The ambidextrous budget
The CFO is the same; so is the Monday. At ten, she must unveil the cost programme: remove eight per cent from opex, item by item, without exceptions or sentiment. At two, she faces that identical executive team to make the case for the growth envelope, insisting that now is exactly the time to keep investing in the future.
Is that a contradiction?
The survey evidence suggests she is far from alone. Gartner’s 2026 CFO priorities shows that 56% of CFOs place enterprise-wide cost optimisation in their five leading priorities for the year, while 47% also put capital allocation to new growth opportunities in that same top five. In other words, half the profession is embracing both positions simultaneously.
This is neither hesitation nor a survey artefact. It reflects the job’s newly configured reality. O’Reilly and Tushman gave the label ambidextrous to organisations that exploit today’s model while exploring tomorrow’s, two decades ago, drawing on James March’s earlier distinction between exploitation and exploration. The idea itself is familiar. Far less attention goes to the point at which ambidexterity is truly resolved: not in a strategy presentation, but in the budget. That is where trade-offs cease to be rhetoric and become arithmetic. Most budgeting systems, however, were designed for only one hand.
One budget, two logics
Growth and cost are not opposite settings on a single control. Each follows a distinct discipline, governed by different physics.
Cost work is deterministic. Set the objective, monitor variance, assign named accountability: a miss constitutes a management failure. Monthly reviews, bridge walks and a no-surprises culture all exist because certainty is the deliverable in cost management.
Growth, by contrast, is probabilistic. You are purchasing options on possible outcomes, not outcomes themselves. A competently managed growth portfolio includes bets that fail — indeed, it must, since a portfolio without failures has not taken genuine options. Certainty is not the output here; learning is.
The usual mistake is to force these two logics through one mechanism. Put growth bets through cost machinery and the ROI hurdle behaves as ROI hurdles always do: it eliminates options, since options cannot yet guarantee returns by definition. Bring growth optimism to the cost programme and the budget’s oldest fiction appears: “we’ll grow into the cost base” — something nobody, in my experience, has ever achieved to schedule. A single process cannot contain both, since each logic defines failure differently.
Why averaging fails
When confronted by this tension, people instinctively seek the middle: cut a little in every area and invest a little in every area. The constraint gets spread evenly, like peanut butter, ensuring that every function sacrifices a little and no one loses a debate.
That approach loses both contests. The reductions are too minor to alter the cost base — three per cent across the board changes nobody’s behaviour — while the investments lack the scale to shift the trajectory. Five inadequately funded bets do not form a portfolio; they amount to five protracted deaths. An averaged budget leaves a firm marginally less capable of operating itself, without making it any better at becoming something new.
What works instead has a barbell shape. One end contains deep, unglamorous discipline in the run-the-business budget: actual targets, actual variance and actual consequences. The other holds concentrated, explicitly staged bets within the change-the-business envelope. The soft middle, where most budgets really reside, is the place where both intentions expire. Comfort is the giveaway: when a budget round closes without difficult arguments, averaging has prevailed.
Two hands, two rulebooks
Accommodating both logics inside one company is not about temperament. It depends on four practical mechanisms.
Separate envelopes, separate cadences. As ever, the cost base is established annually and controlled through monthly variance. The growth envelope operates as a portfolio with stage gates, assessed as a fund assesses its book: modest cheques for many ideas, evidence thresholds before the next cheque, quick termination, and greater backing for survivors. There must be two clocks, rather than one. Reviewing a growth bet on the cost timetable means judging it halfway through its experiment, like grading an examination before the paper is complete.
Different definitions of success. Cost performance is measured by variance against plan. Growth must instead be assessed through learning velocity and option value, meaning that ending a bet on the basis of evidence is a win and must be logged accordingly. If a terminated project damages its sponsor’s record, no one will terminate anything, and the portfolio quietly becomes a museum of the walking dead.
A visible recycling link. Finance the growth envelope explicitly from the cost programme: designate a stated portion of every pound saved for recycling into the portfolio, rather than allowing it to disappear into margin. The calculation matters, but the politics matter more, in the best sense. A cost programme that only protects margin produces compliance; one that visibly pays for the future produces volunteers. This connection turns the two hands into one body.
Separate rooms. The same committee should not approve both during a single meeting. A cost meeting’s atmosphere — sceptical, forensic and intolerant of optimism — destroys valid options. A growth meeting’s atmosphere disguises run-rate expenditure as vision. Hold the conversations in sequence, staff them differently, and allow them to contest one another through the CFO instead of within one agenda item.
The referee’s question
This brings us to the central issue: who carries the ambidexterity? It is not the process. Processes do not sustain tension; they eliminate it, generally poorly. The CFO holds it, serving as referee rather than scorekeeper, and deciding for each line item which rulebook governs.
One question performs most of the classification. For every pound in the budget: is this pound purchasing performance, or purchasing options? Performance pounds belong in the deterministic book, governed by targets and variance. Option pounds belong in the portfolio, governed by stages and kill criteria. The question appears simple, but honest application is difficult. Misclassification is widespread in both directions: pet projects hide inside run-the-business areas, beyond cost discipline’s reach, while authentic capability builds are choked out in plain sight, treated as overheads because nobody made the case that they were options.
In 2026, the ideal stress test has arrived, because the most disputed budget line in most organisations now occupies both books at once. AI spending is partly a performance pound — automation with a measurable payback — and partly an option pound: an investment in a different operating model whose value cannot yet be recorded. In When cognition becomes metered, I argued that this spending line acts unlike anything finance has previously budgeted for. Treat every part of it as cost reduction and you will require paybacks that suffocate the option. Treat every part as a strategic option and you will support a thousand pilots that were simply unmanaged costs wearing lanyards. The first genuine test of the ambidextrous budget is already on the desk.
Writing with both hands
A budget is a firm’s theory of its future, expressed in arithmetic. A one-handed budget embeds one conviction: either the future is merely a cheaper version of the present, or the future can be purchased without first enduring the present. On occasion, each belief is correct for a short time. Neither withstands an entire cycle.
The ambidextrous budget states the honest, and uncomfortable, position: with the same pound, amid the same uncertainty and at the same moment, you must finance today’s model and tomorrow’s under two rule sets deliberately at odds with one another. That is not a defect in the modern CFO’s remit. It is the remit.
The firms that navigate what lies ahead will not be those that selected one hand over the other. They will be those that learned to write with both.