The Robots Are Coming for Phil in Accounting
nytimes.com
nytimes.com
A.I. was always an overblown threat. The real threat has always been a good UI that encodes most of the rules around a system. You don't need fancy A.I. to figure out most of these tasks, you really just need a collection of widgets someone can throw together to build their own semi-custom automaton.
Knowledge worker need to change and adapt like the rest. And they have it better because they already have a degree and skillset which can be incrementally added to. They may not be able to get an accounting job, but they could probably learn some other skills and get an accounting-adjacent job in sales, management, or tech.
Most knowledge workers need to upskill over their career. Whether you're a mechanical engineer or an accountant, there's a very good chance that your job looks different today than it did in 2011 or 2001 (or, if you're old enough, 1991). It's much easier for a knowledge worker to learn one new skill and combine it with their other existing skills, and be able to leverage that into a new career.
If there was a universal API for financial data exchange between companies then maybe it would be accelerated (starting with Accounts Payable Workflows). Would be my dream to work on a project like this. As for the non-data entry part of accounting, we are even further away in terms of automation.
Software engineers will come along without domain expertise saying "how hard can it be?" They inevitably hit a brick wall when they realize why it's difficult. Even the incumbents with new solutions are selling automation snake oil right now - looking squarely at Microsoft and Oracle.
Commercial AP automation solutions (SAP ARIBA, etc.) exist today but afaik there's no universal standard for electronic invoicing portals, so companies trying to invoice their clients find themselves having to manually enter data on different invoicing portals depending on which software their clients have implemented (there is even a processing fee at times). The burden is borne solely by the invoicing party however.
Should be worth about $10b.
A lot of complexity is from buying a general purpose system (e.g. ERP software) to do a handful of specific workflows. 99% of cognitive effort is tip-toeing around all of the unused functionality minefields.
But we can’t blame our $100m or whatever ERP « investment » that makes staff slow and error-prone. It’s the human’s fault.
This almost always goes unrecognized. Not sure if project manager's fault, the actual software or just the horse-blinders everyone puts on when implementing the ERP