The ROI that compounds is growth, not savings.

Every cost saving has the same life cycle. It is announced in a quarterly review, applauded, booked into the baseline, and by the next planning cycle it has vanished into the new normal, never to be claimed again. A dollar saved is a dollar saved exactly once. The current verdict on most business AI activity reflects that ceiling: researchers describe the gains as "marginal rather than game-changing benefits, so far" (Harvard Business Review, June 2026). Marginal is what a one-time claim looks like after the applause fades.

Growth obeys different mathematics. A sustained two-percentage-point lift in organic growth adds roughly 50% to firm value (Harvard Business Review, June 2026), and it does so because each year's gain becomes the base for the next year's gain. The savings curve is flat after the first step. The growth curve never stops climbing. Any ROI argument that treats those two shapes as interchangeable has already lost the comparison, and most business cases written for agentic programs this year treat them as exactly that.

Savings are claimed once. Growth claims itself again next year.

Walk both curves forward five years from the same starting point. The savings program captures its 5% expense reduction in year one, and in years two through five it contributes nothing new, the P&L has simply moved to a slightly lower line, and competitors running the same tools arrive at the same line within a procurement cycle or two. The growth program's two points, by contrast, compound: revenue in year five stands more than 10% above the no-lift path, margin scales on the larger base, and the multiple re-rates because markets pay for trajectory.

That is why the valuation literature finds four sustained points of organic growth more than doubling firm value while even heroic cost programs top out near a 10% lift (Harvard Business Review, June 2026). The asymmetry is not a rounding difference; it is a difference in kind. One curve is a step. The other is an exponent.

What growth-side deployment looks like in practice

Growth-side agentic deployment is concrete, not aspirational, and three patterns recur across Milton engagements. The first is intelligence at deal speed: cross-functional questions answered in about 5 minutes against a pipeline of 8 million records, so the commercial team negotiates with current facts while the counterparty is still scheduling its analyst (documented engagement outcome). Deals move at the speed of the slowest fact-finder in the room. Owning the fastest one changes who sets the tempo.

The second pattern is visibility that wins commitments. One client converted the deeper organizational visibility its named agents produced into a five-year customer commitment, a duration almost unheard of in its category, granted because the customer could finally see, continuously, what it was getting (documented engagement outcome, client anonymized). The third is capacity that scales without proportional headcount: a fleet absorbing the diligence, reporting, and monitoring load of a larger book of business, so revenue grows ahead of the cost line instead of in step with it. Each pattern feeds the next year's version of itself. That is what compounding looks like at the workflow level.

Even the savings serve the growth story

None of this disqualifies operational wins, it reframes what they are for. A Milton engagement delivered a 23% raw-materials inventory reduction measured against the customer's own baseline (documented engagement outcome), and the honest accounting treats that as released capital, not as the headline. Released capital funds the growth side: the intelligence coverage, the paired senior operators, the expansion into the next function. Savings are the fuel. Growth is the destination. Programs fail when the fuel gets mistaken for the trip.

The sequencing matters for credibility too. An operational win measured against the client's own baseline, not a vendor benchmark, earns the trust that growth-side deployment spends. The five-year commitment described above followed exactly that arc: visible operational truth first, then expanding scope, then the long contract (documented engagement outcome). Compounding applies to confidence as much as to revenue.

Targets you can hold someone to

A growth claim invites a fair objection: growth has many parents, and any vendor can take credit for a good year. The discipline is to state outcomes as design targets against the client's own baseline, with consequences attached. Milton's lighthouse stage is designed against a 30–60% function-level improvement target, with a credit policy applied against a defined share of the engagement fee if the target is missed (internal operating record). A design target is not a guarantee, markets move, baselines shift, but a credit policy makes it a commitment with a price, which is the only kind of commitment a CFO should accept.

So the closing arithmetic for the business case is short. Savings: real, bounded near 10% of firm value, claimed once, gone into the baseline (Harvard Business Review, June 2026). Growth: roughly 50% of firm value for two sustained points, compounding annually, and "so far" the part of the opportunity almost nobody has deployed against. The ROI that compounds is the only ROI that can justify the program everyone says they are running. Build the case on growth, bank the savings on the way through, and put a number with consequences on the front page.