AI Only Has Generic, Predictable Opinions. That’s a Problem.

Three ways professionals protect their thinking before the algorithm flattens it

Ingrid had just finished one of the best client presentations of her career. The room was engaged. The questions were sharp. Even the partner who rarely smiled was smiling.

On the way out, a colleague leaned over. “That section on market positioning — really strong. How long did that take you?”

Ingrid paused. She couldn’t remember writing it.

She had reviewed it. She had approved it. The AI had drafted it, and somewhere between opening the document and hitting send, her authorship had quietly left the building.

She smiled, said something vague about the research phase, and moved on. But the question stayed with her.

The machine has no skin in the game

Cormac noticed the problem differently.

After fifteen years as an investigative journalist, he had developed a sharp instinct for language that tried too hard to please everyone. When he moved into brand strategy, that instinct followed him.

Every AI draft triggered it.

Not because the writing was wrong. Because it was always reasonable. Balanced. Safe. The kind of prose that survives a committee because it never risks a clear position.

“It doesn’t have an opinion,” he told Ingrid. “It has a weighted average of everyone else’s opinions. That’s not the same thing.”

He was right.

AI isn’t failing. It’s doing exactly what it was designed to do: produce output that most readers will accept. The issue is that acceptable and useful are not the same. In professional work, acceptable without distinction is just polished noise.

Most people think the risk lies in hallucinations. They fact-check and move on.

The deeper risk is quieter. AI produces work that is correct, coherent, and forgettable. Nothing feels off, so nothing gets questioned. You read it, it sounds fine, and your own point of view never makes it onto the page.

Over time, that’s not a workflow issue. It’s an identity issue.

As Cormac puts it: if a client could swap you out for the tool and get the same output, you haven’t been adding value. You’ve been adding speed.

1. Write before you prompt

Cormac’s first rule is almost stubbornly simple.

Before opening any AI tool, he writes three sentences by hand.

They don’t need to be polished. They just need to be his.

The goal isn’t a better prompt. It’s a better starting point. When you begin with nothing, you’re asking the tool to define the problem. It sets the frame, the tone, and what matters.

By the time you edit, you’re refining its version of the problem, not yours.

The three sentences interrupt that process. They establish a position before the algorithm can.

Ingrid tried it before a competitive analysis brief. Three rough sentences in a notebook, scratched out and rewritten.

When she opened the tool, she didn’t ask for a general analysis. She asked it to pressure-test a claim she had already formed.

The output changed immediately. It was sharper. More relevant. More hers.

The exercise takes less than two minutes. At first it feels inefficient. Then it becomes essential, because the difference in output is obvious.

More importantly, the difference shows up when you have to defend the work. You know what you think, because you thought it before the tool did.

The Mixtape Constraint

Ingrid had been an early adopter of every AI feature her firm released. Research synthesis, slide generation, summaries, draft writing. She used all of it.

Cormac used far less.

“You’re subscribing to everything,” he told her, “and calling it a strategy.”

He compared it to a mixtape. A good mixtape has limited tracks because someone made decisions. They chose what to include and what to leave out. The restraint is part of the value.

Using every AI capability is the opposite. It looks like productivity, but it’s really delegation. And every delegated step is a judgment you no longer make.

The Mixtape Constraint is a simple audit: which tools amplify your thinking, and which ones replace it?

Pick three that amplify. Use them well. Ignore the rest for now.

For Ingrid, those three were:

• Synthesizing research she had already read
• Structuring documents she had already outlined
• Pressure-testing arguments she had already formed

In each case, she brought something to the tool. The AI refined it. It didn’t originate it.

The features she had been using to skip the thinking stage turned out to be the ones eroding her value.

2. Pick what amplifies you. Drop what replaces you.

The distinction sounds obvious, but it rarely feels that way in practice.

Replacing and amplifying both feel efficient. Both produce usable output. Only one leaves you with work you can fully stand behind.

Replacing: you give the AI a topic and accept what comes back. Amplifying: you give it a position and make it work harder than you would alone.

Most professionals are doing the first and calling it the second.

The Mixtape Constraint forces a harder question: which of your habits extend your thinking, and which ones substitute for it?

The answer is often uncomfortable. That discomfort is useful. It points to where your authorship has started to slip.

The Symbiotic Shield

After enough conversations, Ingrid stopped thinking of AI as a tool and started thinking of it as a relationship.

Not one to distrust, but one to manage.

Like any working relationship, it has tendencies. Defaults. Patterns that serve its nature, not yours. AI tends toward the center — toward what is broadly acceptable.

Left unchecked, it pulls your work there too.

Cormac described a “filter layer”: a small set of personal rules that sit between you and the output. Simple enough to apply every time. Strong enough to catch when the algorithm starts to take over.

It’s not a checklist you paste into prompts. It’s a habit.

Ingrid built hers over a weekend. Three rules.

Before: write a single sentence stating your position. Not a topic — a position. Something you could be wrong about.

During: challenge the first draft. Even if it’s good. Find where it softened a point, added a hedge, or stopped short of a conclusion. Push it further.

After: rewrite the opening sentence yourself, without looking at the AI’s version. The opening sets the frame. If you inherit it, you inherit the logic that follows.

Three minutes total. The consistency matters more than the effort.

3. Architect the relationship. Don’t just use the tool.

The point of the filter layer isn’t skepticism. It’s authorship.

AI will reliably produce answers that are reasonable and well-structured. That’s its strength.

Your role is different. You are expected to produce the answer that reflects how you see the situation, what you think should be done, and what the data actually implies.

Those are not interchangeable outputs.

One satisfies most readers. The other justifies your presence.

The difference shows up later

Three months later, Ingrid is in another client meeting.

A partner asks her to walk through the positioning argument.

She does, clearly and without hesitation. She explains the assumptions she rejected, the alternatives she considered, and why this recommendation stands.

Because she wrote it.

The AI helped her refine it. It tested weak points and improved structure. But the thinking was hers before she opened the tool, and the process kept it hers after.

That’s the shift.

Not better AI output. Better ownership of the work.

The decision happens early

The algorithm will always give you something competent. That’s not in question.

The question is whether you bring something into the exchange — a position, a set of deliberate uses, a way of filtering what comes back — or whether you let the tool decide what’s worth saying.

That decision happens before the first prompt.

It happens in the few minutes when you decide whether to think first or generate first.

For most people, that moment is coming up again very soon.

Jean Marie Bonthous (publishing as JM Bonthous) is the author of more than two dozen books, including six on the human side of AI, six about filmmaking, and four about digital/AI art. See his latest books: www.jmbonthous.com

He writes three blogs on Medium:

About the human dimensions of AI: AI in Real Life

About AI art: The Algorithmic Eye

About AI filmmaking: The Solitary Frame