i guess the lessons i learned from writing marketing content for an "AI" "bot" company (read: shitty chatbot which nobody would willingly use, yet VCs threw money at) would be different. pardon the cynicism here.
1. people don't need education about AI. it bores them. they know what the potential for AI is. what they do not know is the potential for your company's product. they need education about what your product is -- and your product is not "AI". when you brand it as "AI" you immediately start to fight a losing battle against the hype cycle of AI and also your own marketing, which will always oversell the product.
2. specific use cases are all that some of these "AI" products are good for. they're so weak in comparison to what their marketing is claiming that they eventually have to rebrand as a tool for the few things they do correctly because it's cheaper than broadening their strengths.
3. free trials work excellent for AI, but they work terribly for "AI" companies, as the author notes. it takes five minutes of demoing to set up a test case and find out that the "AI" product doesn't work. chatbot companies know this, and hate this. users can immediately tell that the marketing is disgustingly oversold and that they will need to invest a lot of time in getting the product to do what was advertised.
note: i'm not accusing the OP/their company of anything, but we seem to have taken different lessons from similar experiences.
Without having any experience in marketing, let alone AI or AI marketing, as a maybe potential client I find the "what buyers get wrong" and the "buyers asked dumb questions" to be a very adversarial approach to selling a product.
Maybe Talla needs some more lessons on the matter that they haven't yet learned (not specific to AI).
As I see it, the buyer has a given set of problems and wishes to pay money to solve them, it is up to the seller to show (patiently and respectfully) how the product can solve those problems, if the buyer was already an expert in AI, knew already how AI works and its limitations, probably he/she wouldn't need the product at all and would have already solved the given set of problems.
My last company was a Saas company, and buyers wanted to discuss maintenance contracts like they would for packaged software. They dont matter for Saas. Buyers disnt always understand that and it took time to change their buying habits. It was goid in those days to challenge customers on that.
I'm one of the data scientist at Talla working on NLP and ML problems. One of the challenges of selling an "AI product" is that users assume the AI is monolithic. With Talla, the user interacts with the product through either through the chat bot (via slack or ms teams) or our web application to create/manage knowledge content, import content, and manage users. So they (fairly) assume that the AI is the bot itself, which is their primary point of engagement.
In reality, many of the "AI" technologies we have developed are behind the scenes and rather invisible. We have deployed various machine learning models and nlp techniques across the product to make it more intelligent. We are working on models to extract context and entities from user questions, support natural language querying of tabular data, extract "useful" user questions from slack channels, and automatically generate diagnostic workflows from user conversations. So ML/AI is more than the bot itself.