My impression, is that government projects often go like this: 1) We want to digitize and make some process X more efficient, 2) Oh, now that we're automating we can avoid "errors" by removing case-by-case handling (X is of course now set up for individual evaluation by a professional bureaucrat). The result is that not only do you get feature/scope creep (handle all the things for all the cases, or declare failure), if you succeed, the system ends up being much worse (it now follows outdated written documentation to the letter, doesn't allow for any leeway).
The obvious way around this, is make a system that automates the easy 80% of the cases. Allow the system to give good information to, the now much fewer, bureaucrats, and allow them to handle the 20% on a case-by-case basis.
After a year, evaluate, and reduce the 20% to another 80/20.
Give up the idea that you should automate policy. It's completely insane, even if it might look good on paper.
It's like with accounting systems: you should always have a way for a clerk to enter in and account for unforeseen things in a way that a) accommodates the system ("-500 towels missing/destroyed") along with a comment-field to elaborate, and to facilitate revision ("Bus crashed through the window, gave away 500 towels for use as bandages").
This is probably the most important part - every digitization effort I've seen fail, failed because they tried to digitize the existing process, rather than look at what information needed to be collected and creating a new digital process to collect that information.
Another difference with government is the intractable complexity.
This plays to the difference between digitization and automation. Automation is the removal of human judgement. It requires simplification of the rules and processes.
My local government revamped their payroll system. The first effort tried to digitize existing rules, and failed utterly. The second effort reworked the payroll rules (automation), allowing them to use software closer to off the shelf, and succeeded.
I consider the projects I worked on for higher ed to be complete failures. Publishing course catalogs, registering for classes, auditing degrees, etc. Conceptually simple stuff. Alas, management tried to implement every single psychotic rule and mutant edge case for each and every stake holder, preserving paper-based processes over 100 years old. Such fun.
My work for a telco wasn't any less nutty making. There's a reason your phone bill is too high: bureaucracy. Their internal IT is terrible. They can't even provide conf call services or roaming wifi to themselves. I sometimes fantasize that Apple or Google will introduce ubiquitous flat rate service, gut the dinosaurs, and making another gagillion dollars just thru operational simplicity.