Not every company is a software company. That's not a criticism, it's just one of those "it is what it is" things. Your company makes furniture, or runs recruitment or you feed four hundred people a day. Software is something you buy, or pay someone to build and then it's supposed to get out of your way and let you get on with getting on.

I don't think AI is going to destroy development jobs (well not soon anyway). BUT I do think it has already changed what that job looks like. Many businesses are already a year+ into that change without even noticing it happening.

Many companies I talk to already have people in them building stuff internally... whether they know and endorse or even ALLOW it or not.

Three people, three afternoons

Fiona in finance built a data validator. It opens one spreadsheet, transforms and calculates a pile of stuff, and drops the result into another sheet to send to manco. It works. It saves her the better part of two days a month.

Riaan in recruitment built a CV parser, it automatically pulls the info of the candidate out of the CV and rates his "fit" for the role.

Koos in the kitchen built a locally hosted site that shows the day's menu and lets people register interest, so he knows whether to cook for eighty or for two hundred.

None of them asked permission. None of them needed to. They described the problem to an LLM, argued with it for an afternoon, and by home time they had a somewhat working thing.

The village

Now as we make our way through the corporate jungle we happen upon a village in a section that has been cleared... some areas cleared with dynamite, others with a machete.

Fiona's validator landed on whatever stack her chosen model felt like that day. Riaan's parser landed on a different one. Koos got a third. One of them ended up on SQLite, one on Postgres, and one is quietly writing flat text files into a folder on a machine under somebody's desk. Each has its own patterns, its own methodology, and its own direct SQL access baked straight into the code.

Nobody planned this, nobody chose it... it just kind of popped up in the middle of the jungle, one house at a time.

And now you have a little village... Every house built by a different person, out of different materials, facing a different direction, and not one of them on the municipal plan. A disconnected little village of villainous software, running around your business doing real work. LOL, as they say.

One house has a combination lock on the door, another doesn't even have a door and still another doesn't even have walls. Authentication, at best, is "you have to log into my pc to run it". Authorisation, meaning who is allowed to see WHAT once they are through the door, mostly does not exist at all.

The quiet one

Missing authentication is the obvious problem, and it is the one everybody fixes first... because it's the loud obnoxious one with a vuvuzela.

The quieter cousin is the one where people build stuff on their personal PCs at home because they aren't "technically allowed" to do it at the office or on the company laptop. They cart spreadsheets home, build their little house... test it and get it working the way they want. Then they pack it up into a shipping container (a flash drive) and ship it back to the office in their pocket.

That house got built where nobody could see it, which is how you end up with what Riaan has. Riaan's CV parser has legitimate database credentials sitting in its source code. That code lives on a laptop nobody has inventoried, and it reaches HR data. Names, ID numbers, salary history, the lot.

Riaan did not do anything wrong. He had a problem he owned, he used the tools available to him, for the company's benefit. That is the behaviour you want from people.

But that is a POPIA exposure, and a POPIA exposure does not land on Riaan.

It lands on the company.

Fiona was right

I'm not going to be out here going "thou shalt not build thine own software!", it can make sense to do things internally. I did my own accounting and I do some of my own bookkeeping functions internally because of where OrganiCode currently is in terms of size and need.

So... Fiona was right to build the thing.

She understood that problem better than any developer would have, because she had been living inside its edge cases for four years. Paying someone like me to gather requirements, quote, build, test and hand it over would have been a waste of her budget.

Same for Koos and his menu. Small tools, built by the person who actually has the problem, are very often exactly the right call. Some of the most valuable software in any business is the ugly little thing somebody built for themselves on a Tuesday.

The tools are not the problem.

Nobody owning the village is.

And it all runs fine until Fiona leaves

Then the validator stops running.

Nobody notices for three months, because nobody knew it existed as a thing that could stop. And manco has spent a quarter making decisions on numbers that quietly went wrong in April.

So this is why I think developers are not going anywhere. But the job is not what it was, and anyone still describing it as "writing the code" is describing the cheap part.

The job was never typing lines of code. The job is knowing what should not be built. Knowing what has to be owned. Knowing which house is genuinely fine as it is, and which one is a liability with a nice front door. And knowing who is holding the keys when somebody resigns.

That job just got bigger, not smaller. Devs do not disappear in this story. They stop being the people you hand a spec to, and start being the people you hand the plan to. And me, I love the idea of co-operating with businesses that want to build stuff. Together we dream, together we come up with what they want. But with experience you're able to identify which of the houses need extra padlocks, which need underground tunnels connecting each other... and that maybe you need a post office to move boxes between the houses... maybe on a schedule.

So what are we actually offering

The pitch here isn't "hire me to build your village", which is the CLASSIC developer move because "I cAn dO tHiNgS bEtTeR" (we devs like to think that we've done our time so we're good at this building things thing ok? - we can end up building the most beautiful 7 storey monstrosity you've EVER seen... when a single storey storeroom was needed). That pitch is an expensive answer to a question nobody has asked yet, and I'm not here to just take people's money just because.

Talk to me, and we can look at your village and survey the land. Let's clear the fog of war together and see where you stand.

Sometimes the answer is paint all the houses the same colour and put locks on the doors. Standardise the stack, put proper authentication in front of the lot, and get those credentials out of the source code and somewhere sensible.

Sometimes the answer is leave Fiona's validator exactly where it is. It works, it is hers, do not touch it. Just write down that it exists and who owns it now.

And sometimes there is a real problem hiding in there, and you would rather meet it on a Tuesday afternoon with us than on a Friday night on your own.

You want to know which of the three you have before you spend anything on it. That is a conversation, not a project, and it costs you nothing but that conversation.

If any of that sounds like your setup, send us a mail and we will tell you which one it is.

If you want to see what it looks like when somebody DOES own it: how we turned a 12-hour, 3-person job into a 5-minute click.