BoSacks Speaks Out: AI, Publishing, and the Last One Percent

BoSacks Speaks Out: AI, Publishing, and the Last One Percent

I recently read an interesting article about what is called “The Plateau Effect.” It was not about publishing, but it might as well have been.

The piece, by Bob Sullivan, used Waymo’s self-driving cars as the example. Thousands of vehicles had to be recalled because some could not properly recognize construction zones: cones, lane closures, and temporary signage. These were not rare anomalies. These were not exotic edge cases involving a yak, a fog bank, and three tourists from Ohio. Construction is as ordinary as traffic lights. In America, it may be the one infrastructure project that never ends.

And yet, the future of transportation stalled at orange cones.

Sullivan’s point was simple and powerful. The first 90 percent of technological progress can feel miraculous. The next 9 percent can still look impressive. But that final 1 percent, the messy layer of exceptions, judgment calls, ambiguity, and real-world unpredictability, can be brutally hard.

That is the plateau.

And publishing is about to hit it.

AI has entered our business at full speed. It writes summaries, drafts headlines, generates images, builds newsletters, mines archives, translates copy, suggests edits, and produces marketing content on demand. It promises to do almost everything except make coffee and explain why the printer is jammed. Give it time on the printer. It will probably blame production.

Let me be clear. I am not anti-AI. I use it. I value it. For publishers trying to do more with fewer people, it can be enormously helpful. It accelerates workflows, organizes thinking, and handles the first pass on routine work.

But speed is not judgment.

That is where the plateau begins.

AI is strong at first drafts, summaries, pattern recognition, and structured tasks. It can turn chaos into something that looks like order. But publishing does not operate in neat lanes.

Publishing lives in the construction zone.

We deal in nuance, tone, context, and verification. We work with facts that appear simple until an expert reads them. We handle quotes that must be confirmed, claims that require scrutiny, images that may carry rights issues, and sponsored content that demands clear labeling. We publish stories that can be technically accurate and still editorially wrong.

That last problem, to be fair, predates AI. The publishing industry has long been capable of producing its own mistakes without machine assistance.

AI simply makes those mistakes faster, cheaper, and more convincing.

That is the risk.

An AI-generated article can read smoothly, sound authoritative, and feel complete while quietly getting something important wrong. It can deliver errors with confidence. And in publishing, the smallest error can carry the largest consequence.

The last 1 percent is not a rounding issue.

The last 1 percent is the libel claim.

The last 1 percent is the fabricated quote.

The last 1 percent is the copyright violation.

The last 1 percent is the angry advertiser.

The last 1 percent is the subscriber who decides your brand is no longer trustworthy.

The last 1 percent is the correction that never should have been necessary.

That is why we must not confuse productivity with reliability.

They are not the same.

A tool that makes work faster is valuable. A tool that makes bad work faster is a liability with a user interface.

We have seen this pattern before. Every major technology arrives wrapped in promises: efficiency, scale, disruption, growth. The demos are flawless. The projections are optimistic. The adoption curve is presented as inevitable.

Then reality intervenes. People are inconsistent. Workflows are complicated. Institutions are slow for a reason. And language is not math with better branding.

Publishing is particularly vulnerable because the pressures are real. Revenue is under pressure. Staffing is under pressure. Traffic is under pressure. Platforms, subscriptions, advertising, search, and production are all under pressure. When AI offers more output with fewer resources, the temptation is obvious.

But more is not better. Faster is not trusted. Cheaper is not sustainable.

The industry has already spent years chasing scale at the expense of credibility: clicks over quality, volume over value, distribution over distinction. AI should not become the next version of that mistake.

AI belongs in the newsroom. It does not belong in charge of it.

The right model is assistance, not replacement. AI as researcher, organizer, draft generator, archive miner, production helper, and tireless assistant. Useful, fast, and occasionally annoying.

But not editor-in-chief.

Because editing is not merely about producing words. It is about deciding which words deserve to be published.

That distinction is everything.

The strongest publishing brands in the AI era will not be those that use it most aggressively. They will be the ones that use it most intelligently and responsibly. They will build workflows that keep humans accountable. They will be transparent with readers. They will protect original reporting, respect copyright, preserve voice, and remember that trust is not a slogan.

Trust is the business model.

And the plateau is coming.

At first, AI will feel extraordinary. Then it will feel efficient. Then it will feel routine. Then the edge cases will arrive.

The source is wrong.

The quote does not exist.

The tone misses the mark.

The image carries legal risk.

The headline misleads.

The summary distorts.

The recommendation engine suggests something absurd.

The chatbot answers confidently and incorrectly.

That is the moment of truth.

Did you adopt a tool, or did you outsource your judgment?

I remain optimistic, but my optimism has reading glasses. I see the opportunity, and I also see the cones in the road.

AI is not going away, nor should it. Like every major shift before it, desktop publishing, the internet, search, social, and mobile, it will become part of the publishing ecosystem. The question is not whether we use it. The question is how wisely we use it.

The first 99 percent of AI in publishing will look impressive.

But publishing has always been judged on the last 1 percent.

That is where credibility lives.

That is where professionalism lives.

That is where the reader decides whether to believe us.

And no machine, however capable, should be allowed to navigate that construction zone without a seasoned human behind the wheel.