Imagine two engineering teams in the same company. Same size, same roles, same models, same AI budget. A year from now, one of them has cut delivery time nearly in half and retired a backlog of manual work. The other has a very impressive token dashboard. Most enterprise budgeting processes would fund those two teams identically next year, and that is the problem this article is about.
The instinct behind that budget is an old one, and it is not a stupid one. For years, enterprise technology budgeting has been built around access and fairness: every department gets a budget, every employee gets roughly the same tools, and every team receives resources based on headcount, role, or historical spending. That model works reasonably well for software licenses. It breaks down with AI.
It breaks down because AI is productive capacity, and the value of capacity depends on where it is applied, how well the surrounding work is organized, and whether the team can turn faster execution into business results. The companies that win over the next decade will be the ones that invest AI capacity where it produces the greatest return, and that has very little to do with who spends the most.
AI budgets should be managed like an investment portfolio, not an IT expense.
Universal Access Is Table Stakes
This is not an argument for restricting AI access. Every knowledge worker should have an enterprise AI assistant with a reasonable baseline allocation, because AI is becoming part of the standard operating environment, much like email, collaboration software, and office productivity tools.
That baseline matters because it creates organization-wide literacy, encourages experimentation, and allows valuable use cases to emerge from the work itself. The team that finds your best use case next year is probably not the team you would pick today, and you cannot identify the best opportunities if most of the organization is never allowed to explore them.
But universal access is only the starting point. The investment strategy is everything that happens after it.
The Problem With Fair Allocation
Most organizations will be tempted to allocate additional AI capacity by role. Engineering gets one budget, marketing gets another, sales gets a different amount, and finance receives its own pool, each sized by headcount. That looks disciplined because it fits the way companies already budget, and the allocation spreadsheet is easy to defend in a planning meeting. It is also too crude, because job titles do not create returns. Work systems do.
Go back to the two teams from the opening. Both used the same models and spent about the same number of tokens. The first team pointed that capacity at its delivery system: it automated test creation, shortened incident recovery, cleared a backlog of repetitive maintenance work, and cut delivery time by 40 percent. The second team generated more code, more drafts, and more documents without improving throughput, quality, or customer outcomes. Job titles explain none of that gap. What separated the teams was the ability to convert AI capacity into measurable results.
No particular tool or vendor causes this. The fairness model does, with its quiet assumption that spreading capacity evenly is the same thing as investing it well. That assumption is why the team, not the role, should be the primary unit of AI investment.
Why Team-Level ROI Is the Better Measure
Individual AI productivity is difficult to measure well. Once companies try, they usually fall back on weak proxies: token usage, prompts submitted, lines of code produced, documents created, or hours supposedly saved. Those numbers are easy to collect and fit neatly on a dashboard, which is exactly why they get misused; they say nothing about how real work gets done.
A developer may generate code faster, but the team may still be blocked by unclear requirements, slow reviews, fragile deployment pipelines, or a testing process that cannot keep up.1 A product manager may produce specifications more quickly, but that does not matter if the team builds the wrong thing. A support analyst may close more tickets while customer satisfaction declines. Local productivity does not guarantee system-level improvement.
Team outcomes, on the other hand, are easy to connect to the business. Did cycle time improve? Did release frequency increase? Did defect rates decline? Did customer response time improve? Did the team eliminate manual work, bring revenue in sooner, cut operating costs, or take on more valuable work without adding headcount?
Notice that none of those are AI metrics. They are business and delivery metrics, and that is deliberate. Judge AI by the outcomes it helps the team produce rather than by how visibly the team consumes it.
Strong Teams Strengthen Weaker Individuals
A team-based model also avoids a serious mistake: treating AI investment as a reward for the strongest individual performers. That approach looks efficient, but it weakens the organization over time.
The best teams do not simply hand more capacity to their strongest people. They use tools, practices, and shared context to raise the performance of the entire group. A strong engineer using AI can create reusable workflows, better test harnesses, improved documentation, stronger review patterns, and automation that makes everyone on the team more effective. A team that learns to use AI well shrinks the gap between experienced and less experienced contributors without pretending everyone has the same skill level.
This is where team-level investment beats individual allocation. The return is not limited to the person using the most tokens; it appears in the system around them. A productive team uses AI to make knowledge easier to transfer, standards easier to follow, mistakes easier to catch, and routine work easier to automate. That lifts the whole team, including the people who would not have produced the same results alone.
Hunting for a handful of AI superstars misses all of that. Build teams that become more capable together.
Measure Outcomes, Not Consumption
AI budgeting exists to maximize the value created from the capacity purchased, not to minimize token usage, and that requires a different set of questions.
Do not ask which team consumed the most AI. Ask which team reduced delivery time, improved quality, removed a recurring operational burden, increased revenue, reduced cost, improved customer outcomes, or created capacity for higher-value work. Tokens are an input; return is the outcome.
A team that spends twice as much but creates five times the value is not inefficient. It is underfunded. And a team that uses very little AI is not necessarily disciplined; it may simply have failed to identify where AI could improve the work.2 Consumption without outcomes is waste, but low consumption without experimentation is missed opportunity, and the budget conversation has to account for both.
Build an AI Investment Portfolio
The best AI budgets will look less like software expense management and more like portfolio management. Every employee receives baseline access, every team gets room to experiment, and additional investment flows toward the teams that demonstrate an ability to turn AI capacity into measurable business results.
Some investments will fail. Some teams will try ideas that do not work, and some workflows will produce less value than expected. That is not a governance failure; that is how investment works. The real failure is continuing to fund low-value activity because the budget was assigned by headcount, title, or organizational politics.
Capital should move. Teams that produce strong returns should receive more capacity. Teams with promising use cases should receive enough room to prove them. Teams that are not producing value should be expected to change their approach, improve the surrounding system, or release budget for better opportunities. The objective is compounding organizational productivity, and equal spending has nothing to do with it.3
Budget Reviews Should Become Investment Reviews
Traditional budget reviews ask how much was spent and whether the department stayed within plan. AI investment reviews should ask what the organization received in return: team-level changes in revenue, cost, delivery speed, product quality, customer outcomes, operational reliability, manual effort, and capacity created for future work.
This does not require perfect attribution. Most enterprise investments cannot be traced to a single clean number, and the standard should not be mathematical certainty. The standard should be enough evidence to make a better investment decision. Teams should be able to explain what changed, what AI contributed, what constraints remain, and what additional investment is expected to produce. That is a much healthier conversation than arguing over token limits by job title.
The Management Responsibility
AI budgeting is ultimately a management allocation problem. Leaders must decide where additional productive capacity will create the most value, and they have to recognize that AI does not fix a broken operating model by itself. A team with unclear priorities, weak ownership, slow approvals, and a fragile delivery process may simply use AI to create work faster than the organization can absorb it. More tokens do not automatically create more throughput. They create potential throughput, and management still has to remove the constraints that prevent the team from converting that potential into results.
This is why AI budgeting cannot be delegated entirely to IT or procurement. Those groups can manage vendors, security, contracts, and cost controls. They cannot decide where the next unit of AI capacity will produce the best business return. That is a portfolio decision, and it belongs with the leaders responsible for strategy, execution, and results.4
Put Your Tokens Where Your Revenue Is
Think about those two teams one more time. Next year's budget will either fund them identically, because that is what the spreadsheet says is fair, or it will move capacity toward the team that turned tokens into delivery. One of those choices compounds. The other just spends.
The winners will be the organizations that learn to move AI capacity toward the teams that can turn it into revenue, lower costs, better products, stronger operations, and faster execution. Spending the fewest tokens wins nothing, and neither does spreading them most evenly.
Give everyone access. Measure outcomes at the team level. Invest more where the returns are real. AI is not another software expense. It is productive capital.
Put your tokens where your revenue is.
Frequently asked questions
Should companies restrict AI access to control costs?
- No. Every knowledge worker should have baseline access to an enterprise AI assistant, the same way everyone gets email. Universal access builds organization-wide literacy and lets valuable use cases emerge from the work itself; you cannot find the best opportunities if most of the organization is never allowed to explore. Restriction is not the strategy. Baseline access plus targeted additional investment is.
Why allocate AI budget by team instead of by role or department?
- Because job titles do not create returns; work systems do. Two teams with identical roles and identical token budgets can produce completely different outcomes depending on how well they convert AI capacity into results. Role-based allocation looks disciplined because it matches existing budget structures, but it funds titles rather than the systems that actually produce value.
How should AI ROI be measured?
- At the team level, with business and delivery metrics rather than AI metrics: cycle time, release frequency, defect rates, customer response time, manual work eliminated, revenue timing, operating cost, and capacity created for higher-value work. Individual metrics collapse into weak proxies like tokens consumed or lines of code produced, which are easy to collect and easy to misuse.
Is high token consumption a sign of waste?
- Not by itself. A team that spends twice as much while creating five times the value is underfunded, not inefficient. Equally, low consumption is not automatically discipline; it may mean the team never experimented enough to find where AI improves the work. Consumption without outcomes is waste, and low consumption without experimentation is missed opportunity. Judge the return, not the input.
Who should own the AI budget?
- Business leadership, not IT or procurement alone. Those groups manage vendors, security, contracts, and cost controls well, but they cannot decide where the next unit of AI capacity will produce the best business return. That is a portfolio allocation decision, and it belongs with the leaders responsible for strategy, execution, and results.
Does more AI budget automatically make a team faster?
- No. More tokens create potential throughput, not throughput. A team with unclear priorities, weak ownership, slow approvals, or a fragile delivery process will simply use AI to create work faster than the organization can absorb it. Management still has to remove the constraints around the team before additional capacity converts into results.
Footnotes
- AI Speeds Up Execution, Not the System Around It why faster local output stalls when planning, review, and deployment stay slow. ↩
- We're Going to Have to Burn Some Trees the argument for spending tokens seriously instead of rationing them reflexively. ↩
- Cost-Adjusted Software Engineering the same discipline applied to engineering decisions generally, judging work by cost-adjusted return. ↩
- The AI Execution Curve the companion piece on how AI maturity turns spending into a CEO-level growth dial. ↩