This article is a departure from our typical content. Wes West, our Chief Analytics Officer, shares his perspective on how AI has unexpectedly transformed millions of knowledge workers into managers, without any of the training or support that role typically requires.
Millions of knowledge workers just became managers overnight. They don't know it yet, and that's precisely why so many AI initiatives are failing. The proliferation of AI tools has fundamentally transformed the nature of knowledge work. Every person who opens ChatGPT, Claude, or any other large language model is now managing an excitable and brilliant, but ultimately contextless intern. The challenge lies in the fact that most of these newly-minted managers have zero training in management and it shows in their results.
I've been extensively pair coding with AI systems, testing different models and exploring how to make these tools genuinely productive for banking analytics. Through this work, I've observed something critical: AI tools behave remarkably like human team members with distinct working styles.
Use the same tool to create ten separate AI conversations with identical prompts, and you'll get consistency, but also surprising variation. Some threads excel at code documentation. Others shine at summarizing complex requirements. A few will be notably better at creating clear next-step descriptions. But here's the kicker — you can't choose which type you'll get. It's completely random, determined by the initialization seed.
Unlike hiring, where you select for specific strengths, each AI session is a roll of the dice. You need to quickly assess what you're working with and adjust your management approach, exactly like being suddenly assigned a new team member you didn't interview. This requires thinking holistically about how to support a wide array of strengths while compensating for various weaknesses, all in real-time.
The implication is profound. Success with AI is no longer solely dependent upon the technology itself. It is also equally, if not more so, about management skills that most knowledge workers have never developed. The technology is more useful than ever in its applications, but now we've hit a new bottleneck: human capability. And this bottleneck will likely persist far longer than it took for the technology to mature.
Organizational hierarchies are designed to foster a natural progression of growth. People typically start as individual contributors, then get promoted to manage small teams, make mistakes, receive feedback and gradually expand their scope. Along the way, many discover they prefer solo work and choose to remain individual contributors.
Now, everyone's expected to be a manager immediately or step back into an entirely new kind of management position. The individual contributors who've spent careers perfecting their craft suddenly need to:
Most people have never learned how to express their internalized expertise. When they type "tell me what's interesting in this dataset," they're making the classic new-manager mistake: unclear delegation with no context.
This is where generic AI tools and their evangelists get it wrong. They promise that AI will magically surface insights, automate decisions and transform your business. "Ask your data anything!" they proclaim, as if the problem was ever about asking questions rather than knowing which questions matter.
I've tested this extensively. If you give an AI tool raw balance data and ask "what grew weird?", it fails spectacularly. These models can see that $250 million and $275 million are different numbers, but they don't inherently know that represents 10% growth, and they don't inherently care whether it's happening in a typical or unusual part of the balance sheet.
A 10% increase in commercial real estate deposits during a Fed tightening cycle? That's unusual and worth investigating. The same growth in retail CDs when you've raised rates? That's exactly what should happen. The AI doesn't know the difference without appropriate context. It's like hiring a brilliant analyst fresh from university and immediately asking them to identify strategic opportunities in your portfolio without explaining:
Generic tools create more work, not less. Product managers end up playing twenty questions with a chatbot, hoping to stumble onto insights. By the time they've typed their fifteenth variation of "what about commercial deposits excluding public funds during non-quarter-end periods?", they could have run the analysis themselves.
The real productivity comes from encoding domain expertise into how AI explores data. At Nomis, we build this into our AI business analyst:
We calculate growth rates, peer comparisons and historical variances before the AI even starts. It knows that a 10% deposit increase means different things in different contexts.
The system understands that deposit behavior varies by customer type, geography and competitive dynamics. It knows to check if unusual growth coincides with a competitor's exit or a new product launch.
Instead of exploring every possible correlation, we guide analysis toward what matters for pricing decisions. Is this growth sustainable? Does it reflect true customer preference or temporary rate shopping?
I've learned that productive AI use requires empathy for the machine — understanding its capabilities and limitations, then structuring interactions accordingly. You can't just hand over a dataset and expect magic. That's a fool's errand, whether you're working with AI or human analysts.
The companies seeing real AI success aren't the ones with the fanciest models or the biggest budgets. They're the ones who understand this fundamental shift: AI tools require skilled management to be productive.
This creates a massive divide. Organizations that recognize the management challenge are building workflows, training and tools that make AI genuinely useful. Those chasing the "full automation" dream are accumulating expensive toys that their teams can't effectively use.
For those of us in the business of delivering insights and not just tools, this reality creates opportunity. While others peddle generic interfaces that turn every employee into a frustrated middle manager with no training, we're building systems that acknowledge the reality that AI only accelerates human intelligence when properly directed.
The management revolution is here, whether we're ready or not. The question is, will organizations recognize this challenge and adapt, or will they keep wondering why their AI initiatives deliver so little value despite so much promise?
If you're interested in how Nomis's pricing optimization tools and AI capabilities can help your team navigate this new reality (without requiring everyone to become an AI whisperer) reach out to us at sales@nomissolutions.com.
Written by: Wes West, Chief Analytics Officer at Nomis Solutions