Will AI Make the Annual Planning Process Obsolete?

By Victor Hoang · Updated Oct 10, 2026

Every fall, the same ritual starts. Finance sends a template, functions build decks, and for six to ten weeks senior people spend most of their time producing documents about next year. Now a CFO or COO looks at what AI tools do with a spreadsheet and a pile of reports and asks a fair question: if a model can build the baseline, run the scenarios and draft the narrative in a few days, why are we still doing this the slow way?

Our answer is that AI will make the annual planning process obsolete, but not planning itself. It compresses the analysis half of planning, which was always the part that took the calendar. It does nothing for the other half: deciding what you will stop, where you will concentrate, and who owns the result. Companies that keep the 10-week cycle will waste the speed AI gives them. Companies that scrap planning will lose the commitments that make an organization move together. The better design is a lighter, rolling cycle anchored on a few hard choices.

Key Facts

• In a survey of 156 senior executives at companies with at least $1 billion in sales (reported Dec. 2005), firms running traditional annual planning averaged 2.5 major business decisions a year, versus 6.1 for firms that set strategy continuously and by issue rather than by business unit.
• In that same survey, 66 percent of companies planned periodically, usually annually, yet 100 percent said they made strategic decisions continuously.
• The Beyond Budgeting principles call for organizing management processes around business rhythms and events, not the calendar year only, and for resources made available as needed rather than through detailed annual budget allocations.
• In a pre-registered experiment with 758 BCG consultants, AI users finished 12.2% more tasks, 25.1% faster, with quality rated over 40% higher, on tasks inside AI's capability frontier. On a task outside it, they were 19 percentage points less likely to reach a correct answer.
• On a forecasting benchmark, expert human forecasters still outperformed the top-performing language model, according to the ForecastBench authors.

What the annual plan was really doing

Strip the ceremony away and an annual plan does five jobs. It gathers information, forecasts the year, allocates money and people, sets targets, and gets leaders to agree publicly on a direction. Only the first two are analytical. The last three are acts of commitment.

The calendar built up around the plan was mostly an artifact of the analytical work being slow. Collecting data from eight functions, reconciling it, building three scenarios and turning them into a coherent story took weeks of analyst time. That's why the process ran once a year. Nobody wanted to pay that cost twice.

AI attacks exactly that cost. Pulling data together, building a baseline, stress-testing assumptions and drafting the first version of the narrative are all things current tools do quickly, and it's reasonable to expect a planning team to compress that stage from weeks to days. We'd treat that as an informed expectation rather than a measured number, because the gain varies with how clean your data is. The point is the direction: the reason for the annual rhythm is weakening.

The cycle was already breaking before AI

Here's the awkward part. The annual cycle didn't match how executives decide even when analysis was slow. The 156-executive survey above, reported by CFO.com, found two-thirds of companies planning on a periodic cycle while all of them said they make strategic decisions continuously. Mankins and Steele's Harvard Business Review article opens with a CEO of a global manufacturer asking whether strategic planning is "completely useless," after reviews that produced very few worthwhile decisions.

Beyond Budgeting practitioners have made a parallel argument about money for years. Their published principles say to make forecasting lean and unbiased rather than rigid and political, to set directional and relative goals instead of fixed cascaded ones, and to organize around business rhythms and events. A CFO magazine piece from 2014 shows the practical split: one Beyond Budgeting voice wanted budgets replaced, while a consultant argued most companies would never give theirs up and should keep it as a starting point with a continuously refreshed forecast layered on top.

So AI isn't killing a healthy process. It's removing the last excuse for a process that already underperformed its purpose.

What AI compresses and what it doesn't

The useful way to redesign planning is to sort each activity by whether it is analysis or commitment. Analysis gets faster with AI. Commitment doesn't, because it's a human act with human consequences.

Read the bottom half of that table as the core of planning. Those are the rows that determine whether next year goes well, and AI doesn't touch them. If anything, cheap analysis makes them harder, because it removes the cover of "we're still waiting on the numbers."

Why a fast model won't give you a better plan on its own

Three limits matter for executives.

Reliability varies and is hard to see. The BCG study's lesson is a jagged one. Inside the frontier of what the model handles well, users were faster and better. On a task deliberately placed outside it, AI users were 19 percentage points less likely to be correct. A plan is full of tasks where you can't tell in advance which side of the frontier you're on. Forecasting is the clearest case: in the ForecastBench work, the authors report that expert forecasters outperform the top language model. A model-built baseline is a draft to challenge, not a number to adopt. This is the same overhead we describe in the verification tax on AI time savings: the hours saved producing the analysis partly move into checking it.

Fast analysis tempts you to plan more, not decide more. When a scenario costs nothing, teams generate twenty of them. More options without a forcing function produce paralysis. The constraint that matters isn't compute, it's the number of real priorities an organization can carry out at once. Our piece on coordination cost covers why adding parallel initiatives usually costs more than it returns.

The ground moves under a 12-month plan. If your plan assumes a specific model, vendor or workflow, the assumption can be stale before the plan is approved. We wrote about designing for this in AI strategy that survives model churn. A plan that bets on durable capabilities, not named tools, survives; one that locks a year of spending to a tool choice doesn't.

The design: a rolling cycle anchored on a few hard choices

If analysis is cheap and commitment is scarce, the planning calendar should spend its time on commitment and refresh the analysis often. Here's the structure we'd recommend, stated as a working design rather than a proven standard.

1. One annual strategy conversation, short and sharp. Once a year, the leadership team makes a small set of hard choices: three to five bets, plus an explicit stop list. This replaces the 10-week deck cycle. It should take days, with AI preparing the pre-reads. The output is a page, not a binder. If you use OKRs, this page is where the top-level objectives come from, and it is the quickest test of whether your AI plans are a strategy or a shopping list.

2. A rolling forecast with a fixed horizon. Instead of a calendar-year forecast that decays by Q3, maintain a forecast that always looks a fixed distance ahead. The CFO magazine article describes consultants suggesting at least four quarters and sometimes 12 to 18 months, and one asset manager running a six-quarter horizon. AI does the refresh. People review the variances that matter.

3. A quarterly decision review, not a quarterly plan review. Take the lesson of Mankins and Steele's article title seriously: the output of a planning meeting should be decisions. Each quarter, the team asks what has changed, which of the bets need more or less, and which stop-list items are actually stopped. Anything that needs a call between reviews gets an event-driven decision, which is the Beyond Budgeting idea of working to business rhythms and events.

4. Resources released against bets, not spread across the year. Fund the few priorities with the ability to move money as evidence arrives. Keep a small reserve for decisions you haven't made yet. This is where the old budget becomes a flexible envelope.

5. A named owner and an AI role for each bet. Every bet has one accountable person. Decide upfront where AI assists (analysis, monitoring, drafting) and where a human signs off, using the same logic as the copilot versus agent distinction and your governance rules for decisions that carry real stakes.

A worked example (illustrative)

Take a hypothetical 400-person B2B services firm. Under its old process, planning runs from September to December: each of six functions builds a deck, finance consolidates, and the executive team holds three review sessions. By February the targets are set; by June half the assumptions are outdated.

Under the redesigned cycle, the firm spends one two-day offsite in October choosing four bets and two things to stop, with AI-prepared pre-reads on market, pipeline and capacity. Finance keeps a six-quarter rolling forecast that AI refreshes monthly. Each quarter, the leadership team holds a half-day decision review. In this example the two stops are a low-margin service line and a regional expansion; naming them is uncomfortable, and that discomfort is the point. The firm didn't save much time on the hard part. It saved weeks on the easy part and spent the time on the decisions.

How to tell whether your plan is the problem

Three questions give you a fast read.

• Count the real decisions. How many choices in last year's plan changed what a team actually did? If the answer is closer to two than ten, the process is producing paperwork.
• Check the stop list. Does your plan name anything that gets defunded? Plans that only add are wish lists.
• Look at the midyear rewrite. If you reforecast heavily in Q2 and Q3, the annual forecast is already a rolling one, without the discipline.

What to do next

You don't need to scrap anything this fall. Run the next cycle as a pilot.

1. Cut the preparation phase in half. Use AI to build the baseline and scenarios, and require a named reviewer to challenge the assumptions before they enter the plan. Plan for the checking time rather than pretending it vanishes.
2. Force the stop list. Before approving any new initiative, require the team to name what stops or slows.
3. Limit the bets. Pick three to five priorities. Everything else is run-the-business work.
4. Stand up a rolling forecast alongside the budget. Keep the annual budget for the board if you need it, as the American Century example in the CFO piece does, and run the rolling forecast for steering.
5. Schedule the quarterly decision review now. Put the dates on calendars for the whole year and define the decisions it is expected to produce.
6. Track the right outcome. Measure whether the plan changes decisions and results, using the discipline in measuring AI ROI beyond time saved. Hours saved on slide production are not the return.

For the mechanics, the library entry on the strategic planning process is a good starting reference.

The bottom line

AI makes the weeks of analysis behind an annual plan hard to justify. It leaves the part that matters untouched: choosing, committing and being accountable. The leaders who benefit won't be those who generate the most polished plan fastest. They'll be the ones who use the freed calendar to make fewer, harder choices and revisit them often.

Learn More

• AI Strategy That Survives Model Churn
• Measuring AI ROI Beyond 'Time Saved'
• The Coordination Cost of Running Operations With AI
• The Governance Gap: What Leaders Get Wrong About AI at Work
• AI Copilots vs. AI Agents
• Strategic Planning Process