Copilot Catalyst Lesson 7: Fertilizer Helps Growth; Crops Justify Funding
A company we follow closely built something I thought was a model for how to do enterprise AI right. They stood up a self-service LLM for their organization.
We were not involved in the original design or build out, but we knew this wasn’t a rushed experiment or a proof of concept someone threw together over a weekend. They made a serious investment. They did the data work, organized file storage so the model could actually find things and built it to answer questions about the business with real accuracy. It was robust. It was functional. It was exactly what I thought a well-designed internal AI tool should look like.
That’s why it came as a shock when I found out they had discontinued it after only a year.
The overall costs ran higher than they’d projected and despite the quality of the build, the visible value never reached the levels that would justify the price tag. The tool worked. It just couldn’t survive the budget conversation.
For a while, I blamed the usual suspects. Maybe they didn’t communicate the value well enough. Maybe people didn’t know it existed. Maybe they needed more time. But the longer I sat with the discomfort of the cancelation, the more I realized the problem wasn’t adoption. The problem was the portfolio. They had built something that was all cost and no quantifiable benefit, and no matter how good it was, that math didn’t survive scrutiny.
AI transformations are expensive and in the absence of direct revenue generating or cost reduction use cases, the words “increased efficiency” are forced to do a lot of heavy lifting. Even good ideas have a particular talent for becoming black holes. Money goes in, demos come out, launch enthusiasm builds for a few months, but then the bill rolls in and suddenly the thing that felt so promising is fighting for survival.
Two Languages: Fertilizer and Staple Crops
The AI investments that survive budget season have learned to speak two languages at once.
The first language is the language of fertilizer. This is the work that makes everything stronger but doesn’t show up on the company’s P&L statement right away. This is the back-office, the departments often dismissed as cost centers. They don’t close deals but make it possible for deals to get done. Efficiency gains and modernization efforts in the underlying infrastructure of the enterprise matter.
Think about what it means when a finance team saves five hours a week on reconciliation and tightens up their forecasting workflows. Or when HR cuts onboarding document prep from four hours to forty minutes. Or when someone can get an answer to an internal question in seconds instead of losing an hour to SharePoint. Those efficiencies grease the internal workings of the company and allow the revenue generation units to move quicker. It shows up in how the place actually runs. These improvements are real, they have real ROI, but that ROI takes time to materialize.
The delay between investment and visible return has three stages. First, people need time to learn the new tool and actually change how they work, a process that can take weeks or months. Second, the efficiencies start compounding, and you see the uptick in output. The same team is producing more, but the real payoff comes further down the line. Third, you see that compounding turns into capacity. Projects that previously struggled start getting delivered. Work that was sitting in the backlog because there was never enough time suddenly moves forward. The return is real, but you can’t draw a straight line from investment to outcome until the whole cycle has played out, and that timeline is longer than most budget cycles.
The second language is staple crops. This is the work you can measure in dollars right now. Revenue went up. Costs went down. A deal closed faster because someone responded to an RFP two days ahead of the old timeline. Equipment stayed running because a pattern got flagged before it became a breakdown. Errors got caught before they became write-offs. These outcomes have dollar signs attached. They survive scrutiny because the cause and effect is immediate and traceable.
That self-service LLM? It was fertilizer. Beautifully designed fertilizer. Given enough time, the value will show up. But if that cycle is not tracked and articulated to leaders clearly, budget reviews won’t wait for timelines to play out.
The Portfolio Play
A trap that some organizations can make is treating this as a choice. Some go all-in on efficiency, accumulate impressive time-savings projects, and then struggle to track those gains and explain why the investment should continue. Others demand immediate revenue impact from every project and kill the exploratory work before it finds anything worth scaling.
Neither approach works well in isolation. Our approach at FlexPoint Consulting starts from a different premise: you need a portfolio.
A farmer can’t walk into a bank and get a loan based on how much fertilizer they spread. The bank wants to see the harvest. But a farmer who skips the fertilizer to save money is going to have weaker crops and depleted soil. You need both. The staple crops keep the income flowing and justify the investment. The fertilizer makes everything around it healthier and stronger over time.
The same logic applies to AI investment. You want some projects that produce clear financial impact in the near term. Those wins earn credibility. They’re the proof points you can put in front of a skeptical CFO. This buys you time for the efficiency work to keep doing what fertilizer does: strengthening the operation.
That’s why the portfolio matters. The crops provide cover. They generate the visible returns that keep the overall AI investment alive. And that investment is what lets the fertilizer keep working.
The company that built that self-service LLM? If they had paired it with one or two projects that showed clear financial impact, the story might have ended differently. The tool would have been one part of a portfolio that was paying off. Instead, it stood alone. And alone, it couldn’t survive the math.
The Takeaway
I think about that discontinued tool sometimes. It worked. The team that built it did the hard things right, and it still died because there was no hard ROI to point to.
Efficiency gains are fertilizer. They make the organization stronger, but staple crops are what keep the farm running. If your portfolio is all fertilizer and no harvest, it won’t survive the next budget meeting.
The fix isn’t to abandon the back-office work and efficiency gains. It’s to be deliberate about the mix. Pair efficiency projects with ones that show measurable impact. Let the wins from one buy runway for the other.
Build a portfolio that’s credible enough to keep the funding and patient enough to let the real value emerge.
Up Next
Next we’ll look at our last lessons learned with Lesson 8: Curate a Business-First AI Strategy, Not AI-First Business Strategy. We’ll look at why “Where can we use AI?” might be the wrong first question for an established enterprise, and what to ask instead.
If you’ve watched something good get cut because it couldn’t answer the money question, you know this feeling. Tell us about it in the comments. Take what’s useful here, leave what isn’t. And if you want to compare notes on building a portfolio that survives, our team at FlexPoint Consulting is always happy to talk.