Ask ten revenue cycle leaders whether they have automated their operation, and most will point to an offshore team or a BPO. That is a reasonable way to cut cost and add capacity. It is not automation.
Automation, by definition, is machines doing work. Sending claims offshore moves the work to different people. It can be the right call for efficiency and expertise, but it does not change the underlying economics the way real automation does, and it is worth being precise about that difference before you buy anything labeled "automation."
When we started Substrate, we assumed every practice needed AI agents everywhere. What we found, even inside large, sophisticated groups running hybrid onshore and offshore teams, was that most of them were barely using the off-the-shelf tools they already had. So we got pragmatic. Here is the order that actually works.
First, use off-the-shelf tools well. If a mature, stable technology already does the job, use it before you reach for anything more exotic. Your clearinghouse, for example, sends back a rejection message when a payer rejects a claim. If you are not writing those rejections back into your system, that claim looks like a no-response claim, and your team burns time logging into a portal to look up something that was never there. Fixing that costs you nothing in AI and saves real hours.
Second, deploy those tools well. Good implementation, correct configuration, data flowing to the right places, and queues mapped so claims move logically without a human retyping information just to advance them. You can get a lot of leverage without ever using AI.
Only third, once the first two are done, should you reach for advanced tools like AI. Skip the first two stages and you will feed bad data into expensive systems and get mediocre, costly results. Do the first two well and you will be surprised how little AI you actually need to be excellent.
The lesson is not that AI does not matter. It is that automation is a discipline, not a purchase. The teams that win treat off-the-shelf tools, clean deployment, and AI as a sequence, not a shortcut.
