Find a very boring problem that you can automate into a product that needs minimal support. By "very boring" I mean: necessary for a certain line of business a

Find a very boring problem that you can automate into a product that needs minimal support. By "very boring" I mean: necessary for a certain line of business and not well addressed by current competitors and not inherently interesting to work on and full of ugly gotchas and corner cases. In general, solutions to these problems will be very profitable because they represent untapped markets. The only difficulties are: finding a problem sufficiently boring, and how much of it you can tolerate before it crushes your soul.