AI and generative AI in 2023: Four top questions answered

What is generative AI?

We usually define AI as computer systems that can gather information from the digital or physical worlds, draw conclusions, then make smart choices and act on them. Generative AI does all that and goes one step further. It is a type of deep learning that can create content. Since it often works on plain language commands, it can be remarkably easy to use.

The business-relevant content that generative AI can generate includes:

• Software code for business processes
  
• Text, images, audio and video that are high quality
  
• Data analysis
  
• Transcriptions, translations, summaries and analysis of business documents, phone calls and meetings
  
• Virtual simulations such as digital twins and metaverse spaces
  

What are generative AI’s business applications today?

As it matures, generative AI’s applications will likely extend to almost every area of business. We’ll see workplaces and internal operations transformed and new business models created as skilled people instruct well-supervised generative AI models to create content. Just look at some of its more valuable and accessible business applications. 

• Customer service. Cut costs through automation and enable self-service that actually satisfies rather than irritates through true personalization and rapid, accurate responses to questions and concerns.
  
• Automating high-volume tasks. Whether it’s processing insurance claims, meeting payrolls, creating “first drafts” of software code or technical writing, you can automate much of the tedious, repetitive knowledge work that humans currently do.
  
• Provide people with insights. Generative AI’s ability to read, listen to, synthesize and analyze text and voices can give your teams a start on the information they need from things like contracts, invoices, customer feedback, corporate and government policies.  
  

At PwC, for example, we’re already using generative AI to turn large volumes of data into richer insights and recommendations for our clients. We’re also using it more and more to find efficiencies and cost and time savings for ourselves and our clients.

How does generative AI work?

Like conventional AI, generative AI runs on models — sets of algorithms that are trained with human help to produce desired outputs. Based on this training, the AI attempts to predict the better answer to the prompt or command given. 

For example, you could say, “Listen to all the calls to our help desk in the last 24 hours and identify the three biggest complaints.” If the AI model was well trained, its output would likely be accurate because of all the data it had already learned from: other human voices, other complaints, other human language in general.

Generative AI’s answers aren’t always perfect, but they can be remarkably helpful, in part because of its differences with conventional AI. 

• Model complexity. Generative AI runs on foundation models, which use some of the most advanced AI techniques in existence. That can make its inner workings opaque, even to its creators.
  
• Model ownership. Generative AI’s foundation models are so complex, only a few specialized companies are making them. Everyone else is selecting one or more of these models and adapting them to their needs. That can speed up deployment and help reduce costs.
  
• Richness of data. Generative AI’s foundation models are generally trained on truly vast quantities of data — hundreds of billions or even trillions of data points. These models, which are the source of generative AI’s often remarkable accuracy, can represent the collective wisdom of the internet.
  
• Data ownership. For any company, accessing such a huge amount of data isn’t possible if they limit themselves to proprietary data. Generative AI’s foundation models are all designed to include mostly or exclusively open-source data.
  

How can I accelerate AI and generative AI in my company — while building trust?

To use AI and generative AI to deliver near-term outcomes, transformative innovation and increased trust is not an easy task. Three guidelines can help. 

1. Implement responsible AI to build trust. Responsible AI offers frameworks, templates and code-based assets for your AI usage and data governance to be ethical, secure, compliant and robust. A responsible AI framework is a powerful tool to instill trust in both conventional and generative AI.
   
2. Act to operationalize AI at scale. AI offers its greatest benefits when you use it at scale. For generative AI, that may require changes in your approach to model development and deployment, your technology architecture and skills.
3. Recognize new ROI opportunities. Boost your bottom line with AI strategies that capture indirect costs (such as new burdens on specialists and the new oversight needs of generative AI) and indirect benefits (such as improved employee and customer experiences).