5 ms·

I regret every RPA project we have to build. It’s just so stupid and avoidable by only buying things with APIs. But it’s apparently the future of a lot of IT,

I regret every RPA project we have to build. It’s just so stupid and avoidable by only buying things with APIs. But it’s apparently the future of a lot of IT, because people aren’t going to stop buying really shitty software or start making architectural demands. It’s going to be an amazing amount of technical debt in a lot of sectors though. I mean, some of the integrations I build before my venture into management, decades ago are still running without the need for maintenance. Yet the most recent RPA project my team build has already used 6 hours of maintenance this week. So stupid.

freedomben 8y ago
For anyone else not hip to the lingo: RPA == Robotic Process Automation? [1] [1] https://en.wikipedia.org/wiki/Robotic_process_automation https://en.wikipedia.org/wiki/Robotic_process_automation
kornish 8y ago
Where does most of the maintenance of the RPA come from? Is it UI changes, edge cases in state (e.g. need to re-login), or other things?
- eksemplar 8y ago
  Well it’s a range of things, this week it was a SAP update that changed the necessary flow because the render speeds changed in the SAP ui. A month ago it was a change to the invoices from a specific company. Every week it’s the business noticing an edge case, or something that could be better. We pile these up, but you’ll rarely have a RPA process that handles 100% so there is an never ending race to get closer and closer. ;) In one of our processes that handles a major process related to payment, we were able to automate the work of two financial workers, but the cost was that it required a new RPA worker to handle the errors, and a RPA worker is more expensive than two financial workers. It was still a gain, because the RPA worker also does other RPA things, but it’s a good illustration of how terrible RPA is at automation compared to the other tools in the toolbox.
  - kornish 8y ago
    Right on. In many ways, makes a lot of sense: UI automation is a quick-to-implement fix which addresses symptoms, with maintainability being the tradeoff. API integration (a la Mulesoft) can require an overhaul in a whole philosophy of organization and delivery, even though it's fundamentally more sustainable. How do you test RPA bots - just wait for them to break? If you're interested, it'd be great to hear more about what you think of other tools in the toolbox. Email is on my profile.
    - macca321 8y ago
      What's cool about UI integration is that the interface is uniform - it's text and clickS and keystrokes. If you are integrating using something like Sikuli, which works by actually looking at the screen pixels (rather than DOM elements, which are an implementation detail), you can use 'smart' techniques like OCR and synonym lists to make a change resistant client.

freedomben 8y ago

For anyone else not hip to the lingo: RPA == Robotic Process Automation? [1] [1] https://en.wikipedia.org/wiki/Robotic_process_automation https://en.wikipedia.org/wiki/Robotic_process_automation

kornish 8y ago

Where does most of the maintenance of the RPA come from? Is it UI changes, edge cases in state (e.g. need to re-login), or other things?

eksemplar 8y ago
Well it’s a range of things, this week it was a SAP update that changed the necessary flow because the render speeds changed in the SAP ui. A month ago it was a change to the invoices from a specific company. Every week it’s the business noticing an edge case, or something that could be better. We pile these up, but you’ll rarely have a RPA process that handles 100% so there is an never ending race to get closer and closer. ;) In one of our processes that handles a major process related to payment, we were able to automate the work of two financial workers, but the cost was that it required a new RPA worker to handle the errors, and a RPA worker is more expensive than two financial workers. It was still a gain, because the RPA worker also does other RPA things, but it’s a good illustration of how terrible RPA is at automation compared to the other tools in the toolbox.
- kornish 8y ago
  Right on. In many ways, makes a lot of sense: UI automation is a quick-to-implement fix which addresses symptoms, with maintainability being the tradeoff. API integration (a la Mulesoft) can require an overhaul in a whole philosophy of organization and delivery, even though it's fundamentally more sustainable. How do you test RPA bots - just wait for them to break? If you're interested, it'd be great to hear more about what you think of other tools in the toolbox. Email is on my profile.
  - macca321 8y ago
    What's cool about UI integration is that the interface is uniform - it's text and clickS and keystrokes. If you are integrating using something like Sikuli, which works by actually looking at the screen pixels (rather than DOM elements, which are an implementation detail), you can use 'smart' techniques like OCR and synonym lists to make a change resistant client.