Why Your Organization Needs an AI Factory: A CIO's Guide to Leveraging Generative AI

If you plan to integrate generative AI (GenAI) into your operations, you may need to update your AI strategy. Your current IT operating model might not be optimized to leverage the full potential of GenAI’s speed and scalability. Instead, adopting an "AI factory" approach can help you rapidly deploy a single GenAI model across various functions and tasks. The result? Immediate gains in productivity and efficiency, and the potential to unlock new business models in the near future.

At PwC, we have extensive experience in deploying GenAI at scale while managing the associated risks. Through a three-year, $1 billion investment, we are embedding GenAI throughout our firm and helping our clients do the same. GenAI offers extensive benefits across multiple business areas with high levels of repeatability. Instead of being confined to discrete projects, its true potential is realized as an enterprise-wide capability. We bring this vision to life through our GenAI Factory. Here’s what you should consider when setting up your own AI factory.

1. Adopt GenAI Boldly — Start Quickly and at Scale

To achieve rapid ROI from GenAI, leverage its ability to apply a single, repeatable "pattern" of AI training and deployment across your value chain. Unlike traditional AI, which often requires a custom-built model for each use case, GenAI can use one model to generate or troubleshoot software code across multiple functions or business units.

Because GenAI models come pre-trained, they usually only require adaptation and customization, allowing you to swiftly move from a proof of concept to a pilot. Although you won't deploy GenAI everywhere at once, you can start with a few key use cases, learning from both successes and setbacks. With the right preparation, you can implement GenAI far faster than conventional AI. A 90-day sprint is often enough to get initial use cases up and running, ready to scale.

To map out your GenAI deployment, bring together both technology and business leaders. Identify use cases, assess their value, and search for patterns that can be replicated to generate value elsewhere. Frequently, a moderately valuable use case with repeatability delivers more ROI than a high-value but one-off solution.

As you build your roadmap, identify gaps in data, technology, and skills, and estimate the cost and timeline to address them. Typical gaps include insufficient or poorly organized data sets, the need for more cloud computing power, cloud engineering expertise, new APIs, and a lack of both specialized data science skills and comprehensive knowledge of responsible AI practices. It's also essential to integrate trust-by-design into your strategy—embedding responsible AI principles from day one to validate results, monitor ROI, and mitigate risks over the long term.

2. Choose the Right Model and Prepare Your Data

A variety of specialized companies, including cloud giants like AWS, Google, and Microsoft, as well as OpenAI, offer generative AI "foundation models"—deep learning algorithms pre-trained on vast amounts of public data. After selecting the right model, work with your cloud provider to establish a private version within your firewall.

Customizing a foundation model with your own data, context, and intellectual property can significantly enhance GenAI’s value while ensuring security. By embedding your most experienced specialists’ knowledge into the model, you can effectively distribute their expertise across the organization—empowering everyone as if they had access to a personalized mentor.

As you upgrade internal technology and cloud infrastructure for GenAI, pay special attention to your data. While GenAI excels in processing unstructured data, you may still need to cleanse and organize certain datasets to reduce the risk of bias. You'll also likely need to build data pipelines, complete with new APIs, to ensure the model has continuous access to up-to-date information. To protect your data and intellectual property, upgrades to data governance and cybersecurity may be necessary to counter GenAI’s specific risks.

3. Establish a GenAI Factory for Fast, Repeatable Outcomes

For efficient and repeatable deployment, construct an AI factory—an operating model based on pods that identify and assess use cases in a targeted domain or line of business, then tailor the foundation model to deliver value. Each pod includes six key roles:

• Pod Leader: Oversees business objectives and monitors the value of each use case.

• Business Analyst: Breaks down use case objectives and success criteria, gathers requirements, and tracks progress. They should be familiar with the processes that GenAI will enhance or automate.

• Prompt Engineer: Designs prompts to refine the model’s outputs for accuracy and relevance, collaborating with the business analyst and model mechanic to adjust as necessary.

• Model Mechanic: Customizes the AI model to optimize results and handles the technical aspects of the solution after deployment.

• Data Engineer: Prepares and organizes the data to meet the model’s requirements.

• Data Scientist: Works with the prompt engineer and model mechanic to maximize model performance and helps streamline the code for future use.

To staff a GenAI factory, you may need to hire specialists and upskill existing business and technology professionals. However, the scalability of GenAI can ease the burden on your workforce. For instance, technical specialists like data scientists and model mechanics can often support multiple pods simultaneously. As we do at PwC, you can also leverage pre-built toolkits with ready-to-use software code and prompts. These toolkits help accelerate deployment and reduce costs for applications such as personalized customer experiences, content creation, research, agile software delivery, support services, report generation, and more.