Sitting through hundreds of startups during YC demo days, you’re not always sure if you’re actually perceiving patterns, or if your mind, like coffee fighting disdain, is inventing business plans in a kind of pareidolia. But this year, the theme was very clear: “AI can do that, maybe! Maybe.”
Certainly today’s AI models are more capable than those of yesteryear. But we have seen time and time again how these systems perform well, but with system requirements or as tools for reliable and repeatable results.
It’s hard not to see this batch as the forerunner of the next wave of AI-powered shovels. Pick a use case, tweak an existing model a bit (no one builds their own), cherry pick some cool examples for screenshots, and bolt on the frontend UI. Congratulations, you are now the first AI social media content curation platform for independent bars and restaurants in the Middle East and North Africa. Buy a couple hundred 5-star reviews and you’re on your way!
Now, it’s not like restaurants in Cairo and Beirut can’t use the internet as a useful tool to gain some interest and attract new customers. As AI currently exists, it’s like admitting that it’s okay for you to do something.
Creating an AI-powered chat agent that answers the phone at your business sounds great when you set it up as a way to never lose a customer. But what does the business that calls the customer think when deciding the welcome AI deserves? Personally, I would hang up and try someone else. How about an AI calling business person to make an appointment? Same thing.
Realizing that your email is simply “personalized” by AI is like telling us, we can’t be bothered to personalize our emails, but we want you to think we do. Don’t you feel cheated? It is a systematic deterrent to customers.
If your first interview with a company is with a chat agent or someone who clearly reads cues from their knowledge base or whatever, do you feel like someone has joined a team or department for an installation level? You are not even worth the full attention of a competent person.
That’s not the vibe I got from every AI startup in this batch of YC, but I definitely got it from quite a few. Here’s a partial (!) list of “AI can do that, maybe” companies I’ve written about.
- Kind of – AI-first document editor.
- Iliad – Create game art assets.
- position – Build workflows in one-line applications such as on-boarding.
- Nucleus – An AI-powered onboarding orchestra that understands the “true nature of the business”.
- Hadreus – SEC-compliant robo-advisor.
- Speedybrand – Marketing content generated for SMBs.
- Kuzal – Language lessons with an AI tutor.
- Booth. No – Generative AI “photographer” for e-commerce.
- Squeak – Natural language math tools.
- Berri.ai – Creating ChatGPT applications as a service.
- Divorce – Financial news insights “enriched” by AI.
- Credal.ai – A ChatGPT-like interface for employees who refer to company documents but protect trade secrets
- Religious – Add an AI data assistant to your app.
- linkgrap – Suggests things from the knowledge base and adds them to a conversation or forwards notes directly in the browser.
- Draw a ship – Automatic sales emails.
- Don’t spill – Automate market research based on reviews and feedback
- Tin no – Convert knowledge base to custom LLM.
- Real wind – Accounting and financial processes powered by AI.
- Smart labs – Collect insights from customer service call data and emails.
- Only paid – Automate invoice payment, handle overpayments to suppliers.
- Cyber – Insurance industry tasks such as answering questions and writing.
- Meru – A platform to train your own LLMs.
- Sameday – AI that can be called to schedule workers such as plumbers and roofers
- She fell – Analyze customer calls live and discussion points.
- Synchronization – AI to analyze customer emails.
- Equivalent AI – Video courses generated using AI.
- hidden – Automating electronic health records.
- Avoca – AI receiver to answer missed calls on SMBs.
About 30 seconds ago, I added thoughts about the companies to these brief and possibly inadequate descriptions. But I realize that the list runs the risk of becoming a lot of complaints (not to mention too long). No one likes to read someone shooting ideas left and right, especially when most of those ideas were worked hard by important people. It’s easy to criticize. So easy that someone in a summer collection can try to make it automatic!
But I ask you to look at that list and not wonder about some of the entries: Is That’s what he said. What is really needed? Doesn’t it require a lot of supervision? Doesn’t this promote accountability or reduce transparency? Has anyone asked customers if they want this? Who will verify and analyze the results – another AI? Who is displaced by these devices? Who trains people on them?
Practically every company that pitched said it went live a few weeks ago and miraculously they were already at a healthy ARR. But a few weeks isn’t enough time to install a major automation tool and read its documentation, let alone evaluate its performance if it’s worth the price tag. I can’t imagine that even half of these are potential customers, let alone used.
I can’t share an example: a creative marketing image company had the following requirement for the system to work with Slides: Our Classic Ketchup is made with only sweet, juicy, red ripe tomatoes for the thick, rich flavor of America’s favorite ketchup signature. AI Copy: Sweet and juicy ketchup for everyone! If I were a marketer at Heinz and he was in a test drive I was given, I would stand up and thank them for their time and open the door.
Some companies have admitted that they are halfway through the program and have recently signed up for this new application. We need to allow for our adventurous and adventurous nature. But do these companies really feel “innovative” to you? They seem to be big fans of creativity as they sneak into the room and try on the clothes. (“Beautiful… here, try it, fintech”)
I know I’m overestimating the amount of work that goes into building even the most authentic AI-powered B2B SaaS service, but many of them are like our old hackathons where someone provides an API and everyone else tries to shoehorn it in. Hoping to get that $1,000 gift card from SAP or whatever to the most accurate voice app. There is joy in the process of creation, but the results do not stand on their own.
Maybe when one of these companies goes unicorn and everyone laughs at the skeptical TechCrunch writer, I’ll be wrong. But I can’t help but feel dismayed to hear the founder assert that their AI can do better when I suspect the conviction is based on false pretenses.