AI speed is reshaping customer expectations for enterprise software

It's a new world. Think and the feature exists. "

Here's an example: This evening I realised one of the FinTech Profile hero images I was compiling had some additional hair strands in it. If you've followed FTP any time recently you'll see that our 'house style' demands a headshot with a transparent background.

This is is all well and good. A while ago I created a publishing process to manage this. It's not fully automatic but it's point, click, click, move-the-photo-about, click, click.

However, today I realised the remove-background functionality just couldn't handle the headshot of this one executive because of the dark background. The resulting photo had some unnecessary and weird looking hairs poking out the individual's head.

Not a good look. I tried a few different remove-background models. All had the same issue.

So I fixed it.

Claude [Code], can you make a function at this point to help me edit out stray bits of hair from the photo?"

I looked away for a moment or so on something else.

Done.

The new function was inserted seamlessly into the existing production process.

Seamless. I just draw a little white circle around to remove the errant hairs and press 'apply'. Workflow updated. I think it too about 40 seconds to write and test it.

What's fascinating is that now I have this functionality personally, I expect it at work. And that, dear reader, is the science bit.

How do you support production, bank-grade software, when the internal customers are expecting things to work this fast? How do you bring the same level of robust methodology that delivers bank-grade outcomes, but in minutes rather than months?

The answer is deceptively simple. I remember, at least 8 years ago, sitting on a panel talking about production releases with a chap from Monzo. I think my bank at the time was lucky to get 1 production change to customers in a week. I remember the Monzo chap calmly explaining that they shipped perhaps 400-500 updates a day to customers. How? Robust, sequential, simple-but-focused, mature processes. This was almost a decade ago.

How do you apply think-and-it-exists technology in a bank?

Well, with some care.

There are a lot of issues to think about, not least (for example) data security and sovereignty.

AI, as my colleague Vaishakh Nadgir observed today, collapses the time demand dramatically. It doesn't excuse the removal of vital stages of diligence and consideration.

"But the build bit? Opus 4.6 is making mincemeat of whatever I've experienced before!