As what I automate mostly anything that is repetitive, if I have to do it more than five times I automate it (not a hard rule).
About the tools I just search if there is anything that can do it, otherwise I will try to code it.
There is not much more to say about it.
kiriappeee, your account is shadowbanned.
In answer to the original question, we're similar to the poster; if it's a non-trivial task that will need to be done more than a few times, we'll set up a script.
I haven't given a long answer in a while. So here's one.
Everything. Almost. I work in an ecommerce company and through an odd set of circumstances joined to head the production team. That's the team that takes photos and QCs stuff and finally uploads things to the site. I inherited a team of 10 and despite work levels climbing to over 20x of what it was at the time I joined the company, the team size is still 10. I was originally hired to streamline processes and convert the roles in the production team into jobs that could be done in an A-B-C "follow-the-manual" way. The purpose was to make the production team low cost with easily replaceable members. If we wanted to scale with the work load, we'd just hire more people since it would be a low paying role anyway.
I opted to approach the problem differently (and I took a big risk in doing so). Instead of making people replaceable, I trained them to be irreplaceable. And to assist them in doing so I built over 30 applications during my 2 years that I have been here.
Some of the applications I built:
1. Newsletter Generator: Since the company started as a daily deals company, we had a daily newsletter to send to our database of users. The database would be broken into "interest based subsets" and the content of the newsletter would change slightly depending on which subset it was going to (ex- people purchasing deals on hotels would get travel related products on their newsletter). Building newsletters would mean editing HTML manually and this process would take up to 2 and a half hours to complete each day (including QC and error correction time). I built an application where the content admin would simply choose what template blocks to use, arrange them, and just paste the product URL and image URL in and then click "generate". The newsletter generator visits those URLs, scrapes info out of the pages, and constructs the newsletter in a matter of seconds. Oh and the links get saved, so if you want to redo a newsletter or include similar products in the next newsletter, you just choose from the saved list. The entire process was brought down to 10 minutes.
2. Converting products to CSVs: This is my personal favourite. We started a shopify like concept where we sign up merchants to have a store on our site. We don't have the infrastructure though to let them have their own portal to manage products. Instead they send CSVs with products to us to upload. Problem is, the product lists were huge. The content admin in charge of this had a top speed of 25 SKUs per day. We were looking at hundreds of SKUs from several merchants each. Thankfully our ecommerce backend supports CSV uploads. Obviously the format for the system couldn't be the one given to the customer. I built a CSV generator to convert the merchant's CSV to a system compatible CSV. To give an idea of how much this saved time, a merchant with 9640 hours worth of work was completed in 3 hours (I had some manual verification work to do with images that took time. The actual CSV generation and upload time took 10 minutes). Our backlog of merchants was cleared in under a week (We were looking a over a 18 months of work prior to this)
3. Resizing images Our graphics guys get cropped images which they then center on a white background and resize to 3 different sizes. When we started the business model described in the previous point, we were overloaded with 1000s of images to process. We were looking at months of work. I spent a weekend writing a script to plug into adobe illustrator that would iterate through a folder and its sub folders, and scoop up psds, jpegs, and pngs, and do the resizing. This weekend worth of work converted the months of work to about half an hour of the computer being left to work by itself while we chilled on bean bags
4. Competitor analysis While we were heavily into the daily deal business, I wrote a script to visit our competitors websites, track their deals, categorise them, get the numbers sold, and at what prices, and look for changes in deals (such as pricing). This was all written to an sqlite db that connected to an excel sheet from which we were able to generate pretty reports of rate of sales and such stuff on our reports. I eventually decommissioned this after all the other daily deal businesses went out of business. I went on to build a lot of other competitor analysis tools that checked on inventory, merchants and other kinds of information that we could use. Yes, this part was kinda dodgy.
5. Automated QCing After uploading the products to the site (this was during the low volume time) I built a little app to read the word document that we were sent with details of the deal to upload. It read it, visited the URL (the path was also in the document), compared the uploaded deal to the details on the word document and gave a QC report.
The final result of me building this stuff was that I got to build a team full of awesomeness as opposed to a team of boring cogs. I kept my job. But most importantly, I actually invented a new niche job for myself and I was given autonomy to walk around the company and help other departments by automating their processes and creating applications for them to use. Been thinking of doing this as a consultancy but not sure how I'd go about it :-/ . But there you have it. Non techie stuff automated all over.
My threshold for considering automation is doing it once :) but if the work to automate is non-trivial, then I usually don't get around to it until the 3rd cycle.
I usually evaluate a few things, in this order: 1. How much time will it take to do manually vs automated? 2. What are the chances I'm going to have to do this again? 3. Is it useful to just me, or other people too? 4. Do I get any value by doing this without automation? (e.g. will I gain insight by taking time to go through this process, that I would miss if I built a tool to just do it)
...but I suppose that's mostly for "tasks" whereas you mean "processes" in the sense that a process is more difficult to automate because it has more steps and dependencies. In that case, I think there are more evaluation criteria like:
-What is the risk of doing this manually and screwing it up?