I’m Tired of All the AI Talk—and I’m Part of the Problem

I need to admit something up front: I am one of the people most guilty of contributing to the endless stream of AI commentary.  I have been doing so for the last three years.  Don't get me wrong. I'm proud of launching the first social impact chatbot 

Almost every post I publish on LinkedIn includes an AI take of some sort. I write about adoption, strategy, governance, ethics, organizational change, and the future of work. I participate in many of the same conversations that are now beginning to annoy me.

So this is not a complaint from the sidelines. It is a critique directed at myself as much as anyone else.

I am tired of so much AI talk because much of it feels disconnected from the daily work people are actually doing.

We debate model performance, governance frameworks, existential risks, environmental consequences, labor-market disruption, and potential future pathways. Some of these conversations are important. But they often seem far removed from where the average employee, nonprofit, government agency, or company is today.

Most organizations are not deciding whether artificial general intelligence will create a future of abundance.

They are trying to determine whether employees are allowed to paste meeting notes into ChatGPT.

They are not redesigning society around autonomous agents.

They are trying to understand whether AI can help someone produce a better report, respond to an email, analyze a spreadsheet, summarize a case file, improve customer service, or reduce administrative work.

Meanwhile, the public conversation keeps swinging between enormous promises and enormous warnings.

AI will create an extraordinary future of abundance.

AI will take all our jobs.

AI will solve our most difficult problems.

AI will accelerate inequality, misinformation, surveillance, and environmental destruction.

Perhaps some combination of these possibilities will prove true. But the constant pontifications, predictions, and proclamations can feel strangely disconnected from lived reality. Honestly, even some field research can feel removed from the messy conditions in which people are actually trying to use these tools.

The people writing software are generally much further ahead in AI adoption than everyone else. They have also experienced model improvements more directly. AI coding tools can now generate substantial amounts of code, explain unfamiliar systems, identify errors, write tests, and help developers move faster.

But even there, the reality is rarely as simple as “the AI did the work.”

There is still a great deal of human back-and-forth. There is prompting, reviewing, correcting, refining, testing, rejecting, and trying again. The model produces something. The person evaluates it. The model revises it. The person discovers another problem. The process continues until the result is useful.

That is not failure. It is simply a more accurate description of how the work gets done.

And that is the conversation I increasingly want to have.

What are you actually doing with AI?

Show me the report it helped you build.

Show me the process that now takes two hours instead of two days.

Show me the employee who is serving clients more effectively.

Show me the program that improved its outcomes.

Show me the administrative burden that was reduced.

Show me where the work became more accurate, accessible, responsive, or humane.

I am much more interested in demonstrated impact than another sweeping prediction about what AI might eventually become.

Part of the problem is that we still lack common denominators for discussing practical AI competence. There are few broadly recognized solution pathways. There are limited certifications that provide credible market signals. There is no widely shared vocabulary for distinguishing casual experimentation from repeatable, responsible implementation.

Someone who has used ChatGPT three times can call themselves an AI strategist. Someone who has built a useful workflow, evaluated its risks, trained employees, measured results, and improved it over time may use the same title.

That makes it difficult for organizations to know what expertise looks like. It also makes conversations about adoption unnecessarily vague.

We need more shared infrastructure around practical AI use.

We need recognized implementation methods.

We need credible ways to demonstrate competence.

We need reusable patterns for common organizational problems.

We need better evaluation standards.

We need case studies that include not only the polished outcome, but also the failures, revisions, human labor, costs, and tradeoffs involved.

Most importantly, we need to focus on impact.

Did the work improve?

Did the organization become more effective?

Did employees gain useful capacity?

Did clients receive better service?

Were risks identified and managed?

Was the result worth the time, money, and organizational disruption required to produce it?

Those questions may be less exciting than debating the ultimate destiny of humanity. But they are much more useful to the people responsible for making AI work today.

I do not think we need to stop discussing governance, model performance, ethics, labor displacement, or the long-term future. Those conversations matter.

But they should be connected more consistently to implementation and lived experience.

The AI conversation does not need more certainty, more hype, or more dramatic predictions.

It needs more evidence.

It needs more examples.

It needs more honest accounts of what worked, what failed, and how much human effort was involved.

And yes, I am saying that as someone who has contributed more than my share to the noise.

My goal going forward is simple: fewer proclamations about AI and more demonstrations of what it is actually helping people accomplish.