Step 5: Waving the white flag

This is the last installment of the series documenting my foray into employing AI as an author/assistant for a book project. You may gain perspective if you read some of the earlier posts before this one.

As we’ve seen in several of the last posts (Steps 1, 2, 3, 4), ChatGPT offers rich text outputs that don’t quite work as intended.

For my viewpoint as a business writer, it’s not really very helpful except perhaps as a preparatory exercise. Virtually none of the text that I generated from my detailed outlines was usable — leaving me exactly where I started, or maybe even a bit behind where I started. A very interesting timesuck, but still a timesuck.

It is indeed true that I benefited from new insights, however inadvertent, from ChatGPT. But after generating tens of thousands of words from my eighteen detailed chapter outlines, ChatGPT failed to generate anything that didn’t sound stilted or artificial.

Reclaiming the writer’s hat

Without really thinking about it, I instinctively took what ChatGPT wrote, threw it away, and began to write my own versions of its paltry prose. After a few days I had written several chapters — integrating some of what ChatGPT had provided but only with severe modifications that made it unrecognizable.

But hold on a second — why was I writing? Wasn’t ChatGPT supposed to remove this drudgery from my workday? Wasn’t its clever purloined IP supposed to make me look as though I could write a book in a few minutes?

Is ChatGPT really a useful tool? Doubtful, except maybe as a prompt to me, the human writer. I did find value in its “ideas” but they were not readymade and useful in any practical sense. I am reminded of those unboxing videos and all the suspense that accompanies the unwrapping of a new toy. Not so much here — the unboxing of the ChatGPT text was lukewarm at best. My immediate reflexive reaction was to rewrite it, which is not an ideal response to a supposedly gamechanger writing powerhouse.

Now what?

My experiment in writing a dormant book proposal with ChatGPT resulted in a disappointing failure. Of course it’s more than possible the failure was mine but the result is the same: the blandness of this ChatGPT text is not bookworthy.

Suddenly I am faced with the prospect of a dead-or-dying book proposal that will not be resurrected by ChatGPT, much to my dismay. What now? Keep at it, by striving to improve my prompt engineering? Give up, and write the damn thing myself?

I am leaning toward the latter. Writing is a natural action for me, like thinking or breathing. Rediscovering the book proposal was reward enough for me to have undergone this disappointing exercise.

For those who are not natural writers, ChatGPT may seem like it’s going to be a godsend. It is not. It will not answer your prayers, or serve your purpose. It will not connect with your most important audiences. Generative AI requires a dedicated human filter with the discretion to know when it is running your communication aims into a digital ditch, as opposed to when it is making its meager contributions of (perhaps) actually useful help.

Everything gets worse from here

The promise of AI will undoubtedly be destroyed by bad actors. AI’s usefulness in practical business situations (and there is indeed some genuine usefulness) will be undermined and engulfed by the bad-actor applications of its freaky power: misinformation. Already we can see that the Russian bots who flood Twitter (i.e., X) rely on ChatGPT to invent and scale politically targeted nonsense in tsunami-level amounts.

As we see above, Russian propagandists forgot to pay their ChatGPT bill. If comedy is tragedy plus time, then we are in for a long spell of tragedy before AI becomes funny again.

Conclusion or concussion?

Now that we’ve all been thoroughly flogged by the ubiquity and the relentlessness of AI headlines, promising amazing results with no end in sight, what conclusions can we draw from this “innovation”?

  • Superficially impressive is not the same as practically useful.
  • Human experience is not replicable by silicon.
  • Believing in shortcuts could direct you off the nearest cliff. Like driving with Google Maps, but without your own eyes.


Your comments are always welcome.

Curious? Contact me. Find me on LinkedIn.


You may find of interest:

Step 4: Fast forward   [7th in this series]

Step 3: Character references  [6th in this series]

Step 2: Injecting personality into AI text  [5th in this series]

Step 1: Giving the outline to AI   [4th in this series]

I asked AI to write the book I never got around to [3rd in this series]

Ways to use generative AI text in marketing [2nd in this series]

Beyond perfect [1st in this series]

Q & AI

Claude vs. ChatGPT


Image source (talked to death)