Why AI is Destroying Your Problem-Solving Skills

(Part 2 of the series “The Human Downgrade”)

A former colleague of mine called me the other week just to vent (as usual). He left consulting to run supply chain at a manufacturer, and he had been CC’d on an email from a warehouse manager addressed to the production planning team.

The subject line read: “Kindly requesting alignment on BOM discrepancies.” The body was four paragraphs of structured, diplomatically phrased nothingness. Every sentence was too careful. Every word was calibrated to avoid offense.

“You know what the problem was?” Well, he asked. A critical component was missing from the Bill of Materials, and the production line had been standing still for six hours. Nobody had picked up the phone.

That email was written by some ChatGPT LLM. And the plant paid the price.

The Illusion of Automated Diplomacy

In my daily work as an ERP consultant, I constantly see that the success of a digital transformation project relies heavily on navigating friction. You have to align production, logistics, and finance. A compelling analysis by Business Insider recently highlighted a growing risk that resonates deeply with my experience: AI is making us worse at handling disagreements.

Professionals are increasingly using generalist AI models as a “politeness filter” to avoid uncomfortable conversations. Instead of writing what they actually need to say, they feed the raw frustration into a chatbot and let the algorithm sand down every sharp edge.

The result is flat, vague messages that lack the urgency required to fix a broken process. In an industrial environment where time is money, this algorithmic diplomacy hides the reality until the supply chain breaks down.

Here’s what a real warehouse manager would have written without the AI filter: “Production planning sent us a BOM with a missing component. Line 3 has been stopped since 6 AM. We need the corrected BOM by noon or we miss the shipping window.”

Direct. Urgent. Actionable. No algorithm required.

Ground Truth vs. Algorithmic Filtering

The root cause here is the exact same issue I often discuss regarding the AI ROI paradox. We’re trying to force the wrong type of intelligence into a context that demands cold precision.

Generalist Large Language Models have zero intrinsic understanding of the physical world. If a warehouse manager is clashing with the production team over an incorrect Bill of Materials, a well-written ChatGPT email asking nicely for cooperation is useless.

You need analysis of the data. You need to pull the BOM, compare it against the physical inventory, identify the discrepancy, and escalate with facts. Using AI to smooth out the tone ignores the physical problem and delays the resolution.

A critical component is missing, and no polite email will magically make it appear on the shop floor. No language model, however well-tuned, will reconcile a BOM. Only a person with access to the system, the inventory, and the authority to escalate can do that.

From Technical to Relational Debt

We talk a lot about technical debt in enterprise IT: those legacy customizations that turn an ERP system into a tech prison. Every shortcut you take today compounds into a bigger problem tomorrow. We all know that.

The same logic applies to communication. When you use AI to avoid a direct conflict, you don’t resolve the underlying issue. You just park it. And just like technical debt, these unresolved tensions accumulate. They sit in a growing backlog of silent inefficiencies, and they start to accumulate interest.

I call this relational debt.

By the time you reach a critical project milestone, a go-live, a system migration, a major process redesign, all those carefully avoided conversations explode at once. The warehouse manager who never got a straight answer about the BOM discrepancy is now refusing to cooperate with the production team. The logistics lead who was always “gently reminded” instead of directly confronted about shipping delays has no sense of urgency. The relational debt has matured, and the interest rate is presenting the bill.

Finance teams track receivables aging on a daily basis. Nobody tracks the aging of avoided conversations. That’s the problem. The cost of relational debt is invisible on a balance sheet, so it never triggers an escalation.

We often recognize that feeding an AI with dirty historical data teaches the system to scale inefficiencies faster than ever. But similarly, using AI to mediate conflicts teaches your organization to ignore warning signs.

Disarming the Politeness Filter

If you want AI to actually help your business without causing leadership atrophy, try to focus on these practical steps.

1. Replace Diplomacy with Precision. Instead of using AI to write better, use it to analyze facts. Ask a specialized Language Model to highlight objective logistical constraints: stock levels, lead times, production capacity. Real data doesn’t need politeness filters. When the numbers say the line is stopped, the conversation must be immediate and focused on resolution.

2. Maintain a High Signal-to-Noise Ratio. Data transparency in a modern Cloud ERP should expose conflicts so you can solve them, not hide them. If the dashboard shows production is halted, the next step is a phone call (yes, they still exist for this reason). Design your communication protocols around urgency tiers, and keep AI out of the critical path.

3. Invest in Hyper-Specialization. Don’t chase the AI model based on the parameters. Focus on tools that understand your specific industry vertical and provide actionable, data-driven insights. For example, a specialized SLM (Small Language Model) that can pull your warehouse routing data and flag the exact discrepancy is infinitely more valuable than a generalist model that writes a beautiful text about it while costing 3 times more for the output.

Reclaiming Operational Friction

True business intelligence requires deep architectural discipline. This applies to your technology stack and your communication framework.

The organizations that will survive the next wave of digital transformation won’t be the ones with the most polished emails. They’ll be the ones where a warehouse manager can pick up the phone, say the BOM is wrong, line 3 is down, and get a corrected version within the hour.

During my years as an ERP consultant, I have never seen a crisis resolved by a nicer email. Every time the line stopped, the person who picked up the phone and walked over to the desk of whoever was responsible fixed it faster than any politely worded request ever could.

Don’t use AI to avoid the conversation. That skill is part of your value as a consultant. Use AI to arm yourself with the data that makes the conversation productive.

Written by Andrea Guaccio 

August 20, 2026