AI is creating new problems for business. Here's how to solve them.

The initial excitement of AI adoption is giving way to a more sober reality. As businesses integrate AI more deeply, they are encountering a new class of problems, including runaway costs, uneven productivity gains, and the long-term risk of deskilling their workforce. The AI Daily Brief highlights these emerging challenges, suggesting a shift from tactical implementation to strategic oversight is now critical for sustainable success.

The AI Daily Brief podcast recently explored the second-order effects of widespread AI adoption. While AI is proving effective at solving existing problems, its integration is creating an entirely new set of operational and strategic challenges for organisations.

These are not minor teething issues; they are fundamental problems that require new ways of thinking about technology, workflows, and people.

The new operational hurdles

The discussion identified five core problems that businesses are now beginning to confront:

• AI Slop: The internet and internal knowledge bases are being flooded with low-quality, AI-generated content. This makes it harder for staff to find reliable information and for businesses to cut through the noise with their own marketing.
• Rising Token Costs: The shift from small-scale experiments to full-scale production use of large language models (LLMs) can lead to unexpectedly high and escalating costs. What seems cheap for one user becomes a significant operational expense across an entire organisation.
• Uneven Productivity Gains: AI doesn't lift all boats equally. Studies are showing that while top performers get a significant boost, lower-performing employees may not see the same benefits, and in some cases, their performance can even decline. This can widen skill gaps rather than close them.
• Workforce Deskilling: Over-reliance on AI for core tasks—like writing, coding, or analysis—risks eroding the fundamental skills of your workforce. If employees only ever edit AI outputs, they may lose the ability to create from scratch, making them less adaptable and innovative.
• Preserving Human Expertise: As AI automates more complex tasks, there is a long-term risk of losing deep, tacit institutional knowledge. If the human experts who built and refined processes retire or leave, their unique insights might disappear if not properly codified.

Here is a summary of the challenges and the emerging responses:

For business owners and operators, these problems signal that the era of naive AI adoption is over. Simply providing access to a tool like ChatGPT or a foundation model API and hoping for productivity gains is a flawed strategy. Success now depends on anticipating and actively managing these second-order effects.

From tool to system

These challenges are particularly relevant as businesses move towards adopting agentic AI systems. An autonomous agent, tasked with executing a workflow, can rack up enormous token costs or produce vast quantities of low-quality 'slop' without tight controls and clear performance metrics. The risk of deskilling also increases when an agent handles a process from end to end, removing the human from the loop entirely.

Implications for businesses of all sizes

• Small Businesses: You are most vulnerable to rising costs and AI slop. A surprise bill for API usage can cripple a small budget, while relying on faulty AI-generated information for a critical decision can have outsized consequences. Deskilling is also a potent risk when your team is small and each member's core expertise is critical.

• Large Enterprises: Your biggest challenges are uneven productivity and preserving expertise. A widening gap between AI-savvy teams and the rest of the organisation can create internal friction and operational silos. The gradual erosion of deep, institutional knowledge across a large, aging workforce is a serious, long-term risk to your competitive advantage. You have the resources to build robust training and knowledge management systems, but it requires deliberate, executive-level focus.

Ultimately, these new problems force a crucial shift in perspective. The goal is not to 'implement AI', but to build a resilient, human-centric organisation that leverages AI effectively. This requires a more thoughtful approach to workflow design, training, and risk management.

Moving forward requires a proactive and strategic approach. It's no longer enough to just adopt the technology; you must manage its impact on your people, processes, and finances.

1. Conduct an AI Risk and Opportunity Audit. Before you deepen your AI investment, map out where you are using it and where you plan to. For each use case, assess the potential for the problems listed above. Ask questions like: What is the budget for this workflow? How will we measure the quality of the output? What core skill might this tool erode over time?

2. Implement Cost and Quality Controls. Treat AI usage like any other utility. Implement dashboards to monitor API calls and token consumption. Set budgets and alerts for specific projects or teams. Explore using smaller, fine-tuned, or open-source models for specific tasks, as they can be significantly more cost-effective than a large general-purpose model.

3. Redesign Workflows with Intention. Don't just bolt AI onto an existing process. Rethink the workflow from the ground up. Clearly define the roles of the human and the AI. Is the AI a drafter? An analyst? A final reviewer? Designing clear 'human-in-the-loop' checkpoints ensures quality and keeps your team engaged.

4. Invest in "How to Think with AI" Training. Shift your training focus from simple prompt engineering to critical thinking and partnership. Teach your team how to evaluate AI outputs, when not to use AI, and how to use it as a sparring partner to refine their own ideas. This builds judgment, not just dependency.

5. Launch a Knowledge Preservation Initiative. The rollout of AI is the perfect catalyst to finally document your core business processes and the expert knowledge behind them. Use interviews, process mapping, and even AI transcription tools to capture the 'why' from your senior experts. This creates a valuable asset that can be used to train both new employees and future AI systems.