Artificial Intelligence at the Speed of Caution
AI is advancing so fast that even the experts have a tough time keeping up. I’m not talking about the progress towards Artificial General Intelligence (AGI), but rather the speed at which new AI-enabled tools, use cases, technologies, and terminologies are being deployed. Every day there is an updated “best practice” for evaluating AI tools, a new software that weaves in AI features into its product, or a new large language model (LLM) that specializes in a novel use case (e.g., deep research). With technology advancing at the speed of light, organizations should be cautious about which IT modernization efforts to fund. Let’s dive into the AI-centric and AI-adjacent IT investments that are safer.
Data Governance and Data Quality
Every single PowerPoint presentation in an organization is dependent upon the maturity of that organization’s data governance and data quality. In one of my previous jobs as a data analyst, our small analysis team uncovered a new sales dataset every 6 months. These sales datasets fed into almost every analysis where a summary would be dictated in a PowerPoint slide that might end up in front of the C-Suite. There are million dollar decisions being made by fortune 500 companies where the “gold layer data” is questionable. Most organizations are no where near the point where further investments in data governance and data quality has diminishing returns. Improvements here will cascade across all reports, analyses, models, and decisions in an organization — enhancing not only traditional outputs but also boosting the accuracy of AI-driven features and insights.
Documentation
While the jury is still out for many AI use cases, one guarantee over the foreseeable future is the increasing importance of unstructured data (e.g., voice, text, images). Some common use cases that have already gone mainstream include video transcription, voice messages, and computer vision for driverless cars. Improving an organizations’ documentation always had some guaranteed benefits (e.g., information flow), but now it is an investment with asymmetric upside. Imagine having years of conversations (or notes) logged between stakeholders in an organization fed into an LLM. The LLM could potentially speed up onboarding/training new employees, improve internal search engines, or automate the development of rough drafts for various documents or presentations. Improving internal documentation (e.g., API metadata, meeting notes, etc.) should no longer be an afterthought.
Open Ecosystem Enables Experimentation
The blocker in many organizations is bureaucracy and centralization. Simplifying enterprise-wide access to data and IT infrastructure is the easiest way to enable limitless experiments. Think of it as funding many small investments across your organization akin to a venture capitalist model. Many will fail, but some will provide strong returns. Amazon is the most well-known company that follows this model; however, Johnson & Johnson recently adopted this model for AI where they experimented with ~900 AI use cases over the past few years.
Modernize your Software when it Supports a Core Strength
If an organization manages a custom software product that supports a core competency, then it is their responsibility to extend its capabilities to incorporate the new AI features. Custom software products are often a central part of an organization. If organizations such as Airbnb, Fanduel, or Thomson Reuters fail to modernize their custom software, they are putting the organization at risk of losing to a competitor. AI is an existential threat to their proprietary software and data. Staying course is riskier than experimenting with AI when software is a central component of an organization’s profit model (or mission).
Conclusion
What do these safer investments have in common? Many of them benefit the enterprise regardless as to whether each AI use case will pan out. When developing your organization’s AI strategy, it is imperative to consider which trends are likely to continue, which investments are safer, which GenAI/LLM use cases make sense, and whether AI is an existential risk to your business.
~ The Data Generalist
Data Science Career Advisor