Turn Generative AI from an Existential Threat into a Competitive Advantage: How to factor the new technology into your strategy
Article originally appeared in the Magazine (January–February 2024)
By Scott Cook, Andrei Hagiu, and Julian Wright
What/Focus
Generative AI can be harnessed to more easily and cheaply improve or create products and services that have traditionally required significant input from people in terms of labour and creativity. However, on the flipside, it also has the potential to disrupt and commoditise many businesses. This article describes a 3-level approach to adding value to offerings in the long term by making generative AI part of company strategy. The authors address the following questions: What can firms do to turn generative AI from a threat into an opportunity? How can they use it to build competitive advantage? They also discuss implementation issues and what types of businesses can benefit the most.
How (details/methods)
The first level of implementation involves companies adopting publicly available tools such as off-the-shelf large language models (LLMs) like Chat GPT and Bard. These tools can be used to improve internal communication documents or create social media marketing posts, among many other applications. Publicly available generative AI tools are becoming increasingly necessary and ever more accurate, applicable, and secure. Fine tuning of tools using company data is also possible at this level. However relying solely on tools available to everyone will only create a temporary advantage.
At level 2 businesses can create their own customised generative AI tools that use the data and know-how generated in interactions with customers to improve the customer experience. Tools can be built from open source models or by using commercially available LLM models. Whatever the setting, the tool is trained using internal data to provide ease of use, customisability, and personalisation in an intuitive
user interface.
Incorporating user feedback further improves the model, leading to level 3 – creating an automatic and continuous feedback loop from customers as they use the product or service. The main challenge is collecting reliable signals in the natural process of customers using the product or service without disrupting their experience.
There are implementation considerations at each level. At level 1 the data shared with publicly available tools raises security and competitive concerns – but providers are taking steps to address this. Level 2 requires technical expertise to fine-tune and customise tools to ensure accurate and relevant outputs. At level 3 the focus is redesigning online products and services as well as internal processes to integrate generative AI throughout the entire customer experience, and so maximise data for the feedback loop. Levels 2 and 3 require greater commitment in terms of firm resources, however customisation of proprietary data is becoming progressively safer, easier and more affordable.
Virtually all companies will benefit from level 1, while the decision to move to level 2 and then 3 raises two questions concerning scope. First is how much of a firm’s current offering can be replaced by generative AI, with implications for the magnitude of potential disruption. In the case of online services provided by humans, businesses obviously should move as quickly as possible to levels 2 and 3. The second scope question concerns the potential for a company to leverage their own data, which will depend on how idiosyncratic their data is, the reliability of the feedback they get from customers, and the cost of getting reliable feedback.
So what
Adoption of generative AI technologies is widespread and growing. The firms best placed to achieve lasting competitive advantage will have access to unique customer data via self-reinforcing feedback loops. CEOs and senior leaders must act to make sure this technology is treated as a fundamental part of company strategy, and not just delegated to IT. Investors should focus on the companies that have the potential to reach level 3 implementation.