Approaching AI Efforts Serially & in Parallel
Generative AI has been and will continue to be a hot topic, yet it’s only one domain of artificial intelligence that your organization can be leveraging to improve business processes. If you’re dipping your toe into the waters of AI, you can approach it serially or in parallel.
Before an ensuing eye roll about “AI”, let’s step back to look at a handful of domains that exist that are likely value additive for your organization:
Generative AI: helps generate new text and rich media content (e.g. ChatGPT, Gemini)
Predictive AI: predicts a value, category, or recommendations at some level of confidence (e.g. customer LTV, “spam” vs “not spam”, Amazon’s “for you”)
Computer Vision & Document AI: reads images and documents, and is often used with another domain like predictive (e.g. Silicon Valley’s “hot dog” vs “not hot dog”)
Natural Language Processing: interprets language to identify sentiment, recognize speech, and translate between languages semantically (e.g. customer review is “neutral”)
Robotic Process Automation: automates tasks and processes on a schedule or reactive basis (e.g. download file from web server to perform tasks and upload elsewhere)
You’ve likely experienced all of these domains and the current technology offerings make all of these accessible to your organization.
The trick is deciding whether to start integrating AI serially or in parallel.
If you’re answering yes to two or more of these questions, it’s best to get started serially.
Is your team unfamiliar with or have limited exposure to AI technologies?
Is your highest priority project a prerequisite to the next priority’s project?
Are you and the organization generally risk averse and/or resource constrained?
Pick a use case that is low complexity and can easily be reverted to current state; don’t let your first project be the determinant for a concrete opinion of how your organization views AI.
If you’re answering yes to two or more of these questions, it’s best to get started in parallel.
Are use cases relatively independent of one another based on the domains above?
Is there a compelling reason to take on multiple projects in one or multiple domains (e.g. strategic plan is predicated on a first- or an early-mover advantage)?
Are you and the organization generally experimental in nature and/or have flexibility with resources?
Pick use cases that are independent of one another, even if those might be integrated later (e.g. an NLP chatbot collects data and documents, documents are parsed for specific information, predictive model identifies fraud).
TLDR: It’s time to get started with AI initiatives and you can approach it serially or in parallel. Not getting started is not an option.
