“Nobody Really Knows What They’re Doing with AI” and Why That’s Not New

A striking line from a recent post by Zach Rattner caught my attention:

“At the end of the day, I’m convinced nobody really knows what they’re doing with AI.”

Translated to plain English:
In the end, I’m convinced that no one fully understands what they’re actually doing with AI.

This feeling is not new.

Twenty years ago, when mobile devices and applications first appeared, we were in a very similar situation. Back then, we started transferring the process of creating a packing inventory from paper forms into a mobile app. At the time, we didn’t fully understand why this would be better than writing by hand on a printed form.

We only understood it after we had done it.

Much like the elephant calf in Kipling’s story, whose nose was pulled by a crocodile and turned into a trunk — the real transformation became obviousonly after the fact.

Now, after more than 23 years of practical use, we can clearly articulate why an inventory created in an application is better than a handwritten one.

What Actually Changed in Practice

Here are some very concrete outcomes of that early technological shift:

1. Entries created in the application are written in clear, printed text, not illegible handwriting.
2. Items with identical descriptions are separated into individual lines using a cloning function.
3. The application allows the use of English terminology via autocomplete, enabling staff with limited language skills to select correct terms.
4. Notes about pre-existing damages, including photos, are immediately linked to the specific packed item by number.
5. Damage descriptions and their locations on the item are selected from standardized lists, not improvised wording.
6. At the end of the job, the inventory is generated as a PDF and shown to the shipper on site.
7. If the shipper agrees, the inventory is signed digitally by both the shipper and the crew leader.
8. The PDF can be sent directly from the device to the shipper via email or WhatsApp or Telegram or WeChat.
9. The inventory is uploaded to a cloud portal, where it can be easily retrieved later.
10. The inventory no longer depends on the crew leader’s device — it is not lost if the device is reset, sold, or replaced.

The Familiar Pattern of Technology Adoption

When you look at this list, a familiar pattern emerges. It aligns perfectly with another idea expressed by Zach Rattner:

They automate the boring work first.
They scale human impact, not headcount.
They rethink what teams can do when tedium disappears.

The overlap is almost exact.

People genuinely dislike writing block letters by hand. As a result, they write in cursive — often so poorly that even they themselves struggle to read their own writing later.

And consider a simple example:
If you have 50 boxes of books, writing by hand usually results in something like:
“1–50 books”.

The application, by contrast, produces:

• 1 – box of books
• 2 – box of  books
• 3 – box of books
• …
• 50 – box of     books

Yes, the inventory becomes longer.
But it also becomes verifiable, line by line, without interpretation or guesswork.

A Familiar Lesson for the AI Moment

Looking back, the lesson is clear:
We rarely understand the full value of a new technology at the moment we adopt it. We discover that value only through real use, repetition, and results.

The current moment with AI feels very similar.
Not knowing exactly where things will land is not a failure of thinking — it is a normal stage of transformation.

The difference, as always, will be made not by slogans, but by what actually works in practice.

‍