Maturity of UI and ability to have no-code conversational analysts working within the tool.
The ability to train contextual entities within an intent, to handle unknown examples of a trained entity.
Strong set of documentation underlying the product. Clear and transparently laid out.
Clear ethical framework, and a high level of transparency of how the product works and how both the users, the developers and the AI companies data are all protected. Clear opt-outs which are respected.
Good start for free model and move forward from there.
Ability to handle large intent and data sets.
New searching of nodes, and real time "Try it out" node analysis of failure.
Simple integration of cloud functions to call and process external cloud services without needing an external orchestration layer.
Easy integration with a variety of iPaaS tools
node-red being free and supporting such a wide variety of Watson and other integrations out of the box, and then with community led extensions.
Currently incomplete movement from single workspace based, to multiple skills (workspaces) based agents, and from a message based pricing to an active user based pricing model. This is going to get there, but at the the moment the product is somewhere inbetween. I would also like more support for analytics in terms of conversational flow and drop outs, the current strategy of pushing the data out into Watson Studio is incomplete as there aren't easy prebuilt notebooks, or dashboards to build the views you would typically have. So whilst you can do it, there feels like there is a step missing. The "Premium" only feature set is very much disliked as it contains things which feel like it should be in the product. Not giving them for a free tier is fine, but even if you pay for standard level they aren't included, instead reserved for a contact only enterprise tier. It makes it impossible to evaluate them properly and means that they don't get the full workout they need to become proper full features.
Ensure you have a plan for your automated conversational testing and your analytics of successful conversations.
Integrate with your DevOps systems to have daily conversational releases, and automated testing of the affects of each days changes.
Engage good upfront conversational design.
Maintain your training sets and testing sets independently
24/7 contact via website, transforming response figures from a variety of social and web platforms. Outbound lead generation, increased inbound customer satisfaction. Large scale effort reduction. Interactive discussions with users to discover things they might not have thought about. Wellbeing conversational styles.