Rethinking the ROI of GenAI
Measure what matters.
The topic of measuring the ROI of generative AI comes up constantly. Everybody wants to know about hours or dollars saved, throughput percentage increases, and improvements to quality. These are all good and important ways to measure the return on investment in GenAI, but they are far from the only way.
Today’s post is a deep dive into how we think about measuring ROI here at Remix.
📰 What’s Happening in GenAI
Claude Cowork
You may have caught wind of the explosion of excitement around Claude Code while we were on break; watch this space for more on our own experiments with it. 🤯 But the learning curve for Claude Code is as steep as El Capitan. Fortunately, the good folks at Anthropic were paying attention and, using Claude Code, created a far more accessible and intuitive version called Claude Cowork. It’s currently still a “research preview” and only available to Max subscribers. Think of what happens when you run a Deep Research report in a chat model, and now imagine that it can not only look on the internet and generate documents, but can also browse your local computer (with permission) and create all kinds of files beyond research documents. It’s a glimmer of the future that we’re quite excited about and think you should experience for yourself.
A GenAI Price War? Winner: SMBs
Microsoft just introduced “Copilot Business” at $21/user/month, but only for organizations with 300 seats or less. It’s a solid price cut at ~35%, but we still can’t recommend Copilot as your primary or only frontier model given what we hear from our clients 🙁. As of January, Gemini is now included in all Google Workspace plans rather than an add-on. This represents a 56% functional price cut - just weeks after they launched one of the most powerful frontier models in Gemini 3. When they say, “intelligence too cheap to be metered,” maybe we should start to believe them.
“96% of SMBs Plan to Use GenAI”
This is a truly excellent round-up of recent research about SMB adoption of GenAI from the Small Business Administration, the US Chamber of Commerce, Salesforce, and Thryv Survey. Most interesting? 96% of SMBs report an intent to adopt GenAI this year, but only 58% report that they are currently using GenAI (up from 40% the year before). In 2026, GenAI is going to truly become what strategy guru Roger Martin refers to as a “flattening technology.” It will move from being a competitive advantage for early movers to table stakes for everyone. Are you ready?
Reframing the ROI of Generative AI
One of the most common challenges we encounter at Remix Partners is confusion about the ROI of generative AI. When we’re working with SMB leadership teams on their AI strategies, there’s rarely confusion about whether GenAI creates value. Most leaders believe it does. The confusion is about how to measure it, how to justify it internally, and how to think about the investment in the first place.
The pattern usually looks like this: a leader is excited about GenAI’s potential but stuck in a conversation with their CFO or board. The CFO wants payback periods. They’re asking for an “hours saved” analysis. They’re pattern-matching to every other enterprise technology purchase they’ve evaluated.
Here’s the problem: that template was designed for six-figure implementations, eighteen-month deployments, and irreversible commitments. When you’re spending $500K on a system you’ll live with for a decade, rigorous ROI analysis makes sense.
GenAI doesn’t work like that.
A ChatGPT Business subscription costs $30/user/month. Claude Pro is $20. You can cancel them tomorrow. There’s no multi-quarter implementation cycle. Your team can begin testing frontier capabilities immediately..
When the “I” in ROI drops by an order of magnitude, the decision framework should change with it.
Leaders trained on big-bet thinking are asking questions like, “What’s the three-year NPV?” for investments that cost less than the monthly coffee budget. They’re treating reversible experiments like irreversible commitments.
The better frame: treat a 90-day GenAI program as a learning investment with option value. The subscription cost is the premium. The deliverables are information and reduced uncertainty. The goal isn’t to perfectly forecast a three-year NPV. It’s to buy high-quality evidence that improves your next set of larger, less reversible decisions.
Here’s how to think about it.
Three Returns Worth Measuring
Most ROI conversations focus exclusively on operational metrics: hours saved, errors reduced, tasks automated. These matter, but they’re the third most important return. The first two are where the real leverage lives.
Return #1: Transformational Intelligence
Early GenAI initiatives function as paid reconnaissance for your broader transformation roadmap.
Here’s what you should actually have after 90 days of structured experimentation. These are concrete artifacts you can act on:
A Prioritized Use Case Map. Not a generic list from a consulting deck, but 10-15 specific use cases ranked by feasibility and impact in your specific context. You’ll know which ones work because you tried them. You’ll know which ones flopped and why.
Tested Assumptions. Every organization has theories about where AI will help. After 90 days, you’ll have evidence. “We thought AI could handle customer inquiry triage, but our edge cases are too complex,” is valuable intelligence. So is “AI-assisted proposal writing cut our time by 60%, but only after we built a custom jig with our past winning proposals.”
Identified Blockers. The obstacles that would torpedo a bigger investment: data quality issues you didn’t know existed, skill gaps on your team, process dependencies that require upstream changes, compliance or security constraints that limit certain applications. Better to discover these with a $5K experiment than a $500K implementation.
A Change Roadmap. Clarity on which activities should be stopped entirely, which should be redesigned before AI is applied, which are ready for immediate automation, and which require infrastructure you don’t yet have.
This is intelligence you can act on. It’s the difference between making your next major technology investment based on vendor promises versus making it based on evidence from your own operations.
What to measure: Number of use cases tested. Percentage of tested use cases that clear a defined value threshold. Time-to-first-usable-artifact. Blockers identified early (before they derail larger investments). Readiness score by function.
Return #2: Leadership Judgment
The most valuable return from early GenAI investment isn’t operational efficiency. It’s what happens to your judgment as a leader, in two distinct ways.
First, you develop informed judgment about GenAI itself.
Every significant technology decision you make over the next five years will be shaped by whether you understand what GenAI can and cannot do. Vendor selection. Build-versus-buy choices. Hiring priorities. Process redesign. Competitive response. Capital allocation across your entire technology portfolio.
Leaders with hands-on GenAI experience make fundamentally different decisions than those evaluating it from a distance. They can distinguish real capability from vendor noise. They know which employee experiments to fund and which to kill. They spot automation opportunities that GenAI-illiterate leaders miss. They understand realistic timelines and can call out inflated promises.
Second, GenAI amplifies your judgment across everything you do.
GenAI functions as an augmentation of human intellect. It’s not just a tool for automating tasks. It’s a thinking partner that can sharpen your judgment on every aspect of your business.
You now have access to an on-demand advisor, available 24/7, that can challenge assumptions, generate alternatives, and stress-test plans without internal politics. It will tell you what the people on your team may never say because of power dynamics.
Used well, GenAI helps you make better decisions about hiring, pricing, market positioning, competitive response, organizational design, and every domain where clear thinking matters. There’s a double bottom line: judgment about GenAI because you understand the technology, and judgment with GenAI because you now have an intellectual sparring partner in your pocket.
This is something we’re deeply passionate about at Remix Partners: helping leaders understand the unique mindsets and approaches required to use generative AI effectively as a strategic tool. The technology is powerful, but unlocking its value for leadership work requires new mental models that most executives haven’t yet developed.
The math is asymmetric. A few thousand dollars in subscriptions and experimentation buys you informed judgment that will shape decisions worth hundreds of thousands…or even millions. Even if year-one operational savings are modest, GenAI fluency pays back quickly by improving the quality and timing of decisions you’ll be forced to make anyway over the next 24 to 36 months.
What to measure: Reduced vendor evaluation cycle time. Fewer “replatforming” regrets. Higher quality investment memos and requirements documents. Faster pattern recognition on strategic questions.
Return #3: Economics That Finally Pencil
Here’s what the “hours saved” frame misses: it only captures value from tasks you were already doing.
GenAI’s true impact is transforming the cost-prohibitive into the cost-effective.
Let’s be honest: SMBs could do sophisticated competitive intelligence before GenAI. They could hire someone, or pay an agency, or re-assign internal resources. Spending $50K/year on competitive monitoring that might yield $30K in value doesn’t pencil. For far too long, so many things that could drive an SMB forward were simply out of reach as they were too expensive and/or required highly specialized talent.
GenAI changed the math:
Personalized customer communication at scale. Before: hired writers, built sophisticated CRM workflows, spent weeks on setup. Cost: tens of thousands. After: one person, one afternoon, dozens of tailored sequences. The capability was always theoretically possible. It just didn’t pencil for a business your size.
Continuous competitive monitoring. Before: dedicated analyst or expensive agency retainer. Cost: $50K+/year. After: structured prompts, weekly synthesis, fraction of an FTE. Same capability, radically different economics.
Custom internal documentation and training. Before: hired a technical writer or pulled senior people off revenue-generating work. Cost: the opportunity cost plus hardline expense made it easy to deprioritize. After: generate first drafts from existing content, refine with subject matter experts. The ROI calculation completely changed.
The pattern isn’t “things you can do faster.” It’s “things that now make economic sense for the first time.” When the cost of an activity drops by 90%, the entire portfolio of what’s worth doing shifts. And these capabilities compound. Each one you develop makes the next one easier.
What to measure: New activities unlocked that weren’t viable before. Incremental revenue protected or created. Pipeline velocity improved. Churn reduced. Risk reduced through better information.
Asking the Right Questions
Wrong question: “Exactly how many hours will this save, and by when?”
This question assumes you already know which tasks to automate, that those tasks won’t change, and that time savings is the primary value. All three assumptions are usually wrong.
Better questions:
For leadership judgment:
What decisions are we making poorly right now because we don’t understand GenAI capabilities?
How would our technology investment strategy change if our leadership team had hands-on experience?
Where in my role would an on-demand thinking partner create value?
For organizational intelligence:
After 90 days, what specific artifacts will we have to guide our next investments?
What do we need to learn about our own workflows and constraints?
What blockers might exist that we can’t see from the outside?
For changed economics:
What capabilities have we decided weren’t worth the investment at previous cost structures?
What do larger competitors do that’s always been economically out of reach for us?
What becomes rational when the cost of certain activities drops by 90%?
How To Make It Stick
Here’s the objection I hear most often: “We already bought subscriptions to a couple of AI apps and ChatGPT. People used them for a month. The novelty faded. Nothing stuck. Why would this time be different?”
Fair question. And honestly, if you just hand people subscriptions and say, “experiment,” you’ll get the same result.
The difference is structured experimentation versus open-ended exploration.
Open-ended exploration looks like: “Here’s a ChatGPT subscription, go find ways to use it.” This reliably produces a few weeks of enthusiasm followed by abandonment. Without specific goals, accountability, or end points, the urgent always crowds out the experimental.
Structured experimentation looks like:
Identify problems and “if I could only do x” scenarios
Specific use cases to test, not open-ended “find something”
A simple log: what we tried, what happened, what we learned
Someone accountable for driving the experiments forward
A 90-day end point with defined deliverables
Regular check-ins to share learnings across the team
The subscription isn’t the investment. The structure is.
When people say, “we tried AI and it didn’t stick,” what they usually mean is, “we provided tools without structure and hoped something would emerge.” That’s not experimentation. That’s wishful thinking.
The Cost of Waiting
The most expensive decision isn’t picking the wrong tool or investing in a use case that doesn’t pan out. Those are recoverable errors with limited downside.
The expensive decision is waiting.
Every month you delay, you make decisions without understanding what’s possible. You’re evaluating vendors, prioritizing initiatives, and allocating capital based on secondhand information rather than direct experience. Your competitors build compounding advantages. Organizations that started 12 months ago aren’t just 12 months ahead; they’re multiple learning cycles ahead. Your team either experiments without guidance or falls behind industry norms. You miss the stacking effect that early movers are already benefiting from.
The investment is a few thousand dollars and some structured attention. The return is clarity on decisions worth orders of magnitude more.
That’s the ROI that matters. ✨ ✌🏻 ✨
Questions? Send us email - we read every message.