For three years, AI lived in the innovation budget. It was a portfolio of pilots, licenses and experiments, small enough to escape the scrutiny applied to a new plant or an acquisition. That era ends in 2027.
AI is becoming a system, not a set of tools. It now touches pricing, service, product, supply chain and the way customers discover brands. When a technology rewires how the whole business creates value, it stops being an IT line item and becomes a capital allocation decision. And capital allocation is the CFO’s language.
This article makes a simple argument. In 2027, AI ROI will be judged the way every other major investment is judged: against a hurdle rate, over a payback period, by its contribution to return on invested capital. The companies that win will be the ones where the CFO and the CMO, and every leader spending on artificial intelligence, already speak that language together.
Key Takeaways
- AI spending is moving from experimental budgets into core capital allocation, and 2027 plans are being written now.
- Most AI ROI calculations fail because they count cost savings and ignore revenue growth, customer lifetime value and strategic option value.
- Finance leaders and marketing leaders need one shared vocabulary: hurdle rates, ROIC, payback and economic profit.
- The highest returns sit at the top of the AI Adoption Pyramid, where AI reinvents how value is created, not where it trims manual process cost.
- A true partnership between the CFO and the CMO is no longer a cultural nice-to-have. It is the operating model for scaling AI.
Why 2027 Is the Year AI Becomes a Capital Allocation Question
The scale of the investment has changed the conversation. Gartner forecasts worldwide AI spending of roughly $2.5 trillion in 2026. At that level, AI is no longer a rounding error in a technology budget. It competes directly with factories, acquisitions, share buybacks and brand investment for the same capital.
The returns have not kept pace. MIT’s NANDA initiative reported in 2025 that around 95 percent of organizations saw no measurable return from their generative AI pilots. Boards have noticed. The question in the boardroom has shifted from “are we doing enough with AI?” to “what are we getting for it?”
From AI Tools to One System
The first wave of AI adoption was about tools. A copywriting assistant for marketing. A chatbot for service. Automated journal entries and a faster close process for finance teams. Each tool had a small budget, a local owner and a local business case.
The next wave is about one system. AI agents now connect customer data, pricing engines, content, service and supply chain. A decision made by an agent in one function changes the economics of another. When AI becomes infrastructure, its costs are shared, its benefits are shared, and its risks are shared. No single function can claim the return or own the cost. That is precisely the kind of investment a CFO is built to govern.
The 2027 Budget Is Being Written Now
Most large companies set next year’s capital plan in the final quarter of this one. The AI line in the 2027 plan will not be approved on enthusiasm. It will be approved on a business case tied to business goals, and it will be measured against it a year later. Leaders who arrive with a productivity anecdote instead of an investment strategy will see their budgets cut or reassigned.
The data points in one direction. In OneStream’s research, roughly three quarters of CFOs now say they lead enterprise AI strategy, yet only about one in three have deployed AI at scale. Bain found that more than four in ten CFOs plan to raise AI investment by over 30 percent within two years. Finance is taking the wheel. The question is whether everyone else in the room can read the map.
Why Most AI ROI Calculations Fail
Many CFOs are frustrated with AI business cases for a good reason. Most of them measure the wrong thing.
Cost Savings Are the Wrong Denominator
The default AI ROI calculation counts hours saved. Two or three processes get automated, a large team gets smaller, and the savings appear in the business case. This is the easiest return to measure and the least valuable one to pursue. Cost savings are finite. Every competitor with the right tool will capture the same savings within months, so the advantage disappears and the margin flows to customers through lower prices.
Klarna illustrates the trap. In 2024 the company announced that its AI assistant was handling two thirds of customer service chats, the work of roughly 700 agents. In 2025 its chief executive said the company would bring more humans back into customer support, because quality and customer experience mattered more than the cost line. The savings were real. The business value was incomplete.
Cost savings are the return everyone can copy. Revenue growth is the return only you can earn.
Pilots Measured Like Experiments, Funded Like Line Items
The second failure is structural. AI pilots are funded as operating expense, measured by activity metrics such as usage and adoption, and owned by a single function. None of that maps to how a CFO evaluates investment. A pilot that reports “80 percent of the team used the tool” says nothing about conversion rates, margin, retention or capital efficiency.
Without a shared language, the conversation shifts to defense. Marketing argues reach and engagement. Finance asks for payback. Technology asks for more data. Everyone leaves the meeting unconvinced, and the budget defaults to whoever told the clearest financial story.
The AI Adoption Pyramid, Read by a CFO
Vivaldi’s AI Adoption Pyramid describes four levels of AI maturity: productivity, transactions, interactions and reinvention. Only around one in ten companies reach the top. Read through a finance lens, each level carries a different return profile, a different payback period and a different way to measure ROI.
| Level | What AI does | Primary return | How to measure ROI | Typical payback |
|---|---|---|---|---|
| Productivity | Automates tasks and manual process work | Cost savings | Hours and cost per unit of work | Months |
| Transactions | Improves pricing, conversion and service resolution | Margin and conversion rates | Gross margin, conversion, cost to serve | One to two years |
| Interactions | Personalizes the relationship with each customer | Revenue growth and retention | Customer lifetime value, churn, share of wallet | Two to three years |
| Reinvention | Creates new business models and value pools | Economic profit | ROIC, new revenue streams, option value | Three years and longer |
The table explains why so many AI investment cases look weak. Companies apply productivity metrics to interaction and reinvention bets. They expect a two-quarter payback from an investment designed to change customer lifetime value over three years. Then they cancel it before the return arrives.
Two Languages, One P&L: Why CMO CFO Alignment Breaks
The CMO and the CFO have historically spoken different dialects. Marketing reports awareness, engagement, leads and brand health. Finance reports margin, cash, ROIC and earnings per share. Each side considers its own metrics the real ones.
AI makes this gap expensive. The largest AI returns sit in the customer: personalization, pricing, service, discovery and loyalty. Those are marketing territories. But the investment required is large, shared and long term, which makes it a finance decision. If marketing cannot express AI value in capital allocation terms, finance will fund the cost-saving use cases it can measure and starve the growth use cases it cannot. The challenge is not the math. It is collaboration.
We have argued before that marketing should raise ROIC, not chase ROI, and that brand spend is argued in marketing language but judged in financial language. AI turns that observation into an urgent operating problem. The conversation is no longer about the marketing budget. It is about how the entire company allocates capital to the system that will run it.
What a Shared Capital Allocation Language Looks Like
Getting on the same page does not mean marketing becomes finance. It means both sides agree on a small set of terms and use them every time AI investment is discussed.
Hurdle Rates, Not Hype
Every AI investment should clear the same hurdle rate as any other capital project. That forces a simple discipline: state the expected return, the time horizon, the risk and your confidence in each. A use case that cannot be expressed this way is not ready for funding. A use case that can be expressed this way earns a fair hearing, whether it sits in finance, marketing or operations.
ROIC Over ROI
ROI on a single project is easy to inflate. Return on invested capital asks a harder question: does the AI investment make the whole business more capital efficient? An AI system that raises conversion rates while requiring less working capital, fewer discounts and lower acquisition cost improves ROIC. A system that saves labor but requires a large data platform investment and constant retraining may not.
Customer Lifetime Value as the Bridge Metric
Customer lifetime value is the metric both functions already trust. Finance recognizes it as discounted future cash flow. Marketing recognizes it as loyalty and relationship strength. AI investments in personalization, service and discovery should be measured by their effect on lifetime value and acquisition cost. That single ratio translates brand and experience into the language of capital.
Option Value for Reinvention Bets
Some AI investments will not pay back inside a standard planning cycle. They create the right to compete in a new market, a new channel or a new business model. Finance has a tool to support these decisions: option value. Treat reinvention bets as a portfolio of options with staged funding, clear kill criteria and defined milestones, rather than forcing them into a three-year payback model they were never designed to meet.
How to Measure AI ROI Across Cost, Growth and Strategy
A credible AI business case covers three layers of business outcomes. Most cover only the first.
| Layer | Examples | Metrics finance leaders accept |
|---|---|---|
| Efficiency | Automated journal entries, faster close process, service automation | Cost per transaction, cycle time, FTE capacity redeployed |
| Growth | Personalized offers, dynamic pricing, AI-assisted sales | Revenue growth, conversion rates, average order value, retention |
| Strategic | New services, AI-mediated discovery, data partnerships | Share of new revenue, customer lifetime value, ROIC, option value |
To measure ROI credibly, set the baseline before deployment, define a control group where possible, and agree in advance which metric will decide success. Ensure data definitions match across finance and marketing, so that “customer,” “revenue” and “conversion” mean the same thing in every report. Most AI ROI disputes are really data disputes.
Five Moves for a True Partnership Between Finance and Marketing
1. Build One AI Investment Portfolio
Replace scattered departmental AI budgets with one portfolio, reviewed jointly by the CFO, the CMO and the technology leader. Classify each initiative by its level on the AI Adoption Pyramid, so that productivity bets and reinvention bets are never judged by the same yardstick.
Key question: can you see every deployed AI initiative, its owner and its expected return on one page?
2. Agree on a Shared Value Dictionary
Write down the six to eight metrics that will govern AI investment: hurdle rate, payback, ROIC, customer lifetime value, acquisition cost, margin, and one brand health measure. Use them in every business case. Retire vanity metrics such as usage and activity from investment discussions.
Key question: would your CFO and your CMO calculate the return on the same AI use case and get the same answer?
3. Fund in Stages, Measure at Each Gate
Scaling AI works best with staged funding. Release capital for a defined phase, measure against the agreed metric, then release more. This protects the business from runaway spend and protects promising initiatives from premature cancellation.
Key question: what evidence would convince you to double this investment, and what evidence would make you stop?
4. Put Growth Use Cases First
Cost savings fund the AI journey, but growth justifies it. Pair every efficiency initiative with a growth initiative in the same customer journey, so that capacity freed by automation is redeployed into revenue growth rather than simply removed.
Key question: where does the time or cost your AI saves go next, and what does it earn?
5. Make the Brand Part of the Business Case
AI changes how customers discover and choose brands. Assistants and agents increasingly shortlist options before a human compares them. A brand that AI systems recommend by default enjoys lower acquisition cost and higher lifetime value. That is a financial asset, and it belongs in the business case alongside the technology. Our work on brand performance metrics shows how to connect brand health to the numbers a CFO uses.
Key question: how would your AI investment case change if it counted the value of being the brand AI recommends?
What This Looks Like in Practice
The companies furthest ahead already treat AI as a capital allocation problem. They review AI investment alongside other strategic bets, not inside the IT budget. Their finance leaders sit in product and marketing reviews. Their marketing leaders prove value in the language of margin, lifetime value and ROIC.
Duolingo’s 2025 “AI-first” memo is a useful contrast. The strategy was sound on productivity, but the public reaction showed that customer and brand consequences belong in the same calculation as cost. An AI investment case that ignores brand risk is incomplete, in the same way that a brand case that ignores capital efficiency is incomplete.
The lesson is not that AI should be slowed down. It is that AI should be governed like what it has become: the largest and most strategic investment program most companies will run this decade. Implementing AI at that scale without a shared financial language is like building a factory without agreeing on what it will produce.
The Conversation to Have Before the 2027 Plan Is Locked
New technology always arrives as an experiment and matures into infrastructure. AI is crossing that line now. When it does, the companies that win will not be the ones with the most AI tools. They will be the ones whose leaders can explain, in one shared language, why each dollar of AI investment earns more than it costs.
That conversation belongs in the next leadership meeting, not the next budget cycle. If you want a starting point, our view on why capital allocation does not create value, future demand does frames the question every AI business case should answer.
In 2027, AI will not be judged by what it can do. It will be judged by what it earns.