The Two Gravities: Why Companies Would Rather Save Than Improve

Picture two proposals lying on the table in a meeting. The first promises that a process will run with three fewer steps from now on, same output, less time, less friction. The second promises that the result will get noticeably better for the customer, though nobody in the room can say exactly when that will show or by what measure. Both proposals are good, and yet you already know, before the discussion even starts, which of them will get the budget in the end. The first will go through because it can prove itself in the next quarterly number, and the second will be postponed because it only proves itself somewhere and somehow.

This is where a force is at work that you find in almost every organization once you start paying attention to it. There are two basic directions a company can send its energy in. One direction wants to improve the result, the product, the service, the customer experience, the outcome, whatever you call it. The other direction wants to avoid effort, meaning to hold the current state of the result but with less work, less time, less cost, fewer errors. Both are legitimate, both go into the calendar as "improvement," and yet the second is the more attractive one, for reasons that have nothing to do with laziness.

The cheaper force proves itself at once, the costlier one only later

The difference between the two forces is at its core not a difference in value but a difference in feedback. When you avoid effort, you get your proof on the same day, because the task took four minutes instead of twelve, the team needed two people instead of five, the error rate fell from eight to three percent, and all of it sits at month's end on a slide that nobody can dispute. When you improve the result, on the other hand, the proof waits somewhere in the future, it arrives indirectly, it shows up as a customer who stays, as a feeling while using the thing, as a recommendation spoken three months later, and every part of it is hard to measure and even harder to attribute to a single decision. One force delivers a number, the other delivers a promise, and organizations are built to reward numbers.

This is the real lever, and it is inconspicuous, because the effort-avoiding force is not better, it is only done proving itself earlier. It does not win the argument, it wins the clock.

The calculation that tips the scale

You do not have to believe this imbalance, you can calculate it, and with exactly the metrics that already sit on the table in every controlling department. ROI is a fraction, return minus cost, divided by cost, and even this fraction has a built-in preference. Effort avoidance attacks the denominator, it lowers the cost, and the saving is at the same time its own return, immediately visible and almost certain. Result improvement has to enlarge the numerator, meaning it has to generate a future extra return that is uncertain and arrives late. A metric with cost or capital in the denominator therefore rewards shrinking more reliably than growing, because lowering the denominator is bounded and calculable, while raising the numerator stays unbounded but speculative. This is exactly what Clayton Christensen meant with his uncomfortable line that in trying to maximize the return on capital we lower the return on capital, because a ratio like return on net assets can rise while the company shrinks in absolute terms.

On top of that come two multipliers, both working against result improvement. The first is discounting, because a euro that only arrives in year five is worth only about sixty-two cents today at a ten percent discount rate, and because savings come at once and result gains years later, the costlier force loses a substantial part of its present value simply by waiting. The second is risk weighting, because a saving happens almost for certain while a revenue gain is speculative, and as soon as you multiply the expected value by the probability of it happening, the speculative gain falls to less than half, before any discounting has even started.

Take two projects that both cost one hundred thousand euros. The first avoids effort and saves forty thousand euros a year, with ninety percent certainty, from the first year on. The second improves the result and promises sixty thousand euros of added margin a year, so the potential is half as large again, but only with forty percent certainty and only from the third year on, because a better customer experience needs time before it shows up in the numbers.

Illustrative worked example, not an empirical survey. Assumptions: 10 percent discount rate, expected values risk-weighted.

The project with the larger potential produces a negative net present value and is rejected, the smaller one goes through, and the only difference lies in uncertainty and delay, not in the worth of the idea. The payback period hits even harder, because many controlling departments filter projects out at a payback cutoff before they even calculate the net present value, and at a common three-year cutoff the result improvement fails before it ever gets a chance at a fair assessment.

And then there is an asymmetry that already sits in the bookkeeping and that no formula fixes. With effort avoidance, both the numerator and the denominator of the ROI formula are observable afterward, the counterfactual is clean, we demonstrably spent less. With result improvement the numerator is an estimate that controlling, under the prudence principle, corrects toward the verifiable, meaning toward zero, so the costlier force gets not just the smaller number but the less believed one, and a safety discount is applied exactly where things have already been marked down.

Why the human brings the same tendency along

One might hope that this imbalance is only a matter of bad metrics and could be fixed with better ones. But it sits deeper, namely in the human who reads the metric. Ever since Daniel Kahneman and Amos Tversky formulated their prospect theory in 1979, we have known that losses weigh more heavily than equal-sized gains, and the later measurement from 1992 puts a fairly precise figure on that weight, because the loss coefficient sits at around 2.25, so a loss hurts about twice as much as an equal-sized gain feels good. Effort is such a loss for our brain. Avoiding it therefore does not feel neutral but like a victory, while the prospect of a better result is only a possible, distant, uncertain gain that the same inner scale automatically marks down.

On top of that comes a second, even older tendency. Psychology has known it since Clark Hull in 1943 as the law of less work, the principle of least effort, according to which humans and animals, for the same outcome, reliably choose the path that costs less energy. Across hundreds of thousands of years this was not a thinking error but a survival strategy, because whoever saved energy in lean times starved less often. So when someone in the meeting finds the effort-saving proposal more attractive, they are not following a bad habit but a very well-built instinct that only happens to sit in an economy where growth comes from getting better and not from saving.

What AI does to the scale

Now comes the part that turns an old tendency into a current problem. The faster an environment changes, the more uncertain it becomes what will come out as a result in the end, and the more uncertain the future gain, the more strongly the scale tips to the safe, immediate side, meaning to effort avoidance. Artificial intelligence is exactly such an accelerator, because it raises the pace of change and with it the uncertainty about every future result, and in the same moment it lowers the cost of saving to almost zero, because it automates effort away faster than any technology before it. Both effects point in the same direction, and the numbers of AI adoption show it too, because in the industry surveys of 2025 seventy-four percent of companies wanted to grow revenue with AI, but only twenty percent already did, and only twelve percent of executives achieved both a revenue gain and a cost reduction at once. The great majority therefore land where the proof arrives at once, at efficiency.

Christensen described where this pull leads long before AI, because he distinguished three kinds of innovation, the performance-improving, the efficiency-improving, and the market-creating, and showed that all three are judged by the same metrics that are wrong for the purpose. Efficiency innovation lowers cost and makes positions redundant, market-creating innovation builds new markets and new work, but because the first pays off faster, capital flows into a loop of ever more efficiency that starves exactly the slower, riskier innovation from which growth actually comes. His line on this is uncomfortably precise: "In our attempts to maximize returns on capital, we reduce the returns on capital." An organization that reflexively saves under uncertainty therefore does not only become more efficient, it becomes systematically less innovative, and the bitter part is that every single one of these saving decisions looked reasonable on its own.

Why appeals change nothing about it

The obvious reflex is to treat the problem as a matter of attitude and answer it with a rallying cry, so more courage, more innovation culture, more long-term thinking. Only that changes nothing, because nobody here chooses the cheaper force out of cowardice, but because the environment rewards it and punishes the costlier one. As long as result improvement shows its benefit late, indirectly, and hard to measure, while effort avoidance delivers a clean number that same afternoon, every reasonable person will reach for the number when in doubt, and they will be right to, judged by what they are judged on. You cannot preach behavior against the incentive situation, against the feedback, and against the human accounting of losses, you can only build it differently.

What an action infrastructure would have to change

And here lies the real task, which no appeal handles, only construction. If the effort-avoiding force wins because it proves itself earlier, then the work is not to forbid saving but to give getting better the same early visibility that saving has by nature. This is action infrastructure, meaning redesigning the work environment, the processes, the metrics, and the feedback loops so that the tendency of everyday behavior tips a little toward result improvement, instead of tipping toward effort avoidance already in the starting state. Concretely this means equipping the slow benefit with early indicators so that a result improvement does not only have proof after three quarters, it means celebrating small result gains as visibly as saved hours, and it means turning loss aversion around so that a missed chance feels like a loss and not just a saved effort like a gain.

Such an infrastructure does not arise on the side, and it belongs to no one as long as it belongs to no one. That is exactly why it needs a dedicated role whose only task is to adjust this scale, someone who treats the behavior of the organization as something designable and not as something that just happens. You can call this role Chief Behavioral Officer or something else, the name is secondary, what matters is that in the end a human is responsible for the costlier of the two forces getting a fair chance at all. Because an organization without such a role does not decide against innovation, it just slides, saving decision by saving decision, to where gravity was already pulling it.