Lately, it’s become nearly impossible to go a day without encountering headlines about generative AI or ChatGPT. Suddenly, AI has become red hot again, and everyone wants to jump on the bandwagon: Entrepreneurs want to start an AI company, corporate executives want to adopt AI for their business, and investors want to invest in AI.
As an advocate for the power of large language models (LLMs), I believe that gen AI carries immense potential. These models have already demonstrated their practical value in enhancing personal productivity. For instance, I have incorporated code generated by LLMs in my work and even used GPT-4 to proofread this article.
Is generative AI a magic bullet for business?
The pressing question now is: How can businesses, small or large, that aren’t involved in the creation of LLMs, capitalize on the power of gen AI to improve their bottom line?
Unfortunately, there is a chasm between using LLMs for personal productivity gain versus for business profit. Like developing any business software solution, there is much more than meets the eye. Just using the example of creating a chatbot solution with GPT-4, it could easily take months and cost millions of dollars to create just a single chatbot!
This piece will outline the challenges and opportunities to leverage gen AI for business gains, unveiling the lay of the AI land for entrepreneurs, corporate executives and investors looking to unlock the technology’s value for business.
Business expectations of AI
Technology is an integral part of business today. When an enterprise adopts a new technology, it expects it to improve operational efficiency and drive better business outcomes. Businesses expect AI to do the same, regardless of the type.
On the other hand, the success of a business does not solely depend on technology. A well-run business will continue to prosper, and a poorly managed one will still struggle, regardless of the emergence of gen AI or tools like ChatGPT.
Just like implementing any business software solution, a successful business adoption of AI requires two essential ingredients: The technology must perform to deliver concrete business value as expected and the adoption organization must know how to manage AI, just like managing any other business operations for success.
Generative AI hype cycle and disillusionment
Like every new technology, gen AI is bound to go through a Gartner Hype Cycle. With popular applications like ChatGPT triggering the awareness of gen AI for the masses, we have almost reached the peak of inflated expectations. Soon the “trough of disillusionment” will set in as interests wane, experiments fail, and investments get wiped out.
Although the “trough of disillusionment” could be caused by several reasons, such as technology immaturity and ill-fit applications, below are two common gen AI disillusionments that could break the hearts of many entrepreneurs, corporate executives and investors. Without recognizing these disillusionments, one could either underestimate the practical challenges of adopting the technology for business or miss the opportunities to make timely and prudent AI investments.
One common disillusionment: Generative AI levels the playing field
As millions are interacting with gen AI tools to perform a wide range of tasks — from accessing information to writing code — it seems that gen AI levels the playing field for every business: Anyone can use it, and English becomes the new programming language.
While this may be true for certain content creation use cases (marketing copywriting), gen AI, after all, focuses on natural language understanding (NLU) and natural language generation (NLG). Given the nature of the technology, it has difficulty with tasks that require deep domain knowledge. For example, ChatGPT generated a medical article with “significant inaccuracies” and failed a CFA exam.
While domain experts have in-depth knowledge, they may not be AI or IT savvy or understand the inner workings of gen AI. For example, they may not know how to prompt ChatGPT effectively to obtain the desired results, not to mention the use of AI API to program a solution.
The rapid advancement and intense competition in the AI fields are also rendering the foundational LLMs increasingly a commodity. The competitive advantage of any LLM-enabled business solution would have to lie somewhere else, either in possession of certain high-value proprietary data or the mastering of some domain-specific expertise.
Incumbents in businesses are more likely to have already accrued such domain-specific knowledge and expertise. While having such an advantage, they may also have legacy processes in place that hinder the quick adoption of gen AI. The upstarts have the benefits of starting from a clean slate to fully utilizing the power of the technology, but they must get business off the ground quickly to acquire a critical repertoire of domain knowledge. Both face the essentially same fundamental challenge.
The key challenge is to enable business domain experts to train and supervise AI without requiring them to become experts while taking advantage of their domain data or expertise. See my key considerations below to address such a challenge.
Key considerations for the successful adoption of generative AI
While gen AI has advanced language understanding and generation technologies significantly, it cannot do everything. It is important to take advantage of the technology but avoid its shortcomings. I highlight several key technical considerations for entrepreneurs, corporate executives and investors who are considering investing in gen AI.
AI expertise: Gen AI is far from perfect. If you decide to build in-house solutions, make sure you have in-house experts who truly understand the inner workings of AI and can improve upon it whenever needed. If you decide to partner with outside firms to create solutions, make sure the firms have deep expertise that can help you get the best out of gen AI.
Software engineering expertise: Building gen AI solutions is just like building any other software solution. It requires dedicated engineering efforts. If you decide to build in-house solutions, you’d need sophisticated software engineering talents to build, maintain, and update those solutions. If you decide to work with outside firms, make sure that they will do the heavy lifting for you (providing you with a no-code platform for you to easily build, maintain, and update your solution).
Domain expertise: Building gen AI solutions often require the ingestion of domain knowledge and customization of the technology using such domain knowledge. Make sure you have domain expertise who can supply as well as know how to use such knowledge in a solution, no matter whether you build in-house or collaborate with an outside partner. It is critical for you (or your solution provider) to enable domain experts who often are not IT experts to easily ingest, customize and maintain gen AI solutions without coding or additional IT support.