A friend of mine runs a small-to-medium Amazon business that sells high quality stationary. He sources from all around the world, which shows how even small companies tap into global supply chains and labor pools.
He uses some SaaS tools, particularly Helium10 and Jungle Scout, for marketing intelligence and other kinds of support, but his internal dashboards and controls are built by manually pulling data from a variety of sources so he can track inventory, lead times, volumes, sales, campaign efficiency, etc.
“What I really need, and what would make AI actually helpful, would be if I could use it to automatically access Amazon’s [selling partner] API.”
Well, now there’s an MCP for that.
Model Context Protocol is a mechanism by which AIs can access APIs programmatically, which lets you build your own agents to get stuff done. LLMs, whatever their other limitations, have solved parsing human speech, so all manner of “glue operations” — such as querying a database, or passing data back and forth across a network — is now within the reach of non-technical operators. At first this means automating a bunch of manual processes, but it also enables entirely new methods, as a new class of people can scratch their own itch.