementing GenAI into your business.

Question #1: What’s the single best use case to start with?

We often get asked what the single best use case is to start with. GenAI is so scalable and it’s usually a missed opportunity to focus too closely on any one use case. Instead, focus on how a single, repeatable “pattern” of generative AI deployment can apply across your value chain. For example, generative AI’s capacity for deep retrieval — extracting actionable insights from unstructured data — may deliver only modest value in a single function. But if you rapidly roll out deep retrieval in every line of business and every function, from compliance to human resources, the ROI can be spectacular.

Question #2: What proof of concept should I consider? 

Since you don’t have to build your own generative AI model — they come “pre-trained,” requiring only adaptation and customization — there’s often no need for proof of concept. Instead, you can frequently take advantage of the models’ off-the-shelf capabilities, perform some customization, and go straight to a pilot. If you do prefer proof of concept, it can often be brief — lasting only a few weeks before your pilot kicks off.

Question #3: How many roles can we consolidate?

That is not the right way to think about GenAI and we are not seeing — or expecting — big job cuts due to GenAI. Instead, we’re seeing demand for new GenAI-specific roles and a surge in the work that existing workforces can perform. We have, for example, seen a tech company use GenAI to help their legal team examine over six million contracts for possible overpayments. That quantity of oversight wouldn’t have made financial sense before GenAI was there to help. Workers get it: In our Global Workforce Hopes & Fears survey, most foresee AI as having a mostly positive impact on their jobs.

Question #4: How should I think about risk when it comes to GenAI?

Generative AI does pose certain new risks. But it’s wise to think less about managing risks, and more about trust-by-design. Your GenAI deployment can start with governance and security, embed oversight to validate outputs, and include a framework to monitor ROI and support trusted, ethical use. Determining an approach to responsible AI should cover strategy (for the CEO and board), control (for risk and compliance officers), responsible practices (for information and information security officers) and core practices (for data scientists and business analysts). 

Question #5: Should I hire more AI talent?

The effective and trusted use of GenAI certainly does depend on specialized skills and you’ll need to hire or develop your tech team. But since you don’t have to build models from scratch, enterprise-wide deployment usually requires fewer hard-core specialists than conventional AI would. What will likely be more important is upskilling your current technology and business professionals. Many may need new skills to adapt, oversee, and use GenAI, whether in your custom models or as embedded in enterprise applications.

Question #6: How can I catch up with the competition?

GenAI isn’t new. Many companies have been using it for several years, but GenAI models that are amenable to business use at scale only hit the market in 2023. So, no one has too much of a head start. Past experience with conventional AI doesn’t always help, since GenAI is deployed and used so differently. A competitive edge will come from learning new ways of working that take full advantage of GenAI and — critically — from quickly developing the new business models that GenAI makes possible. 

Question #7: Which publicly available GenAI model should we use?

Public GenAI models can be powerful, but you almost certainly shouldn’t use them in the enterprise. Instead, license and customize private versions of these models. A private version can enable you to safely input your data and intellectual property, as well as top insights of your top people. All your people can then have a GenAI “co-pilot,” equipped with the best of your organization’s expertise. Naturally, that will require data governance and cybersecurity tailored for Gen AI’s needs, as well as “data pipelines” and updated APIs. Your business will also take advantage of the invisible GenAI that’s being baked into all your business applications, including ERP and CRM.