AI Writing Tools and the Skills They Quietly Replace

AI Writing Tools and the Skills They Quietly Replace

Last year, Sam Altman invited author Dave Eggers to speak to around 200 OpenAI staffers. Eggers has written novels, screenplays, journalism, founded McSweeney's, and started nonprofits that support writers and schools. You might expect him to walk in and offer a polite, measured take.

He did not. He told the room that ChatGPT was silencing an entire generation.

That is a serious claim, made directly to the people building the product. It is worth sitting with, especially if you run a business that uses AI to produce content, proposals, emails, or anything else that carries your voice.

What Eggers Was Actually Arguing

The concern is not that AI writes badly. The concern is that AI writes well enough that people stop practicing writing themselves.

Writing is not just output. It is thinking made visible. When you force yourself to put an argument into sentences, you find out fast whether the argument holds. When you hand that process to a language model, you skip the friction. You get the output without the cognitive work.

For a professional writer, that friction is the job. For a founder drafting a pitch, a clinic manager writing a patient update, or a marketing lead crafting a campaign brief, that friction is also doing something useful: it is making you sharper on the thing you are trying to say.

If you outsource it entirely, you may not notice anything is missing. Until you sit in a room without a laptop and someone asks you to explain your strategy out loud.

This Is Not an Argument Against Using AI

Let us be clear. AI writing assistance is genuinely useful. It saves time on drafts, catches errors, helps non-native speakers write confidently in a second language, and removes the blank-page paralysis that stops a lot of people from communicating at all.

At NUVENAR, we build and deploy AI tools for operators. We use them ourselves. We are not here to tell you to throw them away.

But there is a difference between using a tool and becoming dependent on it in ways that erode underlying capability. That distinction matters at the individual level, and it matters at the organizational level too.

The Organizational Risk Operators Miss

Here is the version of this problem that affects businesses directly.

When everyone on a team uses AI to write internal communications, proposals, and client-facing content, a few things happen over time:

• Voice disappears. Everything starts to sound like everything else. The specific, identifiable voice of your business gets smoothed out into generic professional language.
• Errors compound invisibly. AI outputs confident-sounding text that is sometimes wrong about details specific to your business, your clients, or your industry. If no one on the team has strong enough writing habits to catch the errors, they ship.
• Junior staff stop developing. Writing is how junior people learn to think through problems. If they draft everything with AI from day one, they may never build the judgment to know when the AI is wrong, or when a message is landing badly.
• Critical thinking leaks out. The AI mania problem is real. When decision-makers use AI-generated summaries, briefs, and analyses as their primary inputs, they can lose the habit of interrogating the source. Decisions start to feel informed when they are not.

Simon Willison has been writing about the broader version of this: AI mania affecting organizational decision-making at scale. The pattern is similar whether you are running a five-person agency or a government department. The tool becomes a substitute for the thinking, not a support for it.

Where This Shows Up in SMB Operations

For small and medium businesses specifically, the risk is concentrated in a few places.

Customer communication. If your team uses AI to handle all WhatsApp messages, emails, and follow-ups without review, the relationship quality drops. Customers notice when responses feel generic. In high-trust sectors like clinics, legal services, or financial advice, this erodes the relationship fast.

Sales messaging. The SaaStr data from ICONIQ's 2026 GTM benchmark shows top-quartile enterprise reps carrying quotas of $2.25M and above. At those numbers, generic AI-drafted outreach does not move deals. The reps who perform are the ones who write with specificity, because specificity signals that they understand the prospect's actual problem. AI can help with structure. It cannot replicate the insight that comes from genuinely understanding a customer.

Content that builds authority. If your blog, newsletter, or social posts are fully AI-generated without meaningful human editing, they will be competent and forgettable. Building authority in a niche requires opinions, specific knowledge, and a recognizable voice. Those things require human thinking.

What Good Practice Actually Looks Like

The operators handling this well are not avoiding AI. They are using it with discipline.

Draft, then rewrite. Use AI to get a first draft on the page quickly. Then rewrite it yourself, not just edit it. Rewriting forces you to decide whether the structure is right, whether the argument holds, whether the tone fits your audience.

Keep humans on relationship-critical messages. For anything going to a key client, a prospective partner, or a patient, a human should write or meaningfully edit the message. AI can assist, but not originate and send without review.

Invest in writing skills on your team. This sounds old-fashioned. It is not. Teams that write clearly, think clearly. Short internal memos, regular written updates, documented decisions: these practices build the muscle that keeps your people capable even as AI tools change.

Use AI for the right tasks. Reformatting, summarizing long documents, generating first-pass options, checking grammar: these are good uses. Generating your company's opinion on something it has not actually thought about: not a good use.

If you want to see how we approach AI automation at NUVENAR without replacing the human judgment in your business, our services page covers where we draw those lines in practice.

The Generation Question Is Real

Eggers was talking specifically about young writers and students. His worry is that if you have never struggled to write something difficult, you do not build the capacity to think through difficult things.

For businesses, the equivalent concern is this: if your team has never had to write clearly under pressure without AI assistance, what happens when the tool is wrong, unavailable, or producing output that subtly misrepresents your position?

You need people who can catch that. And you only have those people if you have been building that skill alongside the tool.

OpenAI has been working on teen-specific versions of ChatGPT with age-appropriate controls and parental oversight, which suggests even they are grappling with the question of what appropriate AI use looks like for people still developing fundamental skills. Operators should be asking the same question about their own teams.

A Practical Audit You Can Do This Week

Pick three pieces of content your business sent last month: one internal communication, one client-facing message, one piece of public content.

For each one, ask:

• Who actually wrote this?
• If AI generated it, who reviewed it and what did they change?
• Does it sound like your business, or does it sound like a language model?
• If the content contains a specific claim or recommendation, can someone on your team verify it independently?

If the answers are uncomfortable, that is useful information. It tells you where to put attention.

AI tools work best when the people using them are capable enough to know when the tool is wrong. That capability does not maintain itself. You have to build it deliberately, and keep building it.

If you want to talk through how to structure AI use in your business in a way that keeps your team sharp, book a call with us. We work with operators who want to use automation well, not just use it fast.