The Question Every CEO is Asking: What is Our Generative AI Strategy?

“What is our generative AI strategy?” That is the question almost every executive team is getting asked today at Fortune 500 companies and beyond. The person asking the question is almost always the CEO, who is expecting an answer and rapid results. It is also true that the same CEO will be unhappy if generative AI (GenAI) starts to hallucinate with your customers, strategic data assets are mismanaged, or intelligent applications create a security or performance issue for your company. While this reality will be a boom for management consultants for years to come, some practical guidance is needed to help those C-level executives (including CIOs and CDOs) and their business unit partners develop a compelling plan. In our conversations with enterprise customers, GenAI companies, and cloud service providers, three core questions have emerged that can inform your generative AI strategy:

1. 1. What is the enterprise mindset for successfully adopting GenAI?
   2. How can you capture “quick wins” in the near term?
   3. What are your best strategies for leveraging GenAI in the medium to long term?

Enterprise Mindset for Success

You have likely already heard the expression that AI had its “Mosaic browser moment” with the launch of ChatGPT late last year. We have been using AI for decades (Google search, Amazon Alexa, Netflix/Spotify recommendations), but now individuals can directly and creatively use GenAI to rapidly deliver value. Whether that value is drafting a note, developing software code, or completing a task, GenAI can help anyone be more efficient and effective just by prompting an AI model via natural language. Opportunities abound in commercial settings, and many of your employees are already experimenting!

Since GenAI is so accessible and dynamic, it requires an agile mindset that is often hard to come by in larger companies. Job functions are going to evolve, new skills are going to be required, and many companies are going to be challenged by gen-native businesses. Make no mistake that GenAI will be disruptive, and many employees will resist the change. Some employees will thoughtfully use areas like security and governance to temper your experimentation. Managing those risk areas is very important and will require guardrails and operational best practices, such as the framework developed by WhyLabs. But, many employees, including CEOs, will be hungry to try, learn, and apply GenAI to your business. Finding ways to leverage these new technologies with lower risk and rapid feedback loops is critical to every business.

Before jumping in, think deeply about your company’s domain capabilities and differentiated data assets and how you plan to leverage them to your company’s and your customers’ long-term benefit. Established customer relationships, differentiated data and know-how, and existing workflows and interfaces can all be valuable tools to an incumbent business. Your employees and customers will already be experimenting with GenAI extensions of your products and value proposition. Embrace these innovators and seek to understand what problems they are addressing and why. By setting a tone where smart experimentation is valued and encouraged, you can rapidly learn what is possible and where to prioritize as an organization while managing risk!

GenAI is very approachable from an end-user perspective, but building and operating GenAI models and applications is not.

3-step Approach to Quick Wins

By asking your employees, customers, and trusted technology partners what they are already doing with GenAI, you will discover relevant opportunities for your business. In the near term, we recommend a 3-step approach to getting started and rapidly learning as you develop your company’s GenAI strategy:

1. 1. Embrace “Copilot” productivity gains.
   2. Partner with large cloud service providers (CSPs) and software companies already delivering GenAI offerings in a “gen-enhanced” manner.
   3. Launch a GenAI Learning Lab and hold a hackathon for rapid and agile experimentation that will help identify and prioritize your GenAI capabilities. This Learning Lab can also help identify emerging, “gen-native” companies with innovative products and approaches to complement your internal skills.

Copilots

One of the early success stories of GenAI is the emergence of copilots — like GitHub Copilot from Microsoft or Code Whisperer from Amazon — for improving software development. Similarly, ChatGPT from OpenAI, and others, have enabled an intelligent assistant in many forms that can help employees be more productive in creating written first drafts.

Your developers are already leveraging GenAI tools for writing and refining code, knowledge workers are drafting marketing documents, and project managers are automating specific workflows. These tools are increasing employee productivity today. In particular, your engineering teams can improve software development by driving the adoption of code generation tools. Many software teams are already seeing productivity boosts of 20-30%. It’s important enough that these teams should take time away from current software development approaches to learn how to leverage these new methods. Teams that do NOT use code generation tools for increased productivity will increasingly fall behind in delivering software projects.

Understanding what is already happening within your organization and encouraging productivity gains will help you assess your internal GenAI readiness and categorize which opportunities are most likely to be “Do It Yourself,” “Do It With Me,” or “Do It For Me” initiatives. Said another way, is your company prepared to bake your own GenAI cake from individual ingredients (data, algorithms, GPUs), make a cake with store-bought cake mix and icing (CSPs, managed models), or buy the cake from the local bakery (gen-native and gen-enhanced SaaS offerings)? GenAI is very approachable from an end-user perspective, but building and operating GenAI models and applications is not. Observing what your employees are already doing will create a baseline of internal capabilities and a plan for how you will deliver your GenAI solutions. To succeed, you will need to have an objective assessment of your in-house capabilities for gathering and preparing data, training and deploying underlying GenAI models, and building GenAI enhancements to your existing applications and workflows. Most likely, the early benefits will be a combination of internal productivity gains and capability learnings from buying gen-enhanced offerings.

Directly Leverage Partners

Look to your incumbent application partners already enhancing their data infrastructure and applications with GenAI. You have a business relationship with them, and they are generally secure and compliant providers. You also have shared context and data sets, and your employees understand the UI/UX and workflow offered by these providers. Thus, they can rapidly help you enhance your business processes and workflows. These partners likely have better access to emerging technologies and skill sets to rapidly enhance their offerings with the deeper elements of the GenAI stack.

Near term, your established partners, including cloud service providers (CSPs) like Amazon AWS, Microsoft Azure, and Google Cloud, and software application companies like Workday, Atlassian, and Smartsheet, are already enhancing their offerings for you. There is also a data management layer that the CSPs and data services companies like Snowflake, MongoDB, and Databricks provide for deeper GenAI work you will want to take on directly.

The key to taking an initial buy-over-build approach is to achieve additional quick wins by understanding and leveraging what your established partners are already doing. Then, as you further clarify where your company is going, you can find the best fit for building your own GenAI enhancements.

Building a GenAI Learning Lab

Most of your early wins, whether from internal early adopters or enhanced third-party offerings, will be to increase internal productivity and creativity. But, the more existential need is determining how you will transform your business to better serve your external customers. Many technology companies have already hosted hackathons to energize their internal teams and develop prioritized opportunities for serving their customers. That would also be an excellent place for your company to start! In addition, they have designated a senior executive to be the GenAI cross-functional leader responsible for leading experimentation efforts and developing a GenAI strategic plan. You could think of this collection of efforts as a GenAI Learning Lab.

The more existential need is determining how you will transform your business to better serve your external customers.

While this team will likely monitor and summarize the rapid and dynamic changes in GenAI, the most effective discovery approach is to learn by doing. Getting selected for the GenAI Learning Lab should be positioned as an opportunity for doers — your highest potential builders and leaders in the company. They will need access to data resources and may need to step on some toes in IT and data management to get what they need to experiment rapidly. Senior executive sponsorship will be key to balancing experimentation and risk. The lab (and the whole company) should have some ground rules and guardrails about when and how to share their projects in a secure and compliant manner. While the hackathon that helped launch this group can be a source of inspiration, it will likely take time and some outside expertise for these teams to develop priorities and product offerings your customers can test. Having a bias for action and an experimental mindset will serve the Learning Lab and your company well.

A few specific areas to consider for early priorities are generative chat and generative search for your customer service reps and customers. Generative chat is a natural language way for your customers and service representatives to interact with the data and knowledge resources you already have that can help someone get to an answer faster. The idea is to interact through a chat-like interface and get questions rapidly answered. Generative search is a cousin to generative chat, although it is more about adding reasoning and insights to your search results. Google recently announced a set of capabilities around enterprise search.

The Learning Lab is likely the best place to partner with compelling startups building more gen-native services in sectors like writing, images, video, code creation, and more. Companies like Jasper.ai, Copy.ai, HyperWrite, and of course, OpenAI’s ChatGPT are natively supporting a wide variety of writing use cases. Companies like Lexion in legal contracts or Harvey in legal procedures focus on specific areas of drafting and managing documents. RunwayML is an early leader in the video creation and editing market and is fortunate to have a team that deeply understands the whole GenAI stack. There are also GenAI-focused companies emerging in enabling areas like model testing and deployment (OctoML, Mosaic), data wrangling (Number Station), and data ingestion (Unstructured.io). We believe gen-native companies are taking a fresh look at the customer’s problem, the new technology capabilities required, and the agile methods needed to constantly iterate from the data up to the end user. As Tomasz Tunguz thoughtfully shared a few weeks back, building, deploying, and operating generative apps will create significant product development and organizational and cultural changes. Working with select gen-native companies in your Learning Lab and beyond will inform your longer-term approach to these areas of capability building.

Winning in the Medium to Long-term

Having set your course for some quick wins and capability building, you will likely have earned the appreciation and confidence of others inside your organization (including the CEO) for the more medium to long-term projects. You will also have better calibrated your internal team and process capabilities, data assets, and customer needs. While building on your quick wins, take time to step back and develop a set of longer-term priorities for what you are going to deliver and how you are going to do it.

Many copilot and software partner quick wins will unlock insights into new internal and external opportunities. Where can you deliver more automation, personalization, and, ultimately, cost savings in areas like customer fulfillment and customer success? Do you have proprietary data and/or an understanding of public data sources that can help you build differentiated GenAI models and offerings? Is there a better system, enhanced by GenAI, for designing the next iterations of your products and services? Will iw distribution or partnering channels emerge to get your products to market while continuing to feed your need for data and reinforcement learning?

One of your biggest decisions will be, “How do we do it?” Your quick wins will likely identify internal champions with the skills and the passion to embrace your GenAI transformation. Cultivating these GenAI champions and involving them in longer-term initiatives will substantially improve your prospects for success. In particular, look for champions who are curious to learn but don’t instinctively believe your company must take a DIY-only approach. Most companies will benefit from a mix of DIY and “do it with me” partnering approaches as they seek to leverage the whole GenAI stack to better meet their customer-facing and employee-engaging needs.

In general, here are a few more guidelines for increased probability of success:

1. 1. Initially focus on cloud-based deployments over edge use cases (edge is coming but is still early).
   2. Design from the data up — and differentiate at the application layer over the infrastructure layer. You will likely be leveraging other open-source and managed models (both LLMs and domain-specific models) and cloud infrastructure, so it is your data and workflows that will help you customize models and applications that create distinct value.
   3. Always be pushing to understand your data and metadata assets and strategically manage them for their long-term value (balancing the data leakage risk with the data spoilage risk). You will likely need to upgrade your underlying data stores, such as data warehouses and database types used, to eventually build your own language models that leverage proprietary data and insights.
   4. Embrace a mix of gen-native innovator and gen-enhanced incumbent technology partners to support your GenAI journey. This should help you increase internal productivity and creativity while improving customer experiences and value!

Prepare for a Marathon, Not a Sprint

Almost 30 years ago, I was a young corporate VP at a Fortune 500 company named The Genuine Parts Company (GPC). Our businesses offered exceptional supply chain services from the manufacturer through to the end user customer in different industries, including auto parts (NAPA), industrial products, and office products. In 1995, a disruptive technology called the internet emerged — launching one of the first big VC hype cycles and a flurry of commercial innovation impacting every company. Each week, a new VC-backed company reached out to GPC, exploring how they could be a new technology enabler, a new customer type, or a new partner of ours (which might turn into a competitor!). My role rapidly evolved into screening these companies, assessing their fit with our strategies, and structuring arrangements that would help extend our value while embracing technological change.

As the internet bubble of the late 1990s approached its peak, the whole business world was consumed with e-commerce, dot-coms, and “bricks and clicks” innovation. Traditional companies like GPC saw their stock prices plunge amidst concerns they would be trapped in an “innovators dilemma.” NAPA Auto Parts, Amazon, and Madrona explored a “bricks and clicks” joint venture but realized shortly before the NASDAQ crested at a then-record 5,000 (in March 2000) that the business market wasn’t ready for an online auto parts store. Others were not so lucky, and countless founders, startups, and investors experienced substantial losses as the bubble burst. I will never forget attending a conference hosted by Fortune and Goldman Sachs in New York City in April 2000 that was framed as the battle between the Fortune 500 companies and the dot-coms. Few there believed that the Fortune 500 had a chance, but today GPC, by having a learning mindset, excellent execution, and a long-term perspective, has a market value 8X what it had back then!

In 2023 and beyond, disruptive GenAI technologies, intelligent and generative applications, and more natural user interfaces will produce winners and losers across startups, incumbent technology companies, and established enterprises. Enterprises and their leaders will need a combination of agility, creativity, and long-term perspective to navigate this highly dynamic era. They will also require strategic clarity about their customers’ needs, an objective assessment of internal capabilities and data assets, and engagement with external partners to emerge as winners. That is the deeper reason why every CEO is asking the question: “What is our generative AI strategy?”