AI spending hasn’t moved most top lines because the spend went to tools and the decisions didn’t change. Teams adopted AI one person at a time, outputs got faster, and nobody decided where AI should create value for the business, what work changes because of it, or how to tell whether it worked. Before your 2027 plan, make those decisions explicitly. The tools are the easy part.
What are leaders seeing?
On a networking call this fall, someone who sees dozens of digital health companies a year told me most of them are spending on AI without changing their top line or their workflows. I’ve heard a version of that on several calls since. Everyone uses AI. The outputs don’t connect. Nothing moves.
The research says the same thing at scale. Wharton’s 2026 AI Adoption Report finds that “positive returns are near-universal, though, substantial ones are not.” The split it draws is about decisions, not tools. Among enterprise leaders who:
Track at least one growth measure: 35% report strongly positive ROI, against 13% of those tracking efficiency alone.
Meet all three components of transformational success: 67% report strongly positive ROI, against 19% of those who don’t.
Have standardized on a single approved AI platform: 50% report strongly positive ROI, against 9% of those still experimenting.
Those are Wharton’s figures, from its survey of enterprise leaders. The pattern is clear: the companies seeing real returns decided what AI was for, measured it against growth, and organized around it.
Haven’t we seen this before?
Yes. In the late 1990s I worked as a consultant helping companies rethink their operations, and sometimes their whole business model, around the Internet. “What’s your AI strategy?” in 2026 is the same question as “What’s your Internet strategy?” in 1998.
The lesson from then still holds. Value didn’t come from having a website. It came from understanding customers and the business deeply enough to reinvent how value was delivered, before a competitor did. The companies that treated the Internet as a tool purchase mostly got a brochure online. The ones that went back to people, process and technology got new businesses.
AI’s value won’t come from secret prompts or clever skills either. The disciplines that worked last time — process mapping, root-cause analysis, starting with small tactical projects and expanding — work now. AI lets us run them much faster. It doesn’t let us skip them.
Definition — AI value decision: an explicit choice about where AI changes how the business creates value, what work changes because of it, and how success is measured in business terms rather than activity.
What should you decide before the 2027 plan?
Six decisions, in the order I’d take them. They apply to product builds and to operations work alike.
Where does AI create value for this business? Start from how you create value for customers. Break that into its sources. Then ask where AI could deliver each one better, faster or cheaper. That list, not the tool list, is the plan.
What changes because of it? Name the workflows, handoffs and roles that will work differently. If nothing changes, the spend is a productivity subscription, not a strategy.
What does “done” mean in business terms? Faster output is not done. Define the outcome — a cycle time, a conversion, a cost per case — so review has something to check against. Faster to where? Faster to right?
What must the AI never do? Write the prohibitions: the customer experiences, compliance lines and revenue rules no agent may cross. In healthcare this list is long and specific.
Who decides, and on what platform? Distributed experimentation was the right start. Scaling needs an owner and a standard way of working. Wharton’s platform figures point the same way.
How will you know? Pick growth measures as well as efficiency measures. Efficiency alone is the 13 percent group.
Across all six, check the four dimensions every product and operation depends on: market viability, user experience, operational feasibility and revenue model integrity. A plan that improves one and ignores the others moves work around without moving the business.
What does this look like on a real build?
On the agentic health startup I’m building, we did these decisions before any code. We mapped the family caregiver’s job to be done — tasks, outcomes and the emotional states along the way. Those became guiding constraints for the whole design and build. A business brief anchored the work and gave us a basis to check for drift.
Only then did we gather the knowledge base, run three rounds of spec design, design the surfaces and distill everything into 72 spec documents for agentic coding. It still took four versions of the spec across three releases to get the build running cleanly. But every version was answering the same business questions, so the rework was refinement, not reinvention.
AI made the build fast. The decisions are what made it worth building.
What’s the smallest useful first step?
Pick one workflow where AI use is already happening informally. Write down what it is meant to achieve in business terms, what changes if it works, and what it must never do. Then measure it for a quarter. One explicit, measured project teaches more than a dozen unconnected pilots, and it gives the 2027 plan something real to build on.
How does PBA help?
Product Behavior Architecture (PBA) is an AI-native product operating system that starts at the business model and carries those decisions all the way through to what agents build. It forces each of the four dimensions to an explicit decision before the build, so speed lands on the right thing. Healthcare is where I’m proving it. It works anywhere AI is changing how work gets done.
Takeaway
AI spending moves the top line when it follows decisions, not when it replaces them. Before the 2027 plan, decide where AI creates value, what changes because of it, what “done” means, and what it must never do. Then measure growth, not just speed.
AI builds fast. How are your teams defining and deciding before the build?
Frequently Asked Questions
Why isn’t AI spending improving revenue?
Most spend went to tools while the decisions those tools serve stayed the same. Wharton’s 2026 AI Adoption Report finds that 35 percent of leaders tracking a growth measure report strongly positive ROI, against 13 percent of those tracking efficiency alone.
What should be in a 2027 AI plan?
Six decisions: where AI creates value, what changes because of it, what “done” means in business terms, what AI must never do, who decides and on what platform, and how success will be measured.
Is AI strategy like the 1990s Internet strategy?
In one important way, yes. Value came from rethinking how the business delivers value, not from owning the technology. The same disciplines apply now; AI lets teams run them faster.
If you want to compare notes on your 2027 AI plan, book a discovery call: calendly.com/besserhealth/discovery
Related: Can a product team really run a 24-hour delivery cycle with AI agents? · Business Product Definition Matters More in AI Projects, Not Less. · AI Builds Fast. Fixing What It Built Wrong Takes Longer.
