Decoding AI : Paralysis, Fatigue & Washing

Telmo Silva sits down with Danny Valentino, Director of Digital and Store Technologies at Home Hardware, for a frank, no-hype conversation about the three AI traps quietly draining enterprise tech teams. Two veterans who have lived through AS-400s, the dot-com bubble, and every trend in between take an honest look at what is actually happening on the ground when organizations try to “turn on AI”  and why so many of them are stuck.

Questions We Asked

• Why do experienced tech leaders freeze up when choosing an AI tool, and what does it have to do with deterministic versus probabilistic systems?
• Is the pressure to adopt AI coming from inside the organization, or is it entirely external?
• What does AI washing look like in practice, and is it becoming a regulatory problem?

What We Learned

• AI paralysis is not a knowledge gap. It is the result of too many tools, too many models, and too much pressure to pick the right one before you even know the problem you are solving.
• AI fatigue is real. Technologists are spending more mental energy figuring out how to enable AI than actually figuring out the processes AI is supposed to improve.
• Clean data is not optional. Whether you are running a generative chat or building agents, messy, unmapped data will always surface as imperfect results.
• Start small and stop chasing models. Pick one focused problem, pick one tool that works for it, and move forward.

AI Paralysis: When More Options Mean Less Progress

Danny Valentino did not read about AI paralysis in a report. He lived it. As AI models became more sophisticated, the responses got longer, the tools multiplied, and the decisions got harder. Copilot or ChatGPT? ChatGPT or Claude? And then the pressure from the business arrived: integrate open APIs, monitor outputs, verify accuracy, switch between tools. What was meant to accelerate work started producing something else entirely: an endless loop of second-guessing with no clear exit.

Part of what makes this so hard, Danny explains, is that IT veterans have spent their entire careers in deterministic systems. Same input, same output, every time. AI is the opposite. It is probabilistic by nature, and the gap between what seasoned technologists expect and what AI actually delivers is where the paralysis lives.

We only know a world of deterministic systems. AI is all probabilistic. You ask it something, and I ask it something, and we get different things. That’s part of the paralysis.
Danny Valentino, Director of Digital and Store Technologies at Home Hardware

AI Fatigue: The Hidden Cost of External Pressure

With over $1.5 trillion being invested in AI globally, the pressure to keep up is relentless. Every vendor pitch starts with AI. Chatbots became agents. Agents became agentic. And somewhere in the middle, Dan found himself spending the majority of his mental energy not on the retail and store systems he is responsible for, but on figuring out how to enable AI because the pressure to do so came from everywhere outside the organization.

This is AI fatigue: the mental exhaustion that sets in when technologists spend more time figuring out how they are going to use AI in their processes than figuring out the processes themselves. The irony, as both Telmo and Danny agree, is that most organizations have barely turned on the last thing before the next one arrives.

From Backlog to Pull Request: Agentic AI in the Real World

Not all of the conversation is cautionary. Telmo shares something that genuinely surprised him: his CTO quietly added a new developer to the Agile board. No onboarding. No introduction. Just a coding agent autonomously picking bugs from the backlog, writing fixes, pushing pull requests, and iterating on feedback until the job is done.

It works. But it also raises a harder question: what happens to senior developers when their primary job becomes validating what an AI produces? The answer, Telmo suspects, is that they are not just checkers. They are trainers. And the real value will sit with people who understand how to connect the pieces : chat interfaces, retrieval-augmented generation, AI APIs, MCP servers, agents and without necessarily writing every line of code themselves.

AI Washing: When the Headline Is AI but the Reality Is Something Else

The third term in this episode is the sharpest one. AI washing is what happens when a company exaggerates or misrepresents the role of AI in its products or decisions. The examples are not obscure: BLOCK using AI as justification for mass layoffs, Meta leveraging AI narratives to manage investor expectations, the SEC already fining companies for it. The pattern is consistent: announce AI, stock rises. Hire AI engineers, stock rises. Fire AI engineers, stock rises. The technology almost does not matter. The headline does.

Regulatory frameworks are beginning to catch up. The EU AI Act includes transparency requirements. The SEC is enforcing. But for public companies managing costs and valuations, AI remains a convenient story, and both Telmo and Dan are skeptical that the workforce reductions announced under its banner will reverse.

Where to Start: Focus on the Problem, Not the Model

The episode closes where it should: not with a framework, not with a zone model, but with the simplest possible advice. Before picking a tool, pick a problem. Make it small. Make it low risk. Make sure the data behind it is clean and well-governed. Then pick one tool that actually works for that problem and stop there.

As Dan puts it: “The small thing will solve a big thing and it’ll reduce your work day. Don’t feel the pressure.“

That is the antidote to paralysis, fatigue, and washing alike.