It seemed to me that many of the darlings of investors in the last few years in the data space were "solutions" companies even if they claimed not to be. That is, Mongodb, Cloudera, Hortonworks, Continuum Analytics, etc.
These companies generally had an open source product and then either support contracts (e.g. solution) or proprietary extensions frequently of dubious if not negative value. (e.g. in the devops age, a web user interface slows down how fast you can make a Hadoop cluster, it doesn't speed it up)
Many of these companies have had a reckoning, others still fly high.
In A.I. or anything else that is really new you will have to give customers special help if they are going to succeed with your product, if anything I would point investors to all of the self-service A.I. products out there that are a complete joke (e.g. IBM Watson text recognition, BERT if you care if your results are useful, etc.)
Here is a buzzword I like:
https://en.wikipedia.org/wiki/Software_product_line
You can make a model of a customer engagement that includes some revenue from the product license and some from the "setup" part, probably this is also some maintenance work you need to do too.
You can ask questions like: can you get the setup effort down over time since you do similar things for different customers, build a toolbox, etc? can you or a customer hire out the setup work to a third party?
I know some people who have interesting ideas in the "low code" space that are partnering with a solutions integrator, which is a really great idea.