The Other ROI: Why “Return on Innovation” Is the Number That Matters in the AI Era

In nearly every AI conversation we’re part of, the same question lands within the first ten minutes: “What’s the ROI?” And between us, that’s a lot of conversations.

We come at this from two sides of the same table. One of us spent a couple of decades leading data, AI, and Technology at Fortune 500 Companies and now sits on boards and advises founders. The other is building an AI company, ASPR AI, in the thick of today's market. The question follows us both everywhere we go.

It’s a fair question. It’s also, more often than not, the wrong place to start.

Not because return doesn’t matter. It matters enormously. But because of what people mean when they say ROI. They mean Return on Investment: a spreadsheet exercise that compares a known cost with projected savings to calculate a payback period. That math is comforting. It’s also built for a world that holds still, and AI is not that world.

So we want to propose a small but consequential reframe. In the age of AI, the “I” in ROI should stand for Innovation, not just Investment. Return on Investment is the floor. Return on Innovation is the building.

Traditional ROI thinking optimizes for the known. You can only model the savings you can already see, so the analysis rewards the safe, legible automation and quietly penalizes the experiment whose payoff you can’t yet name. It treats not knowing as a cost rather than the entire point.

That framing has a blind spot, and it’s a big one: it never prices the risk of standing still. The most expensive line item in most AI business cases is the one nobody writes down: the cost of obsolescence. The cost of letting a competitor learn faster than you. Of arriving at the capability twelve months late because you waited for a certainty that this technology will never give you in advance.

We’ve watched this play out across payments, advertising, search, and now AI-native software. The teams that won were rarely the ones with the cleanest upfront ROI model. They were the ones who ran more experiments, learned faster, and were willing to be wrong cheaply and often.

When you fund innovation rather than just investment, you’re buying three things a payback calculation can’t see.

Optionality, or multiple returns if it works. A cost-savings project returns its savings, and then it’s done. An innovation capability, once it works, pays out in directions you didn’t forecast: a new workflow, a new market you can suddenly serve, an asset competitors don’t have. One bet, many payoffs.

Compounding. Innovation capabilities get better with use in a way that fixed automation does not. The data accumulates, the models sharpen, the team’s fluency grows. The gap between the organizations that started and the ones that didn’t widens every quarter. It doesn’t hold flat.

Insurance against obsolescence. This is the return nobody puts on the slide and everybody feels. The point of experimenting now is not only the upside you capture. It’s the irrelevance you avoid.

None of this is an argument for spending recklessly in the name of “innovation.” The opposite, in fact. We’ve long advised teams to think small and internal first, and to treat the proof of concept as the fundamental unit of progress rather than the enterprise rollout.

The job in this phase is not to be right. It’s to find out: to run a lot of small experiments and explore honestly what’s working and what isn’t. Most of the discipline of innovation is the discipline of killing the things that don’t work quickly and without ego, so the resources flow to the few that do. A portfolio of cheap, fast POCs will out-innovate a single expensive, over-justified program almost every time. You’re not buying a forecast. You’re buying information, and information is what lets you place the next, larger bet with conviction.

Here’s the part that surprises the spreadsheet skeptics. When you fund innovation well, the conventional ROI tends to follow. It just shows up as a consequence, not a precondition. The financial return is the icing, not the cake.

Consider two anonymized examples from real trials run on ASPR AI, the agentic sales assistant one of us advises and the other leads. The companies are unnamed. The figures are theirs.

A real estate investment firm put its full sales team on the platform for a single month. The team didn’t change how it worked. Notes, CRM updates, follow-ups, and coaching simply moved into the background. In 30 days, they got roughly 1,330 hours back, automated hundreds of CRM updates with no rep input, and saw their own scoring of their “must-have” sales workflows climb from 1.9 to 3.6, an 89% lift in workflow maturity. Customer sentiment held positive throughout, which matters: the gains came from removing busywork, not from cutting corners. Annualized, the recovered time alone penciled out to a payback measured in weeks.

That’s a clean Return-on-Investment story. But the second example is the one that actually makes our point.

A semiconductor company with sales and field-engineering teams across APAC ran a smaller, deeper trial. The headline outcomes were strong: workflow maturity rose from 1.9 to 3.3, beating their target, and coaching scores improved at roughly three times the typical benchmark pace. But the most valuable thing that happened wasn’t on the first scorecard at all.

Their customer calls run in Chinese, Korean, and Japanese. The capability to translate these transcripts live into English, and to auto-populate the recap, the follow-up, and the CRM from them, didn’t exist when the trial began. It was built during the trial, for their workflow, because the engagement surfaced the need. A pure Return-on-Investment lens would have scored that trial on the hours it saved. The Return-on-Innovation lens sees what really happened: a durable, differentiated capability now exists that didn’t before, and it exists because the team chose to experiment rather than wait for a finished product. One bet, many payoffs. That’s the whole thesis in a single story.

The deepest returns aren’t efficient at all. Their transformation, a change in how the work works.

When the administrative weight of selling drops away, the role itself changes. Reps spend their hours building trust and having real conversations, rather than wrestling with tools and notes. Coaching stops being an occasional event and becomes continuous and customized. Knowledge that used to live in one expert’s head becomes available to the whole team. ASPR AI was built around a deceptively simple idea: put the human back into selling, and that’s precisely the shift these numbers point to. You don’t get a cheaper version of the old motion. You get a different, better one.

That’s the transformation worth underwriting, and you will never see it in a payback table, because payback tables can only measure the world you already have, not the one you’re trying to build.

If you want to put this into practice, a few principles we’d offer:

The organizations that thrive in this decade won’t be the ones that waited for AI’s ROI to become obvious. By the time it’s obvious, it’s table stakes. They’ll be the ones that treated innovation itself as the return, and discovered, as so many already have, that the investment paid for itself on the way.

Ask “What’s our Return on Innovation?” first. The Return on Investment will be the icing.