Since the beginning of time information was valuable to businesses because it was scarce. Companies invested in market research, CRM systems, customer analytics, brand trackers, forecasting models and dashboards proliferated on the assumption that knowing more would lead to better decisions.
But artificial intelligence is changing the economics. Organisations can now generate more information, analysis, scenarios and recommendations than any management team could plausibly consume. So now the problem is knowing what matters.
There’s a real world out there too.
Most don’t realise it, but nowadays businesses do not manage reality directly; they manage representations of it. The CRM system stands in for the customer relationship; net promoter scores for satisfaction; brand trackers for reputation; sales pipelines for future revenue; KPIs for employee performance. Such simplifications are necessary because reality is too complex to run a company without them. But we need to be aware the a budget is only a budget, best practice is only what we know up to today, and an indicator is just an indicator and a process map is only a map.
Jean Baudrillard explored this dynamic in *Simulacra and Simulation*, arguing that representations can become so powerful that they no longer describe reality but shape what people understand reality to be. In a modern corporation culture I’ve seen it way too oftena nd fallen prey to it’s pitfalls. I have been a deeply unhappy customer customer who appears perfectly satisfied in the CRM; a business can hit service KPIs while customers grow more frustrated; a brand can perform well in tracking studies while becoming less relevant; a sales team can meet activity targets while creating little genuine demand.
Managers are surrounded by data and still misunderstand what is happening. The map is useful until the organisation starts managing the map instead of the customers.
When the measure becomes the target This is where Goodhart’s Law enters the picture. The principle, often summarised as “when a measure becomes a target, it ceases to be a good measure”, captures a behaviour familiar to anyone who has worked inside a corporate. Put incentives to a proxy and people learn how to improve the proxy. Measure call centres by average handling time and calls get shorter, though service won’t improve (my time at C&W). Measure MQL and you may generate more leads, but not more customers (my time at Voda). Measure sales by meetings (my time at Energis) and you will probably get more meetings but not stronger relationships.
The representation begins as a window onto reality; over time it becomes the reality the organisation manages. Rory Sutherland argued that organisations value data for its “defensive capabilities”. A spreadsheet, a forecast or a dashboard carries authority. If a decision fails but the data supported it, the executive can explain why the choice seemed rational at the time. If an unconventional judgement fails, the conversation can be less comfortable.
Over time, we have all drifted from asking “What is the best decision?” to “What is the most defensible decision?”. Baudrillard, Goodhart and Sutherland describe different facets of the same phenomenon: representations replace reality; optimising those representations distorts behaviour; and organisational incentives reward the defensible proxy over the messy truth.
I call it the proxy trap. AI could make this problem more acute because it is becoming extraordinarily good at making judgements. It’s hugely valuable that AI can analyse information, build models and produce more plausible recommendations than human teams could ever have generated. It also means companies can create more maps, scores, forecasts, segments, predictive models, dashboards and apparently rational answers.
The danger of limitless data is that organisations become better at constructing sophisticated representations of reality and more confident in treating them as reality itself.
The return of creative compression - an old lesson from advertising
An old lesson from advertising illustrates why the answer is not less data but to apply better judgement.
Great creatives are highly saught after not because they possessed more information than everyone else. Their value is not in processing the most information in the fastest time, but in knowing what to discard.
Client havethe data; researchers have the research; planners understand the audience; product teams know the features; media teams know the channels. But creatives take all of that and identify the one thing that mattered, compressing enormous complexity into a single human insight, an idea people remember.
AI makes production almost limitless. Firms can generate more reports, strategies, campaigns, scenarios and answers. The constraint is now selection, which becomes more valuable than production. Interpretation has always been more valuable than information and judgement more valuable than analysis alone.
Good judgement is the ability to understand what the data can tell you, recognise what it cannot, connect information from different areas, test assumptions against reality and make a decision when the evidence is incomplete. Experience matters because it improves that calibration. Someone who has worked across different markets, technologies, organisations and business cycles has seen what happens after the strategy presentation. They have watched customers behave differently from the research; seen acquisitions that made perfect sense in a spreadsheet struggle in reality; observed companies optimise one part of the business and inadvertently damage another; witnessed apparently strong strategies collapse when they meet culture, competition or customers.
For these reasons, I believe that the experienced generalist will become more valuable in an AI-rich organisation. Businesses will still need specialists probably more than ever but specialists inevitably see the world through their own lens. Someone has to understand how the pieces interact. The customer proposition must fit the commercial model, the brand promise must match the customer experience, the technology strategy must support the business strategy, the investor story must survive operational reality. Optimising one part of a system can easily damage another and the experienced generalist can see the interactions, trade-offs and unintended consequences between people who know different things very deeply.
We can ask the question that organisations sometimes forget to ask: does this still make sense in the real world?
That may become one of the most important questions of the AI era. Businesses competing to find information their competitors missed is a shrinking advantage. . Increasingly, everyone can access the research, benchmark the market, analyse the data, generate a strategy and produce an answer. So value moves somewhere else. The scarce capability becomes the ability to decide which question matters, which evidence deserves attention, which model still reflects reality and which answer is worth acting on.
Perhaps that is the real paradox of limitless data: the more maps we can create, the more valuable the people become who remember to look out of the window. And that’s me.