Who does your algorithm serve?

I recently asked a simple question on LinkedIn. Ever since Elon Musk bought Twitter, has X become less useful?

The responses were fascinating, not because everyone agreed, but because almost nobody talked about the technology. Instead, they talked about how the platform made them feel.

One person wrote that they had built their career on Twitter. Another said they had met some of their best friends there. Several described it as their primary source of breaking news and expert opinion. One comment struck me more than any other: "It simply stopped being fun."

My view is that the algorithms have changed. I no longer see my friends or the news sources I chose to follow. Instead, my feed increasingly consists of strangers, rage-bait and content I never asked to see. Whether that's intentional or a consequence of the way the platform now optimises engagement almost doesn't matter. The experience has changed.

Every algorithm has an owner. That owner decides what success looks like. If success is defined as advertising revenue, the algorithm will maximise advertising revenue. If success is defined as engagement, it will maximise engagement. If success is defined as reducing costs, improving efficiency or increasing shareholder value, that's exactly what the algorithm will do. Algorithms don't make value judgements. They optimise whatever objective they are given.

Everything works when the owner's objectives and the customer's objectives are aligned. Problems begin when they diverge. Whether intentionally or not, Twitter under Jack Dorsey felt more aligned with what many users wanted: conversation, discovery and community. Today, many users feel that alignment has weakened.

Most of us joined Twitter because it was a place to discover interesting people, follow breaking news, exchange ideas and occasionally have a laugh. It wasn't perfect, but it felt like a global coffee shop where journalists, technologists, politicians, entrepreneurs and complete strangers could all end up in the same conversation. Your follower count mattered less than whether you had something interesting to say.

Reading through the LinkedIn comments, that is what people miss. They don't miss the bird logo. They don't miss 140 characters. They miss the conversations. They miss the serendipity. They miss discovering people they would never otherwise have met. In other words, they miss what the algorithm used to optimise.

Many users feel the optimisation target has shifted. Visibility appears increasingly linked to subscriptions. Controversy seems to outperform nuance. Political outrage dominates feeds because outrage generates engagement, and engagement generates revenue. Whether that perception is entirely accurate is almost beside the point. If enough people believe the platform is no longer working for them, they quietly stop using it.

This isn't really a story about X. It's a story about almost every digital business.

Banks optimise for return on equity. Airlines optimise for yield. Retailers optimise basket size. Streaming companies optimise viewing time. AI assistants optimise response quality, speed and cost. Every digital business is becoming an optimisation machine.

The danger is not that optimisation exists. The danger is forgetting who the optimisation is supposed to serve.

Take banking. A branch may be losing money, so the algorithm recommends closure. The spreadsheet is correct. Transactions have fallen. Costs remain high. Return on capital is poor. The decision is entirely rational. Yet the spreadsheet rarely measures the elderly customer who wants someone who knows their name, the local businesses that still need somewhere to deposit cash every evening, or the reassurance people feel because there is a branch in the town if something goes wrong. Those things are invisible because nobody asked the algorithm to value them. The algorithm hasn't failed. It has done exactly what it was designed to do Nobody asked the algorithm to optimise for the customer's needs.

This is why AI governance matters so much. People often worry that artificial intelligence will make bad decisions. I'm not convinced that's the biggest risk. I think the greater risk is that AI becomes exceptionally good at making decisions based upon objectives that humans chose badly in the first place.

The lesson from X may therefore have very little to do with Elon Musk. It has everything to do with optimisation. Change the objective function and, over time, you change the experience. Change the experience and, eventually, you change the community. By the time people realise what has happened, they don't complain. They just leave.

Every algorithm embodies a set of priorities. AI makes those priorities visible at scale. That is why every board should ask one question before deploying AI across the business: who is this algorithm working for? If the answer is always the owner and not the customer, don't be surprised when customers begin looking elsewhere.