New on the Log: a hidden road-block to industrial modernisation, a small, decades-old boxes quietly running most of the industrial world. This article is about why almost nobody wants to touch the code inside them, and the Beam-built startup trying to change that.
PLC: The Box Nobody Wants to Open
Somewhere on a factory floor, a conveyor belt is running on code written before its newest operator was born. Packages move, sensors trigger, motors stop and start. Inside a control cabinet, a small industrial computer keeps the sequence running. It’s called a programmable logic controller, or PLC. While newer generations have become far more capable, many factories still depend on older systems that do their job reliably, day after day. The problem starts when something needs to change—and nobody is quite sure how the code works anymore.
Built to last, which is exactly the problem
PLCs are rugged by design, built to survive tough environments, they can last thirty or forty years on a factory floor without complaint, and that’s precisely why so many of them are still running: a system with near-zero failure rates for decades is, by most plant managers’ math, a lot less risky than whatever untested benefits modernisation may bring. The code running underneath isn’t Python or Java. It’s a proprietary, decades-old language, originally written by electrical engineers who just needed a way to program the things themselves, and it’s wired so deeply into everything around it (older HMIs, SCADA systems, sensors, drives) that touching one machine can ripple across an entire line.
So factories tend to just leave it alone. The math behind that caution is not abstract. Unscheduled downtime costs the world’s 500 largest companies an estimated $1.4 trillion a year, about 11% of their revenue. In automotive manufacturing specifically, an hour of downtime can run as high as $2.3 million, roughly $600 a second. One of our founders, Farhan Abdulbari, spends his working days talking to the people who actually run these lines. He describes meeting a chip manufacturer in Portugal running nine different versions from the same PLC manufacturer, refusing to touch any of them, for exactly this reason. Multiply that instinct across every factory, warehouse, and parcel sorting line still running 1990s code, and you get an industry quietly sitting on a very large, very old pile of software nobody wants to be the one to break.
The people who understand it are retiring
A reason this topic is getting so much attention right now comes down to one thing: the people who understand this code are retiring, and not much of what they know ever made it into a document anywhere. If it broke, or they added a new component, they simply reprogrammed the PLC. Germany, a heavily industrialized nation, is a useful window into how fast that’s happening. Germany had roughly a million 20-year-olds a decade ago; today it has about 800,000. Nearly two million Germans aged 63 drew a pension in mid-2022, some 400,000 more than expected. Over half of German industrial companies report open positions they can’t fill, and about a third of technical apprenticeships went unfilled last year.
Siemens’ Matthias Pohl frames the industry’s response to this in exactly those terms: less about writing code faster, more about knowledge preservation, catching what’s in an engineer’s head before that engineer retires and takes it with them. It’s not a coincidence that interest in this problem is picking up all at once, across large industrial software makers and small independent teams alike. And the market is waiting for a good answer, not a hypothetical one.
Enter PLC Wizz
This is exactly the problem PLC Wizz is trying to solve, in Farhan’s own shorthand: the ChatGPT of PLCs. The idea isn’t to replace the engineer standing in front of the machine, it’s to give them a straight answer instead of hours spent digging through decades-old code nobody left any notes on. It’s one of the companies we’ve built here at Beam: we hand our Entrepreneurs in Residence a real, already partly validated industrial problem instead of asking them to invent one from scratch, and Farhan and his co-founder Prativesh Namdeo are the ones who picked this particular box to open.
PLC Wizz does three things. Query lets an engineer ask a plain question, why is the alarm on Conveyor 3 going off, and get back a shortlist of likely causes read straight out of the actual code instead of a manual line-by-line hunt. A documentation feature turns undocumented legacy code into a plain-language explanation of what each part is actually doing, the exact knowledge-preservation problem a retiring generation of engineers leaves behind, solved one factory’s machines at a time. And a code generation tool drafts new code, for adding a sensor, say, though never as the final word: every suggestion still goes through a person before it goes anywhere near a live system.
All three lean on the same constraint. PLC code is exactly the kind of thing a car manufacturer, a pharmaceutical company, or a defense contractor doesn’t want leaving its own machines, so PLC Wizz runs mainly as a desktop application rather than a cloud service, built to meet the code where it already lives instead of asking anyone to upload it somewhere else.
Farhan wanted to build something with AI. Eighteen months ago, he’d never heard of a PLC. Now he’s deep enough inside the world of the engineers who live with this problem daily, trading notes with them on Reddit and LinkedIn the same way they trade notes with each other, that some of PLC Wizz’s earliest and most engaged users found the product there first. That’s the kind of trade Beam is built around: hand a founder a real problem, and watch them turn into a genuine expert in something almost nobody spends a minute thinking about, yet something that quietly touches everyone’s daily life anyway, one conveyor belt, one packaging machine, one production line at a time.