Not a surprising comment for anyone who has been around in IT for a few decades and one which has likely been made many times recently.  To be clear, I am not saying that AI is a glorified search engine. Rather I'm saying it is enabling huge productivity gains with the way people work which is, of course, exactly what Google did.

Before Google came on the scene and gave the world a functional index to the web, working in IT was very different when it came to working through any issues. We used to have to make long phone calls talking to techs we knew personally; read massive tomes on all sorts of technical subjects; have long, complex ticket conversations with vendor technicians; and so on.

It could take a long time, weeks, to diagnose and find the correct solution to many IT issues. If your issue was obscure enough, then that time could drag on with both users and IT staff becoming increasingly frustrated with the lack of solution.

It was both slow and expensive. Google changed all that.

Not immediately, of course. It took time for Google's index to be built; for tech based sites to grow and document issues and solutions; for vendors to place documentation online and build knowledge bases etc.  However, the end result was that issues which could take weeks to resolve were being resolved in minutes - literally minutes. The productivity increase was tremendous.

Of course, IT complexity also increased over time. Many new technologies and systems have since been developed which have increasingly complex requirements. On the surface it's all nice and shiny but the back end can still be as sweaty as it used to be.

This is where AI comes in. I have recently used it to complete a process which would have taken me several weeks at a minimum to complete if I had to do it myself. I migrated our service desk from one platform to another.

Service desk migration to a new platform

Over the space of a week I have succeeded in migrating our service desk from one platform to another.

This was a process which involved

  • exporting data, attachments, and images.
  • converting XML files to CSV files.
  • importing the CSV files into the new platform.
  • Retrieving new identifiers from new platform.
  • Matching the identifiers with yet more XML files which were converted to CSV.
  • Identifying correct attachment files and ensuring they were matched with the new identifiers.
  • Downloading about 5GB of images and uploading them into a separate site from which they could be linked to in the changed data.
  • Creating yet more CSV files with links to zip files for the relevant attachments.
  • Importing all the CSV files and zip files into the new platform.
  • Checking everything

Using AI I started this on a Thursday one week and the process was complete by the following Thursday - and I completed my usual workload as well. Actual time spent working on this would have been around a day in all.

The process involved the migration of

  • 17,500+ tickets.
  • 135,000+ notes on the tickets.
  • 11,200 files as attachments.
  • 23,000+ image files
  • 100+ clients and hundreds of people.

Not a huge migration but still one that would have been onerous to the point where if I didn't have AI to do the donkey work, we would either have stayed on the current platform or started from scratch on the new platform.

So what did AI do?

The donkey work of working out the full process for the migration and developing the scripts for completing the work, working through any issues which surfaced (and yes, there were issues).

My job was to create API tokens, add them to the scripts, run the scripts, run various checks, ensure what was happening was actually what was supposed to be happening, and generally keep an eye on things. I probably could have let AI run all the scripts too but I'm still not too keen on giving AI access vi API, SSH, or site logins just yet, at least not to my production sites.

If I had had to do this work, it would have required a lot of reading of two or three different APIs, writing various scripts to do all the work, test and re-test, work out the issues which arose, and so on. As I said, it would have taken weeks for me to complete this. 

Two years ago this would have been unthinkable. Two+ years ago I was quite disparaging about AI in the shape of LLMs. As it has become more capable, more accurate, and more importantly, usable in areas I work, I have changed my views.

Over the past year or so I have delved further into it as a tool and am increasingly happy to use it. It has definitely boosted my productivity in my technical work. I'm now using it to rewrite our documentation via reading existing documentation, reading code, and through use of a local site hosting our application. 

It's producing good and mostly accurate work. Of course, everything still needs to be checked by me but this isn't a huge job - particularly as I need to select and add images to the documentation. However, earlier today it produced a set of documents which would have taken me days to write. It also pushed the documents into our new documentation platform and, linked everything correctly. All I do in the platform is set the correct permissions for our users.

I'm happy with that.