Nvidia just had the best quarter in its history. Revenue was $115.8 billion, up 103% in a year.
The next day, the stock fell again. It now sits 37% below its October record.
The money is real. Nobody can say what it buys.
In October, Finland's Ministry of Finance published a paper called Agentic Finland (Agenttinen Suomi). It asks one question. How does a government use AI agents to do real work, not just answer questions in a chat window?
I co-wrote it. My part was the chapter nobody wants to write. How do you measure whether any of this works?
Here is why that chapter exists.
The paper cites a Gartner forecast from June 2025. Over 40% of agentic AI projects will be cancelled by the end of 2027. The reasons are rising costs, unclear value and weak risk controls.
It also names the trick behind many of those failures. Old tools get a new label. An AI assistant, a chatbot or a piece of process automation gets called an "agent." It still cannot reason or act on its own. Only the slide deck changed.
The metric that exposes it
One measure in the framework does most of the work. It is called shadow-to-production conversion.
It is simple. Of all the agents you piloted, what share made it into real production within 12 months?
Most organizations never count this. They count pilots launched. Pilots launched looks great in a board meeting. It says nothing.
The paper's conclusion is blunt about it. The biggest risk may not be failure. It is getting stuck in permanent pilots and never changing how the work is done.
A pilot that never ships is not a test. It is a hobby with a budget.
Sales Owns Retention, Not Just Closed Deals
Pebl’s Chief Business Officer, Dutta Satadip, joins The GTM Garage to show how Pebl merged Operations and GTM under one leader, and how it aligned goals and incentives across the customer lifecycle.
You’ll also see how Pebl uses AI across that lifecycle and the stack behind prospecting, pipeline, and handoff to onboarding.
15 min conversation. 15 min open Q&A. For Seed to Series A teams.
No baseline, no proof
The second rule is harder. Before an agent goes live, you measure how the work is done today. Minutes per case. Cost per case. How often things go wrong.
Without that starting point, you cannot prove the agent changed anything. You also cannot defend the money afterward. You can only tell a story.
Then the framework counts things most AI dashboards skip:
Total cost per agent per year: development, infrastructure, maintenance, training and licenses
Token cost per case
How often the agent hands a task back to a human
How often it states things that are false or made up
How often decisions prepared by an agent get appealed, compared to decisions prepared by people
That last one matters in a country where official decisions must be explained and can be challenged. Legal responsibility stays with a human official in every case.
The number we told people not to lean on
Finland has a hard math problem. Nearly one in three public sector workers retires between 2025 and 2034, according to pension provider Keva. The government is also chasing about €100 million a year in agency savings.
The obvious metric is headcount. Agents in, people out.
The framework says that number is not enough on its own. It ignores where the freed time goes. It ignores the total cost. And if staff read the dashboard as a layoff target, they stop helping the project succeed.
So the framework asks a different question. Where did the saved time actually go? Into judgment, preparation and advice? Or nowhere?
Turn your expertise into working software
Describe what you want to build
Get real software, not a prototype
Lovable handles payments, logins, databases
Change any detail yourself, any time
Put it in front of customers this week
Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.
The asset you can burn once
47% of Finns trust their national government. The OECD average is 39%, according to the OECD's 2024 trust survey cited in the paper.
That trust is what lets a government try agents at all.
One rollout that nobody measured can spend it.
Nvidia's best quarter ever did not stop the stock from falling.
A small government wrote down what to count before it spends more.
404 Found covers AI developments from a European Insider, three times a week.





