From Billable Hours to Seven-Figure ARR: Learnings From Solving a $2M Problem in 72 Hours
Three weeks ago, I got a call from the head of a 500-employee logistics company. They were about to lose $200,000 per month due to a contract dispute over something mundane: measuring cable diameters across 5,000 pallets.
We built an AI solution in three days. Monthly savings for them: $180,000.
Instead of billing for three days of work, we structured it to generate almost seven figures in ARR. That’s not a humblebrag. It’s a warning.
Building custom AI solutions is easier than most CEOs think. Which means it’s easier for your competitors too.
NEW BUSINESS MODELS REQUIRE NEW CUSTOMER PROFILING
After selling Integral, I made a shift. We went from serving massive enterprises (100,000+ employees, $100B+ revenue) to working with mid-market companies. Those Fortune 500 logos looked great on the website, felt great for my ego, and justified premium hourly rates.
Our incentives were, however, misaligned with the value we created. We were exceptionally good at driving business outcomes efficiently. We didn’t just hire strong software engineers. We invested heavily in lean product management and human-centered design, capabilities that cost more but let us focus on outcomes, not just shipping software.
In several cases, we delivered solutions that created 20-30X return on our fees. Those were the products we loved most. The urgency was real. The curiosity was shared.
Ironically, we often made our highest margins doing work we suspected would fail from day one.
The reason was structural. In large enterprises, our buyers were usually mid-level executives. They had budgets but not true authority. We were paid to execute, not to question. We wrote beautifully crafted code, quickly and with care, for ideas that should never have survived first contact with reality.
At smaller companies, We’re talking to the CEO or owner. They have full context and full accountability. They don’t ask about hourly rates. They ask about ROI and how fast we can move.
Now, I only work directly with CEOs and owners, even if that means smaller companies and smaller budgets. I want to be paid for outcomes and ROI, not time. I prioritize leverage over headcount.
More importantly, I want to share risk with my customers. In the right situations, I’ll even fund the development cost entirely. The criteria are straightforward: the problem needs to be meaningful, AI needs to be a feasible solution, the customer needs to be motivated and willing to share a high percentage of the value, and attribution needs to stay transparent and simple enough.
I used to prioritize revenue first, employee count second, and profit third, but now prioritize leverage first, profit second, and revenue third. Employee size is irrelevant, and I’d rather partner with specialized boutiques like yours when I need to call in the cavalry.
THE CABLE EXAMPLE: WHAT OUTCOME-BASED WORK LOOKS LIKE
That cable measurement problem was well suited for computer vision. We built an initial AI agent in three days and trained a custom vision model using Roboflow.
Productionizing that prototype for long-term use cost us $50,000 in development over 3 weeks, with an estimated operating cost under $5,000 per month. The result is savings of more than $180,000 per month for the customer. They pay us 40% of total savings that could result in over $800,000 in ARR.
At first glance, that margin may seem suspiciously high for a bespoke solution. The customer didn’t care. They were saving up to $2 million per year, had no feasible alternative, and desperately needed a solution before year-end. Furthermore, with the customer friction involved, the opportunity cost could be way more.
When pricing is anchored to value rather than effort, margin becomes irrelevant to the buyer.
What happened next mattered even more. Since we delivered value quickly and established a tight feedback loop, trust built faster than anything I’ve seen in enterprise environments. They’ve now asked us to visit on-site to identify other seven-figure savings opportunities. They’re also exploring how AI could help them rethink long-held assumptions about their business model, not just their operations. Especially impressive for a blue collar workforce.
THE REAL RISKS, AND WHY THEY’RE WORTH TAKING
This model is not without risk.
Our contract is downstream of our customer’s contract with their customer. We’ve intentionally discounted revenue projections beyond six months, anticipating renegotiation. Given that our costs are recovered within the first month, we retain ownership of the intellectual property, and have been invited to actively looking for adjacent problems to solve, this risk becomes a rational tradeoff rather than a gamble.
I also assume many AI-enabled solutions will be quick to market and potentially short in lifespan. OpenAI CEO Sam Altman calls this the fast fashion era of apps. Companies like H&M figured out how to optimize for speed of learning and market responsiveness, not perfection or permanence. We need to do the same. Build systems that sense demand in near real time, iterate quickly, scale what works, and discard what doesn’t.
Established companies need to wrestle with that reality rather than design for a world that no longer exists.
WHAT THIS MEANS FOR YOU
If you run a professional services firm, especially an established one, here are the patterns that matter:
1.Understand why your firm exists. In manufacturing, the “5 Whys” are used for root cause analysis of problems. The same technique works for understanding the value you provide the market. What “job” are your customers actually hiring you for? Once you answer that honestly, re-imagine how that job could be done if AI were available from first principles, not layered on top of your existing process.
2. Do not delegate AI to your team or IT. Installing Copilot is not AI transformation. Neither is sending your team to a training. If you want to understand what is possible, rebuild your business mentally from the ground up, back to when you had to do everything yourself. Block one day a week to actively try to disrupt your own company.
3. Iterate on your business model. Most firms inherit pricing assumptions from their industry. If you’re established, profitable, and supported by peers from Collective54, you’re in a privileged position to experiment. Hourly billing made sense in a world constrained by human throughput. AI breaks that assumption.I especially love the work of Ron Baker on this topic.
4. Accelerate your own learning. Speed of learning is now a competitive advantage. Modern research tools make it possible to deeply understand unfamiliar industries in hours, not months. I learned warehousing and logistics to a sufficient level in hours, not weeks.
5. Learn to build an AI agent yourself. You don’t need to be an engineer to learn the nuances of the new paradigm. Platforms like n8n and Langflow dramatically lower the barrier, especially when paired with a coding assistant.
For example, I had no idea I could train a custom vision model myself within 45 minutes using tools like Roboflow. I assumed this took weeks for an army of people, and this is now a feasible tool in our toolkit as we prioritize what’s worth solving.
I initially set a personal goal to build one agent per month and now build more than one per week. These don’t all have to have high ROI during the learning phase. I even built a Facebook Marketplace agent that helps the family list items quickly as we prepare for a move to Chicago from Detroit.
6. You don’t need to be an AI expert to lead with AI. Your customers already trust you. They’re looking for leadership, not tutorials. We transformed a customer from aspiring to use AI to saving 7 figures because of AI. This is a meaningful mindset change that earns trust and long-term contracts.
START SMALL, START NOW, START WITH YOURSELF
Block one day a week for the next incarnation of your business. Route 20 percent of your leads into a new pricing or delivery model. Involve 20 percent of your team.
For those of us who haven’t solved the founder bottleneck, you’ll be running two businesses at once. That discomfort is not a signal to stop or wait. It’s the cost of staying relevant.
WHAT TO DO NEXT
Block one day on your calendar this month and try to disrupt your own business before someone else does.
If you want to hear more about how I’m thinking through these shifts in real time, I explore these ideas in depth on my podcast, Convergence.fm. Recent episodes like this one about an AI-enabled garbage company cover everything from building AI agents to rethinking service delivery models.
And if you’re curious about how to get started, or want to sanity-check an idea, reach out to me. I’m always interested in comparing notes with ambitious operators who are willing to rethink old assumptions.
A NOTE ON TERMS
Some of the concepts used here, such as the 5 Whys, AI agents, jobs to be done, outcome-based pricing, or custom vision models, may be unfamiliar or used differently across industries.