It’s August. It’s also budget and roadmap season. I’ve already been planning for next year for a couple of months. You can’t make this up.
Like most leaders in large organizations, I’ve spent plenty of time in planning conversations. We build roadmaps, forecast budgets, negotiate headcount, debate priorities, and try to make reasonable decisions about what we’re going to need six, twelve, sometimes eighteen months from now.
There is nothing inherently wrong with doing that. Companies need a strategy. Finance needs some idea of where the money is going. Teams need enough direction to make decisions without renegotiating priorities every week.
The problem is what happens after the plan is approved.
Because between now and next August, a lot is going to change.
Customer expectations will shift. New technology will become viable. Some projects will turn out to be harder than expected. Others will turn out to be unnecessary. New opportunities will emerge that we haven’t even considered yet.
And somehow, despite all of that new information, we’ll still be talking about whether we delivered what we said we would deliver back when we were sitting here in August.
I’ve started to wonder whether annual planning and budgeting, at least the way most large organizations practice them, need to die.
The plan becomes the goal
The part that bothers me isn’t planning itself. It’s how easily the plan becomes more important than the reason we created it.
A team can discover halfway through the year that the original solution isn’t the best way to achieve the outcome. Instead of celebrating the learning, we start asking whether changing direction will affect the roadmap.
Or a new opportunity appears that could create significantly more value than something we’ve already funded. Instead of asking how quickly we can pursue it, the conversation becomes a negotiation over whose budget or headcount can move.
Eventually you end up in a strange place where everyone knows the environment has changed, but the organization continues executing against assumptions made months earlier.
That’s not discipline. It’s inertia.
Budgets create some strange incentives
I’ve seen another version of this play out around budgets.
Leaders know that if they give up money or headcount they aren’t using, there’s a good chance they won’t get it back. So the rational behavior is to protect it.
That creates exactly the behavior you’d expect. Teams hold onto allocations. Leaders ask for a little more than they think they’ll need. Open positions become assets that need to be defended. Money and people get trapped inside organizational boundaries even when another part of the company could put them to better use.
The individual decisions make perfect sense. The system doesn’t.
Imagine an investor looking at a company six months after making an investment and saying, “The assumptions have changed and I wouldn’t make this investment today, but I’ve already allocated the money, so I might as well keep funding it.”
That would never happen!
But inside companies, we often call that sticking with the plan.
Roadmaps aren’t much better
I understand why organizations want detailed annual roadmaps. They make dependencies easier to see, give leaders something concrete to communicate, and create a sense that everyone knows where they’re going.
But there’s a difference between knowing where you’re going and pretending you know every turn you’re going to make along the way.
I’ve been in planning discussions where teams are expected to describe detailed work nearly a year into the future. I always struggle with that. If we genuinely know in August exactly what we’ll be building next July, either we’re extraordinarily good at predicting the future or we aren’t learning very much along the way.
I’m betting on the second one.
A roadmap should communicate direction and intent. The farther out it goes, the less precise I think it should become.
AI is making this harder to ignore
This problem existed before generative AI, but AI has made it much more visible.
Think about how many assumptions about engineering have changed in just the last couple of years. The way developers write code is changing. The cost of prototyping is changing. Build-versus-buy decisions are changing. The capabilities we need on teams are changing. Even assumptions about how many people it takes to accomplish certain kinds of work are starting to change.
A strategy created twelve months ago may still be perfectly sound. The assumptions underneath the execution plan may not be. That’s the distinction I think we’re missing.
AI doesn’t just make some work faster. It shortens the useful life of some of our assumptions. If the environment can materially change every few months, an operating model that requires us to predict detailed work twelve months in advance is going to create more and more friction.
I don’t think the answer is less planning
Whenever I make this argument, the obvious counterpoint is that large organizations can’t just continuously change direction. That’s true.
Hundreds or thousands of people can’t operate effectively if priorities change every Tuesday. Finance can’t run a company without forecasts. Long-term platform investments can’t be evaluated on a six-week horizon. Some decisions genuinely require multi-year commitments.
So I don’t think the answer is to stop planning.
I think we need to stop confusing planning with commitment.
Strategy can be durable without every deliverable being durable. We can be clear about the outcomes we want without pretending we know exactly how we’ll achieve them.
The shift I’m arguing for looks something like this:
| Instead of… | Move toward… |
|---|---|
| Funding specific features | Funding strategic outcomes and capabilities |
| Detailed 12-month roadmaps | Near-term detail with decreasing precision over time |
| 100% allocated capacity | Deliberately reserved capacity for emerging work |
| Measuring delivery against the original plan | Measuring progress against the desired outcome |
| Protecting budget and headcount | Moving resources toward the highest-value problems |
| Treating stopped work as failure | Treating stopped low-value work as recovered capacity |
| Annual prioritization | Regular opportunities to reconsider investments |
None of those changes eliminate planning. They make the plan easier to change when reality gives us better information.
Leave room to learn
I also think organizations need to get more comfortable deliberately leaving some capacity uncommitted. That can sound irresponsible during annual planning. If I have 100 engineers, shouldn’t I be able to explain exactly what all 100 will be doing?
I’m increasingly convinced the answer is no.
If I’ve committed 100% of my capacity twelve months in advance, I’ve made an implicit assumption that nothing important will happen during those twelve months that I didn’t predict. That seems like a much bigger risk.
Maybe 70% of capacity goes toward high-confidence strategic commitments, 20% toward emerging opportunities and experiments, and 10% stays available for things we simply couldn’t have anticipated.
The percentages aren’t really the point. The principle is. Unused optionality has value.
We should get much better at stopping
There’s one question I wish we asked more often during portfolio reviews: Knowing what we know today, would we still start this work? It’s a different question than asking whether something is on track.
Imagine we’re six months into an initiative. We’ve spent millions of dollars. Several teams are working on it. It’s on an executive roadmap. But the assumptions that justified the investment have changed.
Would we fund it today?
If the answer is no, the fact that we’ve already spent money on it shouldn’t automatically justify spending more. That’s sunk-cost thinking at organizational scale.
Stopping work isn’t necessarily failure. Sometimes it’s evidence that the organization learned something and acted on it. I’d like to see us reward that much more often.
The operating model has to change
For a long time, organizations could create a plan, fund it, execute against it, and spend the year measuring variance. I think the next operating model looks different.
We still establish direction. We still allocate resources. But then we continuously pay attention to what’s happening around us. We learn. We revisit assumptions. And when the evidence changes enough, we move resources.
The goal isn’t constant reprioritization. That would be exhausting and would make it nearly impossible for teams to execute.
The goal is making reprioritization possible. There’s an important difference.
The companies that get good at this won’t necessarily be the ones that predict the next twelve months better than everyone else. They’ll be the ones that recognize sooner when their original prediction is wrong and can actually do something about it.
That’s where I think the competitive advantage is moving: better sensing, better decisions, and faster reallocation.
And maybe next August, instead of asking whether we delivered everything we predicted a year earlier, we should be asking a much more interesting question:
Given everything we learned this year, did we keep investing in the things that mattered most?
I’d much rather be accountable for that.
