In a recent episode of GAEA Talks, host Graeme Scott sat down with David Reed, Chief Knowledge Officer at DataIQ, for a timely and important discussion on the ROI of AI in business. The conversation was spurred by a recent MIT report that revealed a staggering 95% of generative AI proofs of concept (PoCs) are failing to achieve a return on investment. David explained that this isn’t just a shocking headline; it’s a critical wake-up call that challenges the narrative of AI as an instant, transformative solution.
The problem with the “Go-to-Market” approach
According to David, the high failure rate stems from a fundamental disconnect. On one side, AI developers have adopted a consumer-driven, “freemium” model, pushing for rapid, widespread adoption. On the other side are commercial enterprises, particularly those in regulated sectors, that move at a much slower pace due to strict security and accountability requirements.
David pointed out that this has created a “grey IT” problem, where employees, already familiar with generative AI from personal use, are using these tools for work tasks on their own devices. This bypasses corporate security protocols, leading to significant risks like data commingling and privacy breaches. As David went on to say, “The rules of business—like protecting confidential data and ensuring regulatory compliance—haven’t changed,” and the “rush to go” approach of generative AI often collides with these non-negotiable requirements.
From PoC to Production: Redefining “Success”
David argued that the high failure rate of PoCs shouldn’t necessarily be a shock, since PoCs are by nature experiments designed for testing and learning. The problem, he said, is that many business leaders have been led to expect instant, quantifiable returns, mirroring the “move fast and break things” ethos of the tech world.
To achieve real ROI, David explained that companies must shift their focus from broad, generic models to highly specific, business-centric applications. Instead of using a vast, general-purpose model, businesses should use models trained on their own proprietary data. This approach allows them to address specific problems in a secure and scalable way. For example, he noted that a luxury brand doesn’t need a model to be an expert on all of human knowledge; it needs a model that can authentically represent its brand’s values and language. By focusing on augmentation rather than replacement, David said, businesses can use AI to empower their human workforce, helping them make better decisions and focus on high-value, creative tasks.
The Human Element: The Real AI ROI
Throughout the episode, David consistently highlighted the crucial role of the human element. He explained that the most successful AI implementations are those that integrate with and support human intelligence, which requires not just a technology strategy but a talent strategy.
David introduced a key concept called “connective labor,” which refers to the invaluable, often unquantifiable work of human connection, communication, and emotional intelligence that drives a business. While AI can automate tasks and analyze data, it cannot replace the human interaction that builds relationships with customers, inspires a team, or navigates complex business relationships.
David concluded that this is where the real ROI comes from: using AI to enhance human capabilities, not to eliminate them. According to him, businesses can move from the precarious “investment economy” of AI hype to a practical, demonstrable “ROI economy” by focusing on:
- Introducing AI that solves specific problems and not adopting a one-sized fits all approach.
- Security and responsible data handling from day one.
- A clear talent strategy that empowers through training and education.
Ultimately, David drove home a crucial point. True success with AI isn’t found in a frantic race to adopt new technology, but in a deliberate, intelligent strategy that centers on your talent.
.
For more information, visit our Podcasts page or subscribe to GAEA TALKS on your preferred podcast platform.
