Why the software gets shelved when the foundation is still run by hand
The situation I walked into
I joined a large business and, about two months in, was handed a project that had lost its owner. They had bought an expensive project-controls and analytics system, the kind of tool that promises a single live view of cost, progress and risk across everything the business is running. It had been approved by the person leading it, and that person had left roughly a month after I arrived. The work had stalled. I picked it up part-way through, an orphaned roll-out with no one steering it.
On paper it was a sensible buy. The problem it was meant to solve was real. The old system underneath it was more than a decade old and had been heavily customised on the way in rather than kept close to standard, so it had never been properly updated since. Because of how it was set up, costs were landing against the wrong projects. The cost team was reconciling by hand, pulling transactions out one at a time and checking each against the right project. There was no trustworthy live picture of where the money was. So a lot of the real analysis was being done offline, manually, in spreadsheets on the side.
The problem was not the new software, and it was not the people. It was what sat underneath it.
What I did
I ran a focused sprint, roughly twelve weeks, to get the new system live. Clear scope, a small group of the right people, a hard end date. I leant on two cost-control managers as the technical experts, we configured it, connected what we could, tested it, and delivered a working system inside the window. By the technical measure, the job was done. We had built the thing we were asked to build, on time.
But the closer we got to live, the more one thing kept surfacing. The data the system depended on, the base numbers that made every report mean anything, was still being kept by hand. It lived in spreadsheets, maintained manually, outside the core system the new tool was meant to read from. So there was no clean, live foundation for the analytics to sit on. Every clever report the new system produced was only ever going to be as good as a set of files someone was updating on the side, and hoping they were right.
I flagged it plainly. You cannot bolt a clever analytics layer onto a foundation that is still run by hand. The tool was ready. The ground it stood on was not.
The outcome
The system was delivered, and then it was quietly shelved. Not because it did not work, but because keeping it fed meant keeping the manual spreadsheets going alongside it, and that was never going to hold. A significant investment ended up sitting unused, through no fault of the team who bought it or the team who built it.
What made it click was a new cost-control manager who came in and named the real problem straight away. The foundational data was being maintained by hand outside the core system, not in it, so the clean data the analytics needed was never actually there. Buying an oversight layer to sit on top of manual inputs was working backwards. His plan was the right one, and it was the unglamorous one: kill the spreadsheet dependency first, get the base data into the core system and keep it there, and only then layer the reporting on top. The order had been reversed, and the order is what mattered. It is also worth saying the seat kept changing hands over those years, and every time it did the work reset to square one, which is a big part of why a good idea never got the foundation it needed.
What it means for you
This is one of the most common traps I see, and nobody in it has done anything wrong. When the numbers are a struggle, it is genuinely tempting to buy the shiny system, or now the AI tool, and expect it to sort out the mess underneath. It will not. A tool, any tool, sits on top of your data and your process. If those are held together by hand, the tool either breaks or gets abandoned, and the money goes with it. AI makes this sharper, not softer. Point clever automation at a foundation that is still run manually and you do not get order, you get confident output built on numbers no one can stand behind.
The order that works is the boring one. Get the base data right, get the process underneath it solid and repeatable, get it into one place that can actually carry it, and then automate on top. That is a lot of what I do with owners at PeakRatio, the groundwork that makes the next investment stick instead of sitting shelved. If you are weighing up a new system or an AI tool and you are not sure the foundation under it is ready, I am happy to talk it through.
The Ratio Check is a free, five-minute read on where your business actually stands, and it is the best place to start.