I run a newsletter called Systems For Humans
I'm a software engineer. Not a data scientist, not an analyst, not a philosopher with credentials. But the longer I build software for businesses, the more I see software engineering and information systems collapse into one new practice. Call it behavioural engineering, or behaviour design, or whatever name eventually sticks. The naming is unsettled because the scale is unprecedented. But the practice itself is already running at planetary scale.
So here goes my journey and insights so far.
More people are connected now than at any point in human history. Behind that connectivity sits a cold, monstrous machine that sees people as data points. Zero sentiment.
Every click. Every scroll. Every GPS coordinate your phone hands over. Every call. Every message you sent and every one you almost sent and deleted. You are a collection of data points to the machine, whether you like it or not.
Back in the day, the business that won was the business that was good with people. Now the business that wins is the one that's good with human data points in relation to other data points. Thomas Davenport called this years ago. Every company eventually becomes a data company. By the time Google was running cloud engineering curricula around the idea, it had hardened into industry orthodoxy.
Human vs people
So why call this Systems for Humans and not Systems for People?
A human is the term we use to describe a person as a scientific phenomenon. Your emotions and sentiments aren't analysed for what they categorically are. They're analysed for what actions they drive you to take. Action economics drives the human world. Sentiment drives the people domain.
Back in the day we were told that the business that won was the business that was good with people. That framing was a polite lie.
Coca-Cola didn't win by just being good with people.
It was running one of the most successful behavioural engineering projects of the twentieth century, except it had to do it through television, billboards, and the radio. The craft was always deliberate. What we have today is a higher resolution, low latency, and broader reach version. ==The instruments got sharp enough to operate on you specifically, in real time, on a device you carry everywhere you go.== The practice didn't get invented. It got closer. It's now building an inescapable frame around you.
When an organisation focuses on the human every problem becomes diagnosable. It becomes like a mechanic tuning a car. A surgeon operating on a sedated person. It can engineer systems and manipulate behaviour and anticipate actions. For the first time human experience is officially and truly a scientific economic problem in real observable time.
Why you do what you do as a person doesn't matter to it. What matters is that you did it. We had an interest in it. And we got a record of the event.
Why this matters for builders in Zimbabwe
Look at why most tech startups here fail and why fintech succeeds.
The average indie hacker thinks: what do people want? He builds an app for that. The better question is: what does the system need to be more efficient around this want? Because even if person A wants something, it won't go anywhere if it doesn't make the system more efficient. Fintech apps aren't about making people happy, they are just there to make the finance system more efficient and reliable.
This creates a new problem for developers in places like ours. If you don't have access to data, you don't know what you're really building for. And the people who have the data know its value, so they won't give it away. Most of it is private anyway. In the end, ==we are locked out of the very inputs we'd need to build systems that actually serve the humans inside them.==
The alternative is to build your own data bank. That's a whole new complexity most aren't equipped for, but ethics aside, it's the only honest path for builders in our position. Every fleet management app, every inventory tracker, every tool that gets deployed in Zimbabwe must become a data collection apparatus.
Let's talk philosophy for a bit
You know the Ubuntu adage. Umuntu ngumuntu ngabantu. I am because you are.
For our grandparents, Ubuntu meant we must be empathetic. It confirmed that we build each other in community. From a data oriented point of view, Ubuntu means you as a data point can influence me as a data point upon points of contact.
Ubuntu describes human existence relationally: a person becomes a person through contact with other people. Modern data systems describe humans relationally too, just differently. Social graphs, recommendation systems and behavioural models all reduce people into nodes, edges and probabilities. The philosophical language changed. The relational structure did not.
In the early 20th century, after Europeans got tired of killing each other, a movement called humanism emerged. A philosophy that put human experience at the centre of society to prevent unnecessary suffering. It collected the known data points of suffering and produced a declaration recommending world leaders enhance human life accordingly.
I'm being deliberate with that phrasing. Even the humanist project, the one that's supposed to protect us from being reduced to inputs, has always been a data project under the hood. Governments and institutions cannot respond to suffering one person at a time. Rather they aggregate patterns, statistics and outcomes. The difference is that humanism claims those abstractions should ultimately serve human flourishing rather than optimisation alone.
Where this is going
The future is data-driven. More complex and predictive algorithms will manipulate and drive human behaviour at scales we haven't seen yet. Self optimising profit-driven algorithms are the new overlords.
My biggest fear is this: while humanist organisations fight governments to protect human rights, a new frontier for humanity is already being shaped. Systems for Humans means proactively building social and business systems that protect human interests on this frontier. And do not treat humanity as a consumable input or data point to be manipulated in service of the algorithm's grand design.
The line between consumer modelling and political modelling is thinner than people realise. Rayid Ghani spent ten years at Accenture Technology Labs building algorithms that scored grocery shoppers on how savvy they were at hunting deals, how they assembled their baskets, what would move them to switch brands. In 2011 he left Accenture and became chief scientist for Obama's re-election campaign. The same modelling techniques he had used to predict whether a shopper would respond to a coupon were retooled to predict whether a voter would respond to a political appeal. The campaign's analytics operation is now widely described as having been potentially decisive in the 2012 election outcome. The grocery basket and the ballot turned out to be the same problem.
Cathy O'Neil saw where this leads. In Weapons of Math Destruction she argues that once you post something, or take an action that gets captured, it's no longer yours. It becomes the property of the company that controls the API that captured it. There is a version of you, a collection of every action you've ever taken, that is not yours. It belongs to them. And that version can be used in real time to manipulate your friends, your family, your daughters, anyone on the same platform.
Including you.
This isn't speculation. It's already the operating model.
When OpenAI ships ChatGPT, they're not just engineering a model and serving it via API. They're shipping a system that reshapes how millions of people write, think, search, decide. The product is behavioural.
When Meta optimises the feed, the engineering and the information architecture are both subordinate to a single goal: produce a specific behavioural outcome (more time on platform) by routing the right stimulus to the right person at the right moment.
So where does that leave you?
When you finally buy that upcoming PS6 you've been wanting, is it really you?
Or is the purchase a data point the algorithm has been quietly developing for months? It starts as a reel: gaming is life. You click. You watch. It records. Then comes another: instead of alcohol, buy yourself leisure you won't regret. A steady stream of PS-vs-Xbox content ramps up. Until you buy.
At which point in that sequence did your agency live? The very thing humanists insist sits at the centre of the philosophy, where exactly is it?
If you want the practical companion to this, Mthokozisi Mabhena has written it. Poison Your Data walks through Cambridge Analytica, Myanmar, the 2024 Taiwan AI persona networks, and lands on a concrete defence: data poisoning. Go read it. His piece is probably the data point that nudged me into writing this one.
My question is the one upstream of his prescription. At what point does the humanist project still hold? When the systems are this good at predicting us, what's left of the agency humanism puts at the centre of everything?
