7 AI startups that stood out in YC’s Summer season ’22 batch

It is that point of 12 months once more. This morning, Y Combinator (YC) hosted a demo day for its 2022 summer time cohort – the thirty fifth demo day within the incubator’s historical past. That includes founders from 30 nations and startups from each trade, together with developer instruments, fintech and healthcare, the day was not wanting compelling pitches.

Competitors has been fiercer than regular, because of YC’s determination in early August to cut back the lot measurement by 40% to round 250 corporations in gentle of financial headwinds. However one specific class of startups stood out: these making use of AI and machine studying to unravel issues, particularly for B2B clients.

This 12 months, there have been solely 14 such startups, in comparison with 20 final 12 months, which is smart as a result of the general cohort can also be smaller. However the bundles share a unifying theme: gross sales. Their merchandise largely goal hurdles in gross sales and advertising at a time when companies are going through the pressures of the recession.

Financial challenges apart, the massive addressable market makes gross sales a gorgeous subject for startups. Grand View Search ankle the one gross sales pressure automation software program market at $7.29 billion in 2019.

AI driver

AI driver develops a instrument for gross sales reps that routinely interprets name recordings into structured information that then straight updates a buyer relationship administration (CRM) system. The thought is to save lots of reps time and guarantee their managers that pipeline information is up-to-date.

It needs to be famous that different platforms corresponding to Fireflies.ai and Microsoft Viva Sales do this too. However Pilot AI founder Max Lu, previously a software program engineer at Salesforce, says his product is extra complete than most and might generate a abstract of every name together with information factors that match CRM fields and questions requested by representatives, along with key components of the recipient’s response.

AI driver

Image credit: AI driver

By sort

By type can also be within the retail area, but it surely focuses on textual content prediction in net purposes via a browser extension and server-side API. Initially developed as a smartphone app, Typewise — which claims to have Fortune 500 shoppers within the e-commerce and logistics industries — can auto-complete sentences, insert sensible snippets, auto-reply to messages, and examine type consistency and of grammar.

It appears to be like a bit like TextExpander and Magical. However founder David Eberle says Typewise is suitable with any CRM system and may be custom-made based mostly on an organization’s information, with an analytics element that means phrases and phrases to make use of.

YC Summer season 2022 AI startups that did not fall into the gross sales and advertising know-how class tended to concentrate on developer instruments, one other profitable avenue for progress. Contemplating that 55% of builders battle to search out the time to construct in-house apps within the first place, in keeping with a current investigationVCs actually see a chance right here: they invested $37 billion final 12 months in startups constructing improvement instruments.

Monterey AI

Monterey AI tackles a decidedly totally different a part of the product life cycle: improvement. Founder Chun Jiang describes it as a “product improvement co-pilot” who replaces paperwork with workflows that routinely generate product specs, together with function concepts, measurements, designs and launch plans .

Utilizing Monterey, clients select a product mannequin based mostly on their use case (eg, “software program as a service”) and configure inputs, checking dependencies to resolve conflicts. Jiang says the platform can uncover conflicts and dependencies throughout groups whereas offering a portfolio overview to align options.

Monterey AI

Image credit: Monterey AI

AI improvement instruments

AI development tools might maybe be utilized in tandem with Monterey AI.

Dev Instruments AI gives a library designed to make it straightforward to put in writing checks for net purposes in present improvement environments by merely drawing a body over a screenshot. By making use of laptop imaginative and prescient, it finds components on net pages like search bins and buttons, and might even see controls in net video games. It could possibly additionally take a look at for crawl errors on pages, together with damaged hyperlinks, 404s, and console errors.

As founder Chris Navrides factors out, writing end-to-end net checks is a historically time-consuming course of, requiring digging into the web page code a number of instances as the appliance beneath take a look at evolves. Assuming Dev Instruments AI works as anticipated, this could possibly be a helpful addition to QA testing groups’ arsenals.

Mayan Laboratories

Mayan Laboratories creates a platform to translate pure language into code. Just like GitHub’s co-driver, Maya incrementally generates packages and shows ends in response to steps in English.

One in all Maya’s founders, Sibesh Kar, says the service builds apps utilizing a mixture of conditional logic, AI-powered search and classification, fine-tuned language fashions, and sample technology. Presently, Maya can question and plot information from an exterior supply like Google Sheets, Notion, or Airtable, and carry out actions on that information, like ship an electronic mail, obtain a file, or replace a database entry of information.

The long-term aim is to increase Maya to duties like net searching, connecting APIs, and automating workflows, which given the present state of text-to-speech methods AI, appears throughout the realm of risk.


For individuals who desire a hands-on method to programming, Hello claims to make use of AI to “immediately” reply technical questions from builders with explanations and related code snippets from across the net. The platform is powered by giant language fashions (assume GPT-3) that reference a number of sources to search out the probably solutions, in keeping with co-founder Michael Royzen.

When Howdy customers submit a question, the service extracts and reclassifies uncooked information from Bing’s website, then extracts understanding utilizing the aforementioned patterns. A distinct set of fashions interprets the outcomes into human-readable solutions.


Image credit: Howdy


One other startup with language fashions at its core is NuMind, which gives information scientists, information analysts, and software program engineers with a instrument to create customized pure language processing fashions. Leveraging giant language fashions just like GPT-3, NuMind can be utilized to, for instance, discover job postings that finest match a given CV on a recruitment platform.

NuMind founders Etienne Bernard (the previous head of machine studying at Wolfram Analysis) and Make.org co-founder Samuel Bernard say curiosity within the firm is kind of excessive, with its paying buyer base shifting to 9 inside a month.

Leave a Comment