If engineers knew what RPA was, they'd probably scratch their head at the idea.
RPA stands for Robotic Process Automation, which already seems to be a misnomer because there's not an actual "robot" anywhere. It's about using traditionally human inputs (mouse, keyboard, GUI) in an automated way to accomplish a task. Things like screen scraping, mouse movements, or sending keystrokes into a form count as RPA. The method fills a gap in legacy applications that don't offer APIs, allowing those applications to connect to other applications.
Over time, the market will choose companies that offer proper APIs over RPA. One is simply a 10x improvement - engineers will always choose an API over an RPA solution. APIs are orders of magnitude faster. APIs are more reliable because they work on structured input. APIs can be authenticated, authorized, metered with low overhead.
Two big trends are disrupting RPA. First, the shift to SaaS breaks RPA. With SaaS, the user doesn't control updates, and that means layout updates that break RPA workflows. Buttons change position, text boxes change labels, and RPA can't figure simple changes like that out. Second, the shift from desktop applications to web applications makes RPA fragile. In an ironic twist, the best "RPA" is simply tapping into operating-system-specific APIs, like the Windows Win32 API. Browsers like Chrome don't have an equivalent (and never will, can you think of how much of a security issue that would be?).
That doesn't mean there's significant business value in the problems that RPA is solving. Look at Plaid's (screen scraping for financial data) latest funding round at a $13 billion valuation, or UiPath's (desktop RPA) market cap at $40 billion. But most businesses look great before they get disrupted.