For thirty years, the Beyond Budgeting movement has argued that the annual budget conflates three jobs that should be separated: forecasting what will happen, setting targets for people, and allocating resources. The movement's practitioners separated them by organizational effort, rolling forecasts here, relative targets there, dynamic allocation elsewhere, against constant gravitational pull back toward the familiar annual ritual. Now AI arrives promising to do the first job, forecasting, continuously and almost for free. It is tempting to declare victory. It would be premature, and the movement's own research explains why.
Start with what the field studies actually found when companies dropped their budgets. Østergren and Stensaker, embedded in a large energy company that abolished budgeting, documented the central surprise: control did not disappear, it migrated. New mechanisms emerged, relations between headquarters and divisions rearranged, new dependencies formed. Bourmistrov and Kaarbøe, studying two multinationals applying Beyond Budgeting ideas, showed the productive version of that migration: replacing fixed internal targets with relative, outside-in ones (performance against competitors) pushed managers from a comfort zone into a stretch zone, forcing more interaction with the environment, more negotiation, more learning. The lesson from both: the budget was never primarily an information document. It was a social contract about expectations, and abolishing the document renegotiates the contract, one way or another.
What cheap prediction actually changes
Against that backdrop, AI forecasting solves a real but specific problem: the production cost and refresh rate of the information input. Rolling forecasts, the movement's workhorse, were historically expensive, an accounting team's quarterly grind, which is partly why Matějka, Merchant and O'Grady's survey of 201 organizations found even committed adopters struggling to let go of the fixed annual budget for decision-making and evaluation. When the forecast updates itself weekly from live data, that particular excuse dies. Prediction stops being scarce.
But notice what cheap prediction does not touch. It does not decide what a fair target is, that remains a negotiation between people with unequal power and asymmetric information. It does not decide who gets resources when two divisions both show promising forecasts, that remains politics, in the neutral sense: the allocation of scarce means among competing legitimate claims. And it does not absorb accountability: a model cannot own a miss. The Beyond Budgeting insight, sharpened rather than retired by AI, is that these were always three different jobs. AI industrializes exactly one of them.
The hybrid trap, at machine speed
The failure mode is already visible in the accounting literature. Bukh, Ringgaard and Sandalgaard, in a 2024 case study of the dynamic relationship between budgets and forecasts, found that forecasts drift into performing budget-like functions: the boundary between predicting and committing is far more fluid than the movement's rhetoric suggests, and hybrids emerge with their own control tensions. Translate that into the AI era and the trap is specific: the machine forecast quietly becomes the target. The moment a manager knows the model's number will be treated as a commitment, the game restarts, played now against the data pipeline, what gets logged, when deals are entered, how pessimism is fed to the model, rather than against the budget office. Gaming a spreadsheet took a meeting; gaming a model is a continuous, invisible activity. Firms that bolt AI forecasting onto unchanged incentive structures will get the old sandbagging at new speed, with an added layer of false objectivity, because "the model says" carries an authority "the controller says" never had.
Henttu-Aho and Järvinen's field evidence adds the realistic timeline: even pre-AI, abandoning fixed budgets was gradual and institutionally mediated, moving through hybrid transitional practices rather than a clean switch. There is no reason to expect the AI-era transition to be less path-dependent, and every reason to design for the hybrid phase deliberately instead of denying it will happen.
The design that survives contact
The coherent configuration, on the evidence, looks like this. Let the machine own the forecast, and protect its neutrality with a firewall: forecast numbers are never used, raw, as targets or performance benchmarks, or the data feeding them will rot. Set targets relatively, against market, peers, or trailing cohorts, in the Bourmistrov-Kaarbøe spirit, so that ambition is calibrated by the world rather than negotiated against one's own prediction. Allocate resources on cadence and evidence, not on an annual pledge drive. And keep humans visibly accountable for commitments, because a promise made by a model is a promise made by no one.
Beyond Budgeting spent decades arguing that the annual budget survives on the cost of information. AI has now run that experiment at scale: the cost is gone, and the budget's remaining defenses are nakedly institutional. That is the movement's real opportunity, not that machines will finally forecast well, but that there is no longer anywhere for the politics to hide.
Sources
- Østergren & Stensaker (2011). Management control without budgets: a field study of Beyond Budgeting in practice. European Accounting Review. Abolishing budgets migrates control into new mechanisms rather than removing it. https://www.tandfonline.com/doi/abs/10.1080/09638180903487842
- Bourmistrov & Kaarbøe (2013). From comfort to stretch zones. Management Accounting Research. Relative, outside-in targets move managers from comfort to stretch zones, increasing environmental interaction and learning. https://doi.org/10.1016/j.mar.2013.04.001
- Bukh, Ringgaard & Sandalgaard (2024). Moving beyond Beyond Budgeting. European Accounting Review. Forecasts drift into budget-like functions; hybrids and control tensions emerge; warns against idealizing abolition. https://doi.org/10.1080/09638180.2024.2362681
- Matějka, Merchant & O'Grady (2021). An Empirical Investigation of Beyond Budgeting Practices. Journal of Management Accounting Research. 201-org survey: adopters more decentralized with relative targets, yet many struggle to abandon the fixed annual budget. https://doi.org/10.2308/JMAR-19-010
- Henttu-Aho & Järvinen (2013). A Field Study of the Emerging Practice of Beyond Budgeting in Industrial Companies. European Accounting Review. Adoption is gradual and institutionally mediated, passing through hybrid transitional practices. https://doi.org/10.1080/09638180.2012.758596