The release of ChatGPT and other generative AI systems is a transformational moment in the democratization of AI, enabling more people to use the technology and lowering organizations’ barriers to entry. It’s easy to get swept up in the excitement: ChatGPT set the internet abuzz with its astounding capabilities and amusing failures. Observers quickly predicted the end of high-school essays, lamented the death of creative writing, and raised alarms about the future of work.
Such reactions are to be expected. We’re witnessing the classic Gartner Hype Cycle, and summiting the peak of expectations with the current wave of Generative AI systems. But despite how fast we’re moving, business leaders can’t ignore the fact that, when the hype is stripped away, these systems are just technology. And, like any technology, what ultimately matters most is not just whether businesses use it, but how its use will affect business models and how it can change businesses’ relationships with customers. Organizations that consider the pace of change but stay focused on the fundamentals will reap the biggest rewards from adopting Generative AI. Here are eight pillars that will strengthen their investments and positioning:
1. Understand the capabilities and limitations in the context of your business
Novel generative AI systems have a plethora of use cases, emerging from the democratization of AI, potential for novel capabilities, and increased ease of use. This presents executives with an urgent challenge: deciding how to make meaningful investments to bring these systems into their organization. Executives need an accurate understanding of the feasibility of various use cases matched with the current and near-term capabilities of Generative AI systems. Business leaders who are stringent when selecting pilot areas and targeted rollouts will achieve long-term success—instead of making hasty deployments that can lead to wasted effort and the erosion of customer trust.
2. Adapt the business model to new opportunities
Generative AI systems are powerful, but in the end, they remain a piece of technology used to meet a specific goal: delivering value to the customer, and ultimately creating the basis for a value exchange. Business leaders will continue to need to think critically about the economics and value exchange of the product or service for which they introduce Generative AI, and not just assume that it will automatically create demand and grounds for a financial transaction (it might do so for a free service in the short run, but in the long run, customers will want concrete value).
3. Establish tight customer feedback loops to steer the development roadmap
While Generative AI systems based on foundation models might require less data to be fine-tuned for specific purposes, they also present a new modality of interaction for customers. As such, product and service development processes need to pay more attention to how customers engage with these systems. Creating tighter feedback loops that capture the unarticulated needs of customers will allow the organization to achieve product–market fit more quickly, and ultimately deliver payoffs from that investment. These feedback loops need to focus on how customers respond to novel capabilities (such as intuitiveness of the interface and delivery of value) and utilize qualitative feedback channels to unearth friction points in adoption and usage.
4. Address Responsible AI issues such as bias and privacy
In many cases, Generative AI system capabilities will be consumed via APIs (such as in ChatGPT and DALL-E 2), meaning there is little control over and transparency into the operations and development of the system. This exposes the organization to the risk of exacerbating biases and privacy intrusions, undoing investments into Responsible AI. Furthermore, Generative AI systems in the future might be trained on data generated from the current crop of systems, amplifying existing ethical issues. Ethics and safety should be a top priority in the design, development, and deployment of any AI system. But in the case of Generative AI systems this becomes even more urgent, since they can demonstrate run-away effects that might emerge suddenly and exponentially but be too subtle to catch. Investments into monitoring capabilities that can track the evolution of bias will help organizations be more proactive in addressing issues as they arise, rather than waiting for incidents to spark action.
5. Achieve and maintain regulatory compliance
Given the expanding set of capabilities that Generative AI systems offer, with limited transparency and explainability, compliance with some of the emerging regulatory requirements might be even more challenging and complex. Scrutiny from regulators will intensify as these use cases become widespread and impact people in meaningful ways. Proactive executives will urge their organizations to accelerate their understanding of these challenges, to identify gaps in compliance that might emerge from the use of Generative AI systems. They will also prompt the risk and compliance, privacy, legal, and ethics functions in their organizations to set up foundations through technical infrastructure, processes, policies, and governance to preempt failures. For example, utilizing techniques such as tracing and auditing in the AI lifecycle can help catch emergent failures early—they alert relevant staff to act on the issues before they spread, impact customers, and run afoul of regulatory requirements.
6. Implement robust cybersecurity protections
Whereas malicious actors need to be right only once, the “good guys” must be right all the time to protect their systems and stakeholders. The same goes for Generative AI systems, which introduce novel attack surfaces that companies must be aware of, such as software supply chain vulnerabilities and adversarial attacks. There are many creative ways to break these systems. For example, simulating a VM in ChatGPT shows us how we can trigger arbitrary code. Such an attempt points to more challenges ahead for those trying to manage risk and will require defense teams to invest in red-teaming the system to find points of vulnerabilities. Tying up loose ends and instituting tighter monitoring, controls, and safeguards—especially those tailored toward machine learning–specific vulnerabilities—will be even more essential.
7. Create mature infrastructure to ensure the supply of high-quality, proprietary data
For those looking to achieve a competitive advantage in the market, simply using an off-the-shelf Generative AI system that is also available to your competitors isn’t going to deliver stellar business results. You’ll need to fine-tune and adapt the model to your specific context and use case. Having access to high-quality proprietary data from your own customer interactions (with informed customer consent) will give your organization’s investment into Generative AI a boost by making it more performant, reliant, and better suited to meet the needs of your customers. This can be achieved through investments into the core features of a mature data infrastructure, such as data pipeline robustness, reproducibility, and stability, feature stores for easier application of AI-powered insights, and enhanced model management in the AI lifecycle. Businesses will also need talent with engineering stills to integrate Generative AI systems—to create meaningful, well-designed experiences.
8. Acknowledge that humans aren’t going anywhere
To make the most of Generative AI systems, company leaders need to keep well-informed humans in the loop throughout, from decision-making about how these capabilities are brought in to how they are governed and where these systems are piloted and deployed. Companies will need to embark on a change management journey with their employees, so that they feel empowered, not fearful. Finally, while Generative AI is a massive accelerator, leaders need to train employees to exercise deliberate reliance (for example, reflecting on where and how they are interacting with and using new capabilities) to ensure that Generative AI doesn’t introduce unnecessary dependence and vulnerability in critical areas.
Much will change with the introduction of Generative AI systems, but there will also be many things that won’t change. The pillars above should feel familiar to business leaders who have already been grappling with the integration of other emerging technologies. Prioritizing these core issues will be a differentiating advantage for organizations that want to lead their industries in 2023.