I was asked to write an analytical brief on introducing AI into public institutions.
One chapter was devoted to the question of how e-government was implemented in Russia, and what lessons, strengths, and weaknesses I would highlight so they can be taken into account today when introducing AI.
I spent the weekend in my archive, wrote detailed answers, and thought this might be interesting to many. Here is the core of it.
What went wrong
Supply was built without demand. Portals and systems were created top-down, in a hurry, to keep up with other countries. Nobody asked citizens. In the year of record digitalisation reports, two-thirds of the population did not use the internet.
Old processes were digitised instead of being changed. On the ground, only the simple and safe parts were implemented; anything that disturbed the familiar order was discarded. The result was the same bureaucracy, only digital.
Indicators became the goal. The country jumped 32 places in the UN e-government ranking. That same year, my field tests were finding dead email addresses, officials unable to open an attached file, and agencies refusing to accept forms from their own website. The ranking measured the shop window, not the service.
Citizens went unnoticed. In parallel with the state, people were building their own services: city problem maps, search-and-rescue networks, mutual aid platforms, all funded by citizens themselves. The state ignored them for years, then simply copied the best of them at public expense.
What worked, and why
The national services portal and the network of one-stop service centres became genuinely popular and convenient. Those successes shared five conditions: clear mass demand; a single owner of the service instead of a dozen agencies; strict standardisation; a real, not reported, reduction in the steps a person has to take; and continuous development after launch instead of abandonment after the deadline.
The main lesson for the AI era, as I would put it: technology does not fail on its own. It fails where it is built without demand, and it works where it shortens a person's path to a result. That rule has not aged a day.
If you are introducing AI, start not with the tool but with a question to your customers, your staff, your residents: what do they need? It is the cheapest part of the whole project, and the most important one.