# The world has sped up. Annual planning hasn't.

Annual planning was born in a world that changed slowly. Today's technology won't wait for next October.

- Columnist: [Daniel Farias](https://ceoone.com.br/en/columnist/daniel-farias)
- Published: October 29, 2026
- Address: https://ceoone.com.br/en/article/the-world-has-sped-up-annual-planning-hasnt
- Em português: https://ceoone.com.br/artigo/o-mundo-acelerou-o-planejamento-anual-ainda-nao.md

![A thick planning binder with coloured tabs on an office desk, next to printed reports and an open laptop, in the late afternoon](https://pub-b80e49b2737940b3ab5d3f20d0578192.r2.dev/artigos/o-mundo-acelerou-o-planejamento-anual-ainda-nao.webp)

Produced by AI

## Key points

- Annual planning, decided between September and November, carries out its last decision fifteen months after approval; in 2026, an executive saw three or four generations of AI models released.
- Uber spent in four months the budget planned for the whole year on AI coding tools, and corrected course with a monthly cap per employee, real-time measurement and continuous review.
- The author proposes splitting the plan by how long each part stays valid: capital on the annual cycle, technology bets on quarterly cycles and AI initiatives on a continuous cycle of testing, measuring and adjusting.

In my first column here, I wrote that artificial intelligence won't wait for a fiscal year of planning. I'm coming back to the subject because this is the time of year when many companies decide the entire following year.

Every plan carries small errors in its assumptions. A small error can wait when time is long. In business, time has shrunk and the risk of waiting has grown.

## The October ritual

Between September and November, companies of every size repeat the same ritual: they sit down to plan the entire following year. Budget, targets, hiring and projects, all decided at once. The last decision in that plan will only be carried out in December of the following year, fifteen months after it was approved.

The format is a little over a hundred years old. In 1922, James O. McKinsey, an accounting professor at the University of Chicago, published Budgetary Control, regarded as the first definitive work on budgeting, at a time when budgeting was a practice of governments, not companies. Four years later, he founded the consulting firm that bears his name. The annual budget spread because it solved a real problem: it gave ever larger companies a way to agree on targets, spending and responsibilities in advance, and to hold each area to what it had promised.

For decades, this worked, because the landscape changed little between approval and execution. The factory, the price of raw materials, the competitor and the customer in December were, with few surprises, the same as in January. Today they are not. An executive who lived through 2026 saw three or four generations of artificial intelligence models released, tried two of them on a phone and switched favourite tools at least once. Even so, they will sit down in October to decide the whole year at once.

The company that wrote in this year's plan that it will study AI in 2027 will start 2028 behind.

## The budget that lasted four months

In December 2025, Uber rolled out AI coding tools to its engineering team. By February, usage had doubled. By April, according to Bloomberg and The Information, the budget planned for the whole year had run out, and the chief technology officer admitted that the planning had to be redone.

Four months.

The problem wasn't that the tool failed. It was that it worked faster than the budget had planned for. Engineers used it, approved of it and started using it more, and the cost followed the usage. No spreadsheet made in October 2025 could have known that.

The shift also happened in the opposite direction. Forrester projects that, in 2026, one in two layoffs attributed to AI will be reversed, and more than half of employers already say they regret the cuts. Ford rehired more than 350 engineers when it found that their knowledge had never been documented and that AI covered only part of the work. The assumption there was that the machine would do those people's work. It stopped holding when it became clear that part of that work was in no manual, system or document. Only they knew how to do it.

None of these companies is badly managed. All of them planned seriously. All of them redid the plan in the first half of the year.

## The plan expired before its deadline

Every plan rests on assumptions: how much it costs, who gets hired, what the technology delivers. Every assumption has an expiry date. In matters tied to technology, that window has shrunk to months, while the plan remains in force for a year.

Some assumptions appear in the plan. Others sit inside the decisions, without anyone writing them down. A three-year software licence contract, signed at today's price and with today's technology, assumes that nothing better or cheaper will appear in that period. A hire approved for September assumes that the work of that position will still be done the same way in September. A system due for delivery in December assumes that the supplier chosen in October is still the best choice.

Uber did the maths with the data that existed. The cost assumption expired in February, and the plan only scheduled a review in December. What stands out is the correction that came afterwards: a monthly spending cap per employee, real-time measurement of usage and continuous review. Without announcing it, Uber stopped operating on annual logic.

This doesn't mean throwing out annual planning. In 1957, already President of the United States, Dwight Eisenhower repeated a line he said he had heard long ago in the Army: plans are worthless, but planning is everything. The explanation came right after: the very definition of an emergency is that it is unexpected, so it is not going to happen the way you are planning. The man who commanded the Normandy landings wasn't dismissing the plan. In the same remarks, he added: when the unexpected comes, the ready-made plan is thrown out, but if you haven't been planning, you can't even start to work intelligently.

Capital, contracts, the fiscal calendar and the board still call for a twelve-month cycle. The change lies in splitting the plan by how long each part stays valid. Capital commitments, such as a factory, real estate, an acquisition or a long contract, stay on the annual cycle. Technology bets, such as replacing a system or building a data team, move to quarterly cycles, with review. AI initiatives, such as a customer service pilot or a new tool for the team, run on a continuous cycle of testing, measuring and adjusting, with a set spending cap and someone responsible for looking at the numbers every week.

Planning for 2027 now includes a task that wasn't on the agenda: deciding how long each decision is valid.

## The bank that abolished the annual budget

The objection is legitimate: without an annual budget, the company loses control. That was the fear Handelsbanken, one of Sweden's largest banks, faced when Jan Wallander, its chief executive from 1970, abolished the annual budget. To him, the budget was unnecessary at best and harmful at worst. In its place, the bank adopted targets relative to the market, continuous monitoring and decision-making autonomy in its branches. Branches came to be compared with one another, and the bank with its competitors, instead of with a number written down the year before.

Control was not lost. According to studies of the case, the bank beat its competitors' return on equity in 46 of the last 47 years and kept up its performance even during global crises.

A bank. The most conservative sector there is. Fifty years ago.

The case didn't stay in Sweden. In 1998, an international consortium, CAM-I, launched the Beyond Budgeting Round Table, a forum that brought together companies such as Royal Mail, Boots and Texas Instruments to study those that had abandoned traditional budgeting. Two consultants linked to the group, Jeremy Hope and Robin Fraser, made Handelsbanken one of the central cases of their 2003 book, Beyond Budgeting. In 2005, the Norwegian oil company Statoil, now Equinor, also abandoned traditional budgeting. In 2010, it went further and decided to "kick out the calendar" from its decisions wherever possible.

I keep wondering how many decisions in the plans being approved right now really needed to wait until October.

## Five questions before approving 2027

Reviewing more often doesn't threaten the target. It protects it. A 2025 survey by Actio indicates that companies that review their plans every quarter are 32% more likely to hit their annual targets than those that adjust once a year.

The target can stay annual. The assumption can't.

The first question is about capital: what in this plan is a real commitment, and what is on the annual cycle only out of habit? A factory, real estate and a long contract call for the annual cycle. A lot of other things come along only because the budget is made once a year.

The second is about assumptions: which ones, in technology matters, expire before the plan itself? The price of a tool, the cost of usage, the expected productivity gain and what the competitor will launch can all change before March.

The third is about reserves: is there money for what doesn't have a name yet? The tool your team will ask for in May hasn't been launched yet. A reserve with no set destination, with clear rules for its use and someone with the authority to release it, avoids redoing the entire budget halfway through the year.

The fourth is about authority: who can revise the plan in March, without waiting for next October? If every change has to go back to the board, the plan only changes once a year.

The fifth is about measurement: is technology usage in your company measured in real time or discovered on the invoice? Uber started measuring usage in real time after the budget ran out. Whoever only looks at the invoice finds the problem a month after it happens.

## Before March

None of this requires predicting what technology will do in 2027. It requires deciding, right now, when each decision will be reviewed, and by whom.

March is closer than next October. How many assumptions in your 2027 planning will already have expired by March?

CEO ONE · https://ceoone.com.br