Governments aren’t struggling to innovate. They’re struggling to scale.  

It came up in the session on AI in public services. It came up in the panel on legacy modernization. It came up, most bluntly, in a session about smart cities that spent an hour asking why so many good ideas never leave the pilot phase. Across three days at G4I 2026, one problem kept resurfacing in nearly every room: the gap between a pilot that works and a service that ships. 

The pattern is familiar to anyone who has worked in or around government technology. A pilot launches, performs well, gets presented at a conference, and then quietly stalls. No one owns the decision to scale it. The budget for what comes next was never secured. By the time production is finally on the table, the technology has moved on, or the enthusiasm has. 

We went back through the sessions where this problem was discussed most directly, “AI in Public Services: Pilots, Scale, and Digital Sovereignty” and “Smart Cities”, and pulled together what practitioners actually said works, and what reliably kills a pilot before it reaches anyone it was meant to help. 

The guide is built around three things. First, the conditions that need to be in place before a pilot has any real chance of scaling – from having a named, accountable owner to redesigning the workflow around the tool rather than bolting the tool onto an unchanged process. Second, the failure modes that came up again and again, sometimes described almost identically by speakers who had never met each other – pilots built around easy wins with unusually motivated volunteers, solutions chasing a fashionable label rather than a real civic need, and the quiet trap of a free or low-cost tool that becomes too embedded to remove. Third, a short checklist you can run before committing resources to move a pilot toward production. 

None of this is about being more cautious. It is about being honest, early, before the pilot begins, about what production requires – so that success is not measured by how many pilots get launched, but by how many services genuinely reach the people they were built for.