Emerging Wedges in Vertical AI Startups
On voice, search, unstructured data parsing and content generation
This is a weekly newsletter about the business of the technology industry. To receive Tanay’s Newsletter in your inbox, subscribe here for free:
Hi friends,
The past few years have seen the rise of numerous vertical AI application startups. These companies have gained traction quickly by using AI to solve pain points often not addressed by existing software. I’ve observed four common "AI wedges" across many of these startups, which I’ll discuss further in this piece:
Voice
Unstructured data parsing
Verticalized search
Content generation
These wedges1 have served powerful entry points into a number of industries, enabling Vertical AI startups to gain traction and insert themselves in a position from which they can broaden impact over time.
Voice
Voice has primarily been used as a wedge to automate conversational interactions, making it a critical tool for businesses that deal with high volumes of customer communication. By using voice AI agents for handling inbound and outbound calls, startups can sell a system to various businesses that performs tasks once handled by human agents, such as answering inquiries, routing calls, or booking appointments. This improves efficiency, reduces costs, and ensures 24/7 availability.
Common Workflows
Support: Handles inbound inquiries, appointment scheduling, and routine support tasks in customer service and experience workflows.
Sales: Automate lead follow-ups, qualification, and outbound call campaigns in sales.
Back Office: Performs back-office automation for tasks traditionally handled via phone calls (e.g., healthcare insurance calls).
Industries
The most prominent adopters of voice AI include customer service-heavy industries such as retail, finance and various SMBs across industries. Healthcare also leverages voice AI for patient interaction and administrative tasks, while logistics companies use it to manage supply chain communication.
Example Startups
Slang: Automates conversational interactions and appointment booking for restaurants and physical stores.
HappyRobot: Automates phone interactions for logistics companies using voice AI, enhancing efficiency and providing support for customer inquiries.
Assort Health: Streamlines patient engagement with voice AI, automating appointment scheduling and health inquiries to improve accessibility and the overall patient experience.
There have also been a number of horizontal startup leveraging this wedge including Bland and Thoughtly.
Unstructured Data Parsing
Unstructured data parsing involves parsing various forms of files (audio/video/documents/web pages) and typically extracting data from it / structuring that data. Startups leveraging this wedge convert unorganized and at times inaccessible data into structured, actionable insights and typically automate tasks such as note-taking, data extraction, data entry which humans were doing. This allows businesses across industries to improve accuracy, accelerate workflows, and reduce reliance on manual data entry.
Common Workflows
Transcriptions: Transcribing virtual or in-person meetings and structuring notes/follow ups2
Data Extraction: Parsing documents such as invoices, contracts, claims and extracting out key fields to streamline workflows
Data Entry: Scraping web data to automate manual data entry style workflows
Industries
We’ve seen startups leveraging a version of this wedge across most industries, with transcription being particularly common in healthcare, and document parsing being used a lot in various financial, real estate and paper-based workflow industries.
Example Startups
Ambience and Abridge: AI-based medical scribes which listen to patient interactions and take notes and write back to systems of records in structured formats.
Vic.ai: Automates invoice processing for finance teams using AI, enhancing efficiency and accuracy in accounts payable across industries
Raft: Streamlines document processing and workflow management for logistics companies to enhance efficiency and reduce manual tasks
Clay: Not vertical focused, but one of the poster children around web-scraping and data structuring for a specific use case (sales prospecting)
Verticalized Search
When ChatGPT launched, many roles realised that having a version of ChatGPT with the unique datasets they care about (both internal and external) could greatly help them speed up research and decision-making workflows. Vertical search (usually on steroids with chat based question answering) emerged as a natural wedge. The focus for startups here is on extremely high-quality domain-specific information retrieval coupled with access to the important datasets in that vertical. One note is that over time, this wedge may morph a bit with unstructured data analysis and content generation (covered below).
Common Workflows
Research: Improved Research workflows by enabling professionals to sift through large volumes of unstructured qualitative and quantitative data.
Industries
Naturally, the legal and finance verticals are the roles where we’ve seen this the most given the volume of data that professionals sift and research through, although there are specific roles more generally where similar tools may also be applicable (support, etc). In addition, we’ve also seen the horizontal search tools such as Glean continue to gain strong adoption.
Example Startups
Harvey: Simplifies legal research through an AI-powered platform tailored for law firms.
Rogo: Empowers investment banks and private equity firms with AI-driven search on internal and external data, streamlining research and decision-making in the financial services industry.
Content Generation
Content generation was probably the number one early use case for Generative AI, with companies such as Jasper, Writer and Copy helping businesses create content for marketing purposes with AI. This was soon followed by various forms of personalized content generation of emails in the context of sales. But content generation in vertical AI can take a pretty different form as well, as it relates to generating specific documents or assets needed in vertical specific workflows.
Common Workflows
Marketing asset/website generation: Generating websites / product or marketing photos for a given vertical
Report generation: Generating specialized reports / assets that are needed within an industry (the latter usually also involves unstructured data parsing, but the generated output is what is being sold)
Industries
The use case around website / verticalized marketing content generation can apply to any industry, and is particularly relevant for SMBs that may not be otherwise able to access these services easily. The specialized reports or assets can apply both to industries such as e-Commerce that are more image and video heavy as well as to industries such as legal or finance where there are very specific reports that need to be created as part of workflows that may not look like a content generation wedge (but ultimately are selling the creation or generation of various reports with a lot of work going on under the hood).
Startups
Topline Pro: Uses AI to automate website building and social media content creation for home service businesses.
Flair AI and Creative Force: Generates product shots and similar visual content for eCommerce and retail brands.
EvenUp: Leverages AI to analyze case documents to create demand packages for personal injury lawyers
Closing Thoughts
These four wedges—voice AI, unstructured data parsing, verticalized search, and content generation—highlight the strategic depth and market potential of vertical AI startups. By addressing sharp pain points with a pointed solution, these companies can insert themselves into various industries and get much faster adoption than in the past.
Over time, most of these startups will have to expand their offerings, perhaps offering many of the AI features discussed across wedges above. In addition, while many work with existing systems of record initially, as they become a central part of their customers stack and get access to large amounts of usually multimodal and usually unstructured data, many may attempt to become a new system of record in their industry. Others may opt to expand the functions they serve within their customers and become sorts of an AI OS for them.
If you’re building something interesting leveraging any of these wedges (or a fifth different one I missed!), I’d love to chat. Feel free to reach out at Tanay (at) wing.vc.
Thanks for reading! If you liked this post, give it a heart up above to help others find it or share it with your friends.
If you have any comments or thoughts, feel free to tweet at me.
If you’re not a subscriber, you can subscribe for free below. I write about things related to technology and business once a week on Mondays.
These wedges have also been proved successful for horizontal startups as well
One could argue this fits under voice, but I prefer to use voice to refer to voice agents, and think of this more around processing unstructured data.