What the Green Dashboard Hasn’t Proved
The last thing I ask before a go-live is for the team taking over to walk one customer order through from start to finish, and I watch where they reach for a workaround.
By that point the status report is usually green, and usually for good reasons, because the work the project committed to is done and spending is within tolerance. The walk-through answers a different question, which is whether the people who will run the business on Monday can actually do it.
A project can earn every green light on its dashboard without proving that. Each team’s reporting can be accurate from where it sits, and the trouble starts when those measures are asked to answer the bigger question of whether we are ready to operate. Several stories in this week’s reading came back to that gap. One was an AI infrastructure company that has sold far more capacity than it runs today, and another was a survey of recovery plans that bring systems back before they bring the business back.
Signed Demand Comes Before Delivered Capacity
In its September 25 analysis of Nscale, Reuters Breakingviews reported that the company had signed contracts worth $103 billion as of August 31, while only about 5% of the chips it planned to deploy were operating. The contracts establish commercial commitment, and delivering the computing capacity behind them is still a very large job.
Securing customers before building capacity is a sensible way to finance construction. The risk is letting confidence in demand stand in for confidence in delivery, so an executive assessing that business needs to know what has to happen before the commitments can be met, and which of those conditions the company actually controls.
AI pilots show the same gap inside ordinary companies. Deloitte’s 2026 State of AI in the Enterprise research found that only a quarter of respondents had moved at least 40% of their AI pilots into production. Some of those experiments should stop and others need more work, but either way the evidence from a pilot is evidence about an experiment, and a working deployment needs its own proof.
Think of an AI pilot that produces a useful answer with a specialist sitting beside it. Before that becomes an everyday service, someone has to decide what happens when the answer is wrong and who has the authority to step in, and the operating team has to show it can handle the exceptions at the volume it will actually see. A successful demonstration can leave all of that untouched.
Restoring Systems Doesn’t Restore the Business
Cyber recovery shows the same gap from the other side: the business was running until something interrupted it, and the team now has to show that it can run again.
Cohesity’s September research, based on a commissioned survey of 3,200 IT and security decision-makers, found that 78% of respondents’ recovery plans put more weight on restoring systems than on keeping critical operations going. Only 22% had both documented and tested a “Minimum Viable Company,” its term for the essential operation that has to continue during recovery. These are survey answers rather than an audit of each respondent’s capability, but the distinction they draw is useful.
A recovery team can bring an application back online and correctly mark that task complete, and still leave the business unable to fulfil a customer order. The order might depend on an identity service, a supplier feed, a pricing table and someone with the right permissions, and any one of them can still be down or waiting for clearance. Every task on the list can be finished while the work that matters stays stuck.
In each of these cases the measure is accurate for the question it was built to answer. The problem comes when it is used to support a conclusion it was never designed to carry.
Evidence Should Tighten as Launch Approaches
The evidence behind a green status should get stricter as a project moves toward launch. Early on it is reasonable to proceed on evidence of demand and a credible delivery plan, and demanding full operating proof at that stage would stop worthwhile investments before they start. Closer to launch, leaders should expect evidence that the people taking over can do the essential work under realistic conditions, because carrying early-stage evidence all the way to go-live is how a green project turns into a bad Monday.
This matters most when a project is already in trouble. Before adding people or compressing the schedule, the sponsor needs to decide what the revised plan has to demonstrate. Otherwise the team can finish more activities without resolving the uncertainty that threatens the launch, and the dashboard gets greener while the risk stays exactly where it was.
Let the Receiving Team Walk It Through
The most useful readiness test I know is a walk-through run by the team that will own the work, using the real systems and the real handoffs rather than the project team’s knowledge of how things are supposed to work. The moments that matter are the ones where someone says “normally we’d just…” and reaches for a spreadsheet or a colleague who happens to know the workaround.
The version I use before a go-live has four steps:
Pick one essential customer outcome, such as a single customer order from start to finish.
Have the receiving team run it end to end on the real systems, with the project team watching rather than steering.
Log every workaround, and every dependency that could be unavailable on the day.
Put whatever stays unresolved into the launch decision itself, with a named person accountable for accepting that exposure or reducing it.
The walk-through doesn’t need to reproduce every possible failure. It needs to be proportionate to the decision and to the cost of getting it wrong, and sometimes the right answer is a limited release instead of a full launch, provided leaders understand its limits and have a workable response if performance falls short.
Would your steering committee reach the same conclusion if the operating team had to demonstrate readiness instead of reporting it? If you are not sure, the next review has a good place to start: the work the business must be able to do, and the evidence still missing before you ask it to take over.
About this series: After weeks of writing about the real-life implications of news stories, I started noticing common threads between them. Each Saturday I pull on one of those threads and follow it into the real world to see what comes of it.
About the author: Derrick Greenwood is the founder of GovernIT Consulting, helping founder-led organizations in Western Canada move from debate to done. He has spent about 25 years leading large-scale transformations across Canada, the US, the UK and China, and he still asks for the walk-through before every go-live.
Comments