Where is "AI" Waldo? (Post #2)
What are the common ways professionals use generative AI, and how should you decide where generative AI can help you?
Initially, Massive Adoption but Limited Mainstream Value
In January 2023, Open AI reached 100 million active users just two months after launch. The tonnage of news and press releases from big-name tech companies leaves little doubt about the potential of generative AI.
With such a strong adoption curve, one might expect stories of Gen AI producing value to follow soon. However, that didn’t happen. McKinsey fielded its Global Survey of AI in April and reported that the percentage of businesses and business functions using AI in 2023 was relatively consistent with levels in 2021 (survey from April 11-21, 2023; n=1684).
Generative AI, by definition, creates content. This is simultaneously one of its most significant advantages and biggest risks. Generative AI can be WRONG. Additionally, 71% of senior IT professionals believed generative AI introduces new security risks, according to a March 2023 Salesforce survey. Thus, it isn’t surprising that many businesses are moving more slowly to mitigate risk.
Creating Content Emerges as a Valuable Initial Use Case
Many of the first companies to use or productize generative AI did so by enhancing user-supplied content:
Writing a Book: Reid Hoffman co-wrote “Impromptu: Amplifying Out Humanity Through AI” with ChatGPT-4. Hoffman released the book in March 2023. Within Impromptu, Hoffman prompts ChatGPT with questions and his thoughts, providing readers with two perspectives on various possible uses for generative AI across education, law, journalism, and more.
Copy-Editing: Copy.AI helps business professionals copywrite and improve their content. Copy.AI was an early adopter of ChatGPT-3 and started experimenting with generative AI in 2021. As of October 2023, Copy.AI sites Salesforce, Survey Monkey, and Zoom among its customers.
Image Generation and Editing: Adobe announced a set of generative AI models for image creation and text effects called Adobe Firefly. On March 21, 2023, Adobe launched Adobe Firefly beta, built on their existing Adobe Sensei AI infrastructure. By October 10, 2023, users had generated over 3 billion images through Adobe Firefly. On the same day, Adobe announced a new Firefly Vector model capable of “human quality” vector and pattern outputs.
Idea Creation: Grammarly launched GrammarlyGo in Beta on March 9th, 2023, using ChatGPT 3 to help users generate new content and offer new ideas based on the context of the already created text.
Actionable Insights: A Roadmap to Follow
Several key commonalities across these examples serve as a roadmap for those looking to follow a proven path to value.
Allow for time to experiment and build: All companies noted above used ChatGPT before November 2022 or had their own AI tool kit in-house.
Expect to be an editor: Each company communicates that someone should decide whether or not to accept the AI’s output.
Use a trusted AI tool: OpenAI was a commonly chosen AI provider among early adopters, with Adobe as a notable exception, given its existing AI capabilities.
Launch in beta to manage customer expectations: Most providers labeled their initial generative AI product a beta when releasing it to a large audience, thus securing optionality for changes and customer patience with issues.
Identify your risks and have a plan to mitigate them: Gen AI has numerous risks, from IP to security to inaccurate responses. Each provider tackled these items head-on to build customer confidence.
Next Step: Identify Your Use Case and Preferred Approach
AI is capable of driving growth or creating efficiencies. As with most types of innovation, businesses are proceeding down one of two different paths: trying to be first-to-market or a fast-follower.
A First-to-Market Approach
Novel ideas and applications of generative AI come from some combination of generative AI’s unique capabilities used to solve problems in a new way. The list of potential solutions and use cases for generative AI is as large as your imagination. The below provides an overview of this approach.
Identify a problem you want to solve. If your goal is growth, start with a customer problem. If your goal is efficiency, start with an internal issue.
Consider generic AI capabilities. Evaluate how each of the below might be used to help solve the proposed problem:
Enhance existing text, images, computer code, or video content
Build new, creative content (text, images, computer code, or video) in response to prompts such as topic, tone, style, and scenario
Guide decisions based on reviewing large quantities of content, identifying possible errors or inconsistencies, and suggesting possible improvements
Allow a broad audience to engage based on a conversational and accessible user interface
Answer questions providing technical and non-technical responses in near real-time, using an immense set of data
Identify new connections in unrelated, potentially unstructured data and suggest insights
Summarize key ideas across text, audio, video, or other content
Create a proposed solution. Combine the most beneficial capabilities into a proposed solution that you can test.
Test your solution. Because generative AI is so now, you will need to evaluate the technical feasibility of the solution, how much of the problem is solved, and the cost-effectiveness of the solution.
A Fast-Follower Approach
You can use proven solutions other companies have already implemented to solve similar problems. Even if your issues are unique, example use cases and solutions are helpful stimulus material for brainstorming new solutions. Many technology and consulting companies provide example customer stories and generative AI use cases, hoping you will choose their services. The following are use-case examples or solution frameworks from a select set of companies:
Cloud infrastructure providers: Microsoft Azure, Amazon AWS, and Google Cloud
Consulting firms: McKinsey, BCG (Marketing, IT), and Deloitte
What Now?
For those who want to increase their usage of generative AI, the above provides a strategic framework to identify potential use cases along and more tactical guidance built from the commonality of early winners. Hopefully, these frameworks will help you think about your next actions.