Now: Generative AI Interest and Adoption Soar, Promising Disruption
Generative AI captured the public’s imagination when it burst onto the scene in the second half of 2022 and first few months of 2023. Few technologies have ever debuted to such fanfare. Adoption and use of generative AI have been sudden and rapid among the public. OpenAI reported reaching 100 million users within 60 days of releasing ChatGPT to the public; in comparison, it took TikTok nine months to reach that milestone. Midjourney’s image generator has around 16 million users. There are 1.5 million daily users of Dall-E 2. Google’s Bard chatbot had 10 million page views in July. Growth in the use of generative AI in enterprise settings has been no less impressive, according to Deloitte’s 2023 CEO Priorities Survey.
What’s made generative AI so impactful is a convergence of factors. First, advanced hardware—primarily specialized AI chips used in training models—have helped produce more advanced models such as large language models (LLMs). These tools have gone mainstream due to a seamless user experience, enabling even nontechnologists to engage with very advanced models.
All this attention has kicked off a gold rush among investors. Investors are pouring money into startups that have generative AI technology at their core, betting that we’re witnessing the dawn of a new paradigm for business technology, one where insights are surfaced automatically, contracts review themselves, and a never-ending stream of content is generated to keep brands in front of their audiences.
While there’s been plenty of talk about how AI may threaten jobs, there’s no real indication that business leaders are planning on using it to automate knowledge jobs at any kind of scale. In a survey of leaders, improving content quality, driving competitive advantage, and scaling employee expertise were the most common reasons for deploying generative AI. Reducing headcount was one of the lowest priorities. It looks more likely that AI will liberate workers from rote, repetitive tasks and free them to focus on more creative aspects of their jobs.
The picture that’s emerging is that AI is coming, and for some, it’s already here. But, as the saying goes, leading businesses know they can’t shrink their way to growth—that is, minimize risks or costs as a path to growth. This means the most productive uses of generative AI won’t be about replacing people but instead will focus on arming employees with tools that help them advance and enhance their productivity, knowledge, and creativity—which, in turn, will help drive innovation in the enterprise.
Executives are increasingly under pressure to speed this transition and stay ahead of their competitors. According to one survey, 64% of CEOs say they’re facing significant pressure from investors, creditors, and lenders to speed the adoption of generative AI. But just as leaders know they can’t shrink their way to growth, they also know the importance of leading with need. Shoehorning generative AI into any and all processes just because it’s a shiny new thing is unlikely to deliver meaningful gains. Instead, businesses may benefit from a more strategic approach to implementation that focuses on leveraging generative AI’s unique capabilities to solve existing problems and help businesses differentiate themselves from competitors. That’s the approach innovative enterprises are taking today.
New: Enterprises Aim for Scalability and Domain Expertise
The true value of generative AI is likely to be unlocked when organizations can use it to transform business functions; reduce costs; disrupt product, service, and innovation cycles; and create previously unachievable process efficiencies. To get there, business leaders may consider more of an evolutionary approach to their enterprise data and technology strategy.
Becoming an AI-fueled organization takes careful discipline and a focus on maintaining systems and algorithms. Just as a rocket needs a launch pad and flight controls to reach its destination, generative AI tools need infrastructure and control systems to succeed in enterprise settings. The good news is a lot of the muscle memory businesses have been developing over the past several years while building up data analytics and machine learning capabilities also applies to generative AI, though some practices may require subtle retooling.
Generative AI typically requires terabytes of data on graphics processing unit–enabled high-performance computing clusters. Since few businesses have this infrastructure, most will access it as a service. Via application programming interfaces, engineers can weave generative AI capabilities into their existing software without needing to build out new infrastructure. While AI vendors are prioritizing ease of use in their products, it’s still important for enterprises to keep these engineering requirements in mind.
Additionally, it’s important to pick your use cases wisely. AI can be used to reduce costs, speed up processes, reduce complexity, transform customer engagement, fuel innovation, and build trust. The specific application of generative AI will vary from business to business, but looking for projects that drive improvements in one area is a good place to start.
Here are some additional considerations from businesses that have already adopted the technology.
Data is the Fuel that Drives the Generative AI Engine
Businesses need to ensure their data is architected properly and accessible to AI applications to enable model training as well as next-generation use cases.
This was one of the learnings for Enbridge, the largest natural gas utility in North America. Several years ago, when it began an ambitious cloud migration journey, it didn’t set out to pioneer new generative AI uses. The primary goals were to modernize its infrastructure and eliminate technical debt by reducing the size of its on-premises data centers. Along the way, it built a centralized data repository that collects data from across the enterprise, including regulatory, marketplace, HR, and other data. This centralized data marketplace replaced what used to be hundreds of silos.
Once generative AI arrived on the scene, Enbridge’s leadership knew this centralized data marketplace was the perfect engine to drive new AI-fueled efficiencies. The technology team rolled out a generative-AI-based copilot tool that helps developers quickly and more efficiently build out code. It also supplied the company’s office staff with a copilot tool to help them navigate productivity applications.
The goal, says Joseph Gollapalli, director of cloud, IT ops, and data at Enbridge, is to “accelerate our delivery and drive innovation and efficiency. These AI solutions have the potential to enhance our operations, improve safety, elevate the customer experience, and enhance our environmental performance.”
Governance is More Important than Ever
Without effective governance guardrails, AI can’t scale. A governance framework should define the business’s vision, identify potential risks and gaps in capabilities, and validate performance. These types of considerations not only safeguard the business but can also help scale projects beyond the proof-of-concept stage.
At CarMax, the largest used car retailer in the United States, effective use of generative AI is predicated on a systematic, enterprise-wide approach that embraces the power of the technology while also putting in place guardrails to ensure employees are using it effectively. One of CarMax’s most prominent applications is a tool that adds AI-generated content to research pages for vehicles. These pages summarize information from thousands of actual customer reviews to let shoppers quickly see what other buyers had to say.
Shamim Mohammad, executive vice president and chief information and technology officer at CarMax, says these kinds of use cases deliver the most business value when they are done in a controlled manner. CarMax has prioritized governance, which may not feel like the most exciting aspect of generative AI but is key to scaling it. The company has created an AI governance team dedicated to ensuring teams across the organization are using AI appropriately. The key is that this team is not charged with simply saying no to new use cases. Part of its mission to help scale impactful applications across the enterprise by standardizing how models are trained and used. The goal is that generative AI is used beyond just technology or product teams.
“We’ve done a lot of cool things through machine learning and AI,” says Mohammad. “What I’m focused on now is ensuring we’re using it in a responsible manner and making sure that, as a company, whatever we deploy, it’s being done in ways that are consistent with our core values.”
Make Sure you have the (Copy)Right
Generative AI has altered the copyright landscape. Now anyone can create images, video, text, and audio with a few clicks. However, some models have been trained on content that comes from third parties. One US court recently ruled that this makes AI-generated content ineligible for copyright protection. Additionally, training models on copyrighted material scraped from the web may present legal risks, including intellectual property infringement.
However, these don’t have to be problems. The content provider Shutterstock, for one, has shown that it is possible to use generative AI in ways that both respect the rights of the original copyright holder and ensure that AI-generated content can be used for commercial purposes.
Shutterstock recently unveiled an image-generating tool that creates visuals based on users’ prompts. Like other image generators, the tool was trained on images created by third-party artists. However, unlike other image generators, every artist whose work was used in training the model agreed ahead of time to participate. Participating artists are also paid when their work is used to train a model and when a user licenses an image generated on the platform. Shutterstock licenses its content as data, which allows it to offer added legal protections to end users.
“Everyone is creating content, from CEOs to folks who work in retail,” says Michael Francello, director of innovation at Shutterstock. “The need to create content was absolutely exploding. We saw an early opportunity to look at our content as data that could train generative AI models. It’s about protecting the core of our business, but also respecting the core, which is the artists and the contributors.”
Crawl, Walk, Run, Fly
This approach has for years been an effective way for enterprises to scale up their use of service offerings. Generative AI is no different. In the crawl stage, applications may be ad hoc and require lots of manual effort. These eventually graduate to the walk stage, in which processes become more defined at the foundational level and automated. In the run stage, use cases get standardized and become pervasive at the enterprise level. When it’s time to fly, the organization leverages the work it has already done to embrace next-generation capabilities.
That approach helped chemical company Eastman begin developing generative AI-based internal services. The company has a long track record of using data and analytics in an industry that isn’t typically known for it. For example, it has an advanced intelligence service (with proprietary thermal stability measures) that will predict when a heat transfer fluid used in its customers’ industrial processes is likely to degrade, allowing engineers to maintain optimal fluid quality, forecast predictive maintenance needs, and avoid costly downtime on manufacturing lines.
Building on this experience, the company is now experimenting with how generative AI can enhance its sales processes. It built an AI-enabled tool that can read through natural language text files. Still in the development stages, the tool is being tested on extracting insights from notes from sales calls. These documents are generated by sales teams after every call but rarely get read by anyone, even though they hold significant intelligence. Now, with the help of generative AI, the company is starting to unlock those insights.
“It lets us, a chemical company, bring a digital service layer to the table to differentiate ourselves in the market and create a competitive advantage,” says Aldo Noseda, chief information officer at Eastman.
Given the pace with which generative AI is progressing, it may be wise to apply this kind of framework to new enterprise use cases. Let proof-of-concept projects lead to standardized practices that become standard operating procedures across the enterprise. Once a business has achieved this kind of maturity, the sky is the limit.
In the near future, it may become even easier for businesses to reap the benefits of generative AI within their industries thanks to the emergence of models that are trained on more specific data. Today, most enterprises that are using generative AI are using tools built on foundational models that were trained on general-purpose data. That tools with such a general knowledge base can be used in very specific subject-matter areas shows the power of LLMs. But the next generation of LLMs is likely to be more hyper-focused and tailored to businesses’ specific needs.
This is a trend that’s already begun to emerge. NVIDIA has introduced a tool called BioNeMo, an LLM aimed at the biotech sector. Google’s Contact Center AI is a tool trained to handle customer service interactions. BloombergGPT is designed to answer finance industry–related questions. ClimateBERT is a model trained on climate change research and can advise businesses on their climate-related risks.
As businesses realize the benefits of models trained specifically for their sector, we’re likely to see more demand for these types of services. More than one-third of enterprises are already planning to train and customize LLMs for their business needs in the future. Private LLMs are likely where the true potential of generative AI lies for businesses. They are developed and maintained by organizations that keep underlying code proprietary and closed to the public. These LLMs are purpose-specific, hosted securely, and trained on private data, and they can offer tremendous competitive advantage to organizations. This is likely the next wave in the generative AI journey.
Next: Imaginative Executives Wanted
The motivational poster has gone from mere corporate cliché to its own category of meme, but one overused aphorism may reclaim its stature as an enterprise imperative: We’re only limited by our imagination.
While you might have heard the saying before, teams and organizations have always been bound by limiting factors. They don’t have enough data or the right data. Leadership is skeptical. Or, most dreaded of all: “That won’t move the needle.”
But in a generative AI world, imagination truly is the only limit. It’s now possible to create constant streams of content, identify new operational efficiencies, or scan regulatory filings or customer reviews in minutes. Now the only question is, what do you want to know?
Asking better questions will become a crucial skillset in enterprises that have adopted generative AI. This trend may create demand for a new type of leader, one that is driven more by creativity than we’ve seen in the past. The past 20 years or so have seen leaders rewarded for steering their organizations based on data and insights, rather than gut and instinct. But the next few years could see more imaginative leaders leap ahead. Give an image generator a boring prompt, and it will produce a boring picture. The same is true of generative AI applications at the enterprise level. Unimaginative use cases produce limited impact. As more businesses attempt to differentiate themselves from their competition, leaders who can find creative new applications for generative AI may separate themselves from their peers who are busy just following data.
This isn’t to say that data-driven decision-making will become passé. In fact, it will be as important as ever, if not more so. But the definition of what it means to be data-driven may change because the range of data that leaders can access will increase, thanks to generative AI. So much of an enterprise’s data is buried in natural language text files, machine logs, and, increasingly, intelligent products. Generative AI gives organizations the ability to make sense of this digital exhaust. The creative leader will understand what these oft-overlooked data sources have to say about their business and will use generative AI to ask intelligent questions of the data sources. And they will ask these questions at the speed of thought, rather than waiting for their weekly report.
But all that barely scratches the surface of generative AI’s full range of likely impacts. We’re pretty sure it’s going to be seismic. We just don’t know exactly where the ground will shift the most.