Automating a mess yields an automated mess - Michael Hammer
Just because you can, doesn’t mean you should
I am a big fan of automation, I have spent a fair amount of time automating “stuff”. But I have learned, out of sheer luck, the easy way, that automation is not without it’s problems. Often we either forget or never consider some corner case, and that’s when the “fun” begins.
But the real problems begin when software engages the physical realm, like how DOGE’s “Workforce Reshaping Tool” is used to reduce the US governments workforce.
Seemingly, “Workforce Reshaping Tool” is a reworked version of a 2 decade old software called AutoRIF with a “friendly” web interface that provides targets for lay-offs faster than the previous manual process.
It’s unclear if this was used from the start, but DOGE has already been linked to the 260000 job cuts with more job cuts expected to follow.
Source:
https://arstechnica.com/tech-policy/2025/05/doge-to-roll-out-shadowy-software-to-speed-mass-govt-layoffs-report-says/
There are natural concerns about the process. We have already seen issues with the nuclear weapons program. The use of software for mass job cuts raises issues with labor rights, privacy not to mention, errors and unknown criteria and logic behind the tool.
Proponents of automated decision-making systems often highlight the supposed objectivity. Algorithms don't experience emotional bias, or personal favoritism. However, this perspective often overlooks the fundamental reality that these tools are created by humans who embed their own assumptions, priorities, and blind spots into the code. They also ignore that faster doesn't necessarily mean better, especially when errors could mean wrongfully terminating essential personnel who ensure critical government services continue uninterrupted, based on metrics that are no more than a simplification of the work being carried out.
Like with most decisions, it is likely that second order consequences are overlooked, one such example is the previous mentioned nuclear weapons program workers. Another seconder order consequence, is the IRS forecast of a loss of $500 billion of revenue.
Source:
https://www.independent.co.uk/news/world/americas/us-politics/irs-doge-cuts-tax-filing-b2719911.html
The report from the IRS highlighting that the loss of jobs in the IRS workforce will make it less effective and potentially cause more revenue loss.
To compound the issues around DOGE, this week was know that a software engineer working for both the DOGE and the Cybersecurity and Infrastructure Security Agency (CISA) had his computer infected by info-stealing malware, compromising his credentials, that later appeared in multiple public leaks.
The engineer, had access to FEMA’s core financial management system, raising concerns about potential security breaches, after the credentials surfaced in several datasets.
This has not been, unfortunately in my opinion, the first operational security failure from the current US government administration. A trend that definitely won’t make the US citizens or the US alliances sleep easy.
Source:
https://arstechnica.com/security/2025/05/doge-software-engineers-computer-infected-by-info-stealing-malware/
As technologists, we often rejoice with turning something abstract into software, making a piece of code a proxy for real world. We should not forget that we often use simplification and approximation to achieve our goals, and that every time something is approximated, it loses fidelity. It might be good enough, but won’t scale well. We also must remember that automating requires a very good understanding of what is being automated, and when it comes to dealing with humans, that is very unlikely to ever be the case.
Do you have any good stories around automation? Have you ever been caught by a corner case that made you rethink automation?
We do use automatization, and enjoy this ofc. However we need to always bear in mind, that N+1 knowledable people to maintain these scripts, and a recurring review of the processes with agile and fast reposnisibilty, if comes to automatization to a problem. At US end, I feel that the country leads turning this country onto a bubble mode, without realizing the decisions side effect and damage (somehow feel Hungarian government does the same) So autocratie is over democracy, where damage is secondary or even lower priority even. They will fix manually and polish the damage. But citizens will feel it more and more, as the time pass. Which already has been strated. Wishing for them wisdom, patiente and to be awaken as soon as they can before things goes very bad.