Superflows lets a software product add an assistant without building one. Upload the API spec, integrate with a line of code, and users can ask analytics questions that require API calls, take actions with confirmation, and query the product documentation conversationally.
Requiring confirmation before actions is the design decision that makes this shippable. An assistant embedded in someone else’s product acting directly on a customer’s data is a support incident waiting to happen, and an approval step keeps the vendor’s liability where it belongs. A playground for testing and interaction analytics cover what happens after launch.
Learney AI describes three tiers, Testing, Scale and Enterprise, with no figures published. The footer notes the company pivoted to this, which is worth knowing about a dependency you would embed in your own product. Assistant quality depends entirely on how well the API spec is written, and embedding a third party in your product surface makes their downtime yours.
