Opportunities

The AI Productivity Paradox: Why Aren’t More Workers Using AI Tooks Like ChatGPT?The Real Barrier Isn’t Technical Skills — It’s Time to Think Despite the transformative potential of tools like ChatGPT, most knowledge workers aren’t utilizing them effectively. Those who do tend to use them for basic tasks like summarization. Less than 5% of ChatGPT’s user base subscribes to the paid Plus version, indicating that a small fraction of potential professional users are tapping into AI for more complex, high-value tasks. Having spent over a decade building AI products at companies such as Google Brain and Shopify Ads, the evolution of AI has been clearly evident. With the advent of ChatGPT, AI has transitioned from being an enhancement for tools like photo organizers to becoming a significant productivity booster for all knowledge workers. Most executives are aware that today’s buzz around AI is more than just hype. They’re eager to make their companies AI-forward, recognizing that it’s now more powerful and user-friendly than ever. Yet, despite this potential and enthusiasm, widespread adoption remains slow. The real issue lies in how organizations approach work itself. Systemic problems are hindering the integration of these tools into the daily workflow. Ultimately, the question executives need to ask isn’t, “How can we use AI to work faster? Or can this feature be built with AI?” but rather, “How can we use AI to create more value? What are the questions we should be asking but aren’t?” Real-world ImpactRecently, large language models (LLMs)—the technology behind tools like ChatGPT—were used to tackle a complex data structuring and analysis task. This task would typically require a cross-functional team of data analysts and content designers, taking a month or more to complete. Here’s what was accomplished in just one day using Google AI Studio: However, the process wasn’t just about pressing a button and letting AI do all the work. It required focused effort, detailed instructions, and multiple iterations. Hours were spent crafting precise prompts, providing feedback, and redirecting the AI when it went off course. In this case, the task was compressed from a month-long process to a single day. While it was mentally exhausting, the result wasn’t just a faster process—it was a fundamentally better and different outcome. The LLMs uncovered nuanced patterns and edge cases within the data that traditional analysis would have missed. The Counterintuitive TruthHere lies the key to understanding the AI productivity paradox: The success in using AI was possible because leadership allowed for a full day dedicated to rethinking data processes with AI as a thought partner. This provided the space for deep, strategic thinking, exploring connections and possibilities that would typically take weeks. However, this quality-focused work is often sacrificed under the pressure to meet deadlines. Ironically, most people don’t have time to figure out how they could save time. This lack of dedicated time for exploration is a luxury many product managers (PMs) can’t afford. Under constant pressure to deliver immediate results, many PMs don’t have even an hour for strategic thinking. For many, the only way to carve out time for this work is by pretending to be sick. This continuous pressure also hinders AI adoption. Developing thorough testing plans or proactively addressing AI-related issues is viewed as a luxury, not a necessity. This creates a counterproductive dynamic: Why use AI to spot issues in documentation if fixing them would delay launch? Why conduct further user research when the direction has already been set from above? Charting a New Course — Investing in PeopleProviding employees time to “figure out AI” isn’t enough; most need training to fully understand how to leverage ChatGPT beyond simple tasks like summarization. Yet the training required is often far less than what people expect. While the market is flooded with AI training programs, many aren’t suitable for most employees. These programs are often time-consuming, overly technical, and not tailored to specific job functions. The best results come from working closely with individuals for brief periods—10 to 15 minutes—to audit their current workflows and identify areas where LLMs could be used to streamline processes. Understanding the technical details behind token prediction isn’t necessary to create effective prompts. It’s also a myth that AI adoption is only for those with technical backgrounds under 40. In fact, attention to detail and a passion for quality work are far better indicators of success. By setting aside biases, companies may discover hidden AI enthusiasts within their ranks. For example, a lawyer in his sixties, after just five minutes of explanation, grasped the potential of LLMs. By tailoring examples to his domain, the technology helped him draft a law review article he had been putting off for months. It’s likely that many companies already have AI enthusiasts—individuals who’ve taken the initiative to explore LLMs in their work. These “LLM whisperers” could come from any department: engineering, marketing, data science, product management, or customer service. By identifying these internal innovators, organizations can leverage their expertise. Once these experts are found, they can conduct “AI audits” of current workflows, identify areas for improvement, and provide starter prompts for specific use cases. These internal experts often better understand the company’s systems and goals, making them more capable of spotting relevant opportunities. Ensuring Time for ExplorationBeyond providing training, it’s crucial that employees have the time to explore and experiment with AI tools. Companies can’t simply tell their employees to innovate with AI while demanding that another month’s worth of features be delivered by Friday at 5 p.m. Ensuring teams have a few hours a month for exploration is essential for fostering true AI adoption. Once the initial hurdle of adoption is overcome, employees will be able to identify the most promising areas for AI investment. From there, organizations will be better positioned to assess the need for more specialized training. ConclusionThe AI productivity paradox is not about the complexity of the technology but rather how organizations approach work and innovation. Harnessing AI’s potential is simpler than “AI influencers” often suggest, requiring only