THE HIDDEN COST OF TOE-DIPPING WITH AI

Decades of social science research have demonstrated a predictable consequence of disruptive new technologies like AI: uncertainty. Anyone who has lived through a couple of tech revolutions like me has seen this.  What’s more, we instinctively FEEL this uncertainty simply because of the sheer number of questions being asked:

• Strategic Questions: Which model should we use? Should we build or buy? Where’s the ROI? What is the real cost of training, running, and deploying AI? Will our tech stack or models be obsolete the minute we launch?

• Operational Questions: How do we integrate probabilistic models with existing systems? What happens if our vendor tweaks its model? How do we address hallucination risks? How do I prevent my employees from compromising privacy and security?

• Employee Questions: Will I lose my job? Does our organization know what it’s doing in AI? Will this create more work for me or make my role redundant? Am I cheating by using AI? What if my manager realizes it only takes me an hour to complete a task that used to take all day?

The Hidden Cost of a Toe Dipping AI Strategy

Faced with such uncertainty, most organizations respond by delaying decisions. They adopt a “wait and see” approach, until a winning platform and a proven ROI emerge from the experience of early adopters.  Scientists call the sheer number of choices we face, choice overload, and the number of decisions we face, decision fatigue. And the response to both is consistently the same: delay.

It’s no wonder that so many organizations adopt a “toe-dipping” strategy for AI. They experiment in a few departments or run small-scale pilots to test the waters.

At first glance, it makes sense. In the face of enormous uncertainty, it seems prudent to pause and to experiment with AI. But here’s the problem: toe-dipping comes with significant hidden costs.

THE HIDDEN COSTS OF TOE-DIPPING

1. Costly Tactical, Not Strategic, AI Use Siloed experiments focus on departmental applications. This limits cross-functional innovation, collaboration, cost savings, and the orchestration of AI towards a more transformative value: competitive advantage. Instead, AI becomes an expensive, incremental tool deployed piecemeal and inefficiently as resources get duplicated. 

2. Silo’d learning and expertise Siloed experiments delay progress, because the learning curve is steep, and the know-how, relationships, and technology investments are fragmented.

3. An Erosion of Confidence from Stockholders, Customers, and Employees Toe-dipping signals a lack of commitment. So employees may question whether they should seek opportunities elsewhere. Customers may look to competitors, already knocking on their door. Investors may lose confidence, potentially impacting stock performance and leadership credibility, putting your job at stake.

   
   

A BETTER WAY FORWARD: ORGANIZATION-WIDE UP-SKILLING

Rather than dipping a toe into AI, organizations should focus on upskilling all employees with company laptops. By training employees in AI fundamentals and relevant use cases, organizations can:

1. Accelerate Organizational Learning While Buying Time Upskilling accelerates learning across teams while providing breathing room to make more informed strategic decisions later on, once models sort themselves out and uncertainties abate.

2. Prepare for a Unified Strategy & Competitive Advantage in the Future Broad AI literacy and adoption creates a workforce ready to collaborate, innovate, and implement a unified AI strategy when the time is right.

3. Ensure Confidence Among Stakeholders Employees, stockholders, and customers see a proactive leadership team that’s committed to staying ahead. Upskilling builds employee trust, enhances morale and retention, positioning AI as a career enhancer instead of a career disrupter.

THINK BIG, BUT START SMALL

While organization-wide upskilling is the goal, start smaller. Aim to initially train ~20% of your workforce – your early adopters and innovators and a few more. About 20-25% is the penetration level at which adoption begins to take off more organically. By initially limiting the number provisioned to the early adopters you actually accelerate adoption. Once the early adopters have validated the tool, scarcity of supply makes the licenses even more in demand among the later adopters who have to wait. You will save time and money over the long term by starting small first and gating their availability. Keep in mind, adoption, does not just mean trained in AI, but use AI daily with increasing proficiency.

CONCLUSION

Toe-dipping is pervasive because it seems like a pragmatic approach to AI adoption. However, the hidden costs can far outweigh the benefits. By investing in widespread up-skilling today, organizations can build smarter, more confident teams equipped to capitalize on the promise of AI tomorrow. Widespread upskilling enables a more strategic and informed approach while avoiding the downside of siloed learning and resource inefficiency. It creates a more certain way forward in an uncertain future while motivating your teams now.

AI STRATEGY

January 2025