Enterprise trends from YC W23 Demo Day

Recently, the?YC Winter 2023 ?batch presented their new businesses to an invitation-only group of investors, press, and others in the startup community. The two days included 265 startups, who were selected by YC from a pool of 20,000 applicants.?

In this post, we look to highlight the major themes and trends in the batch - so as to ascertain what might be next for enterprise tech. YC has an outstanding track record of success, with 300+ companies in its 18-year history valued at $150M+ and 90 companies valued at $1B+.

Before we begin, we would like to congratulate all YC W23 founders! It was a privilege to participate in Demo Day and to listen and learn about your teams, technologies, markets, and business models. Congratulations to YC W23.

In the below chart, we map the number of YC companies by batch over the past 10 years and the percentage of those companies that are tagged as B2B by YC. The W23 batch saw a remarkable increase in the focus on B2B, which kept us busy over the two days and which speaks to the scale and breadth of the opportunity in enterprise tech.

As with previous posts, we divide this post into two sections: 1) our favorite quotes and 2) the key themes and company descriptions.?

1) Favorite Quotes

The following quotes struck us as memorable. Although we may agree or disagree to varying levels, they point to potential problems in the world that are yet to be solved.

‘Browsers can be much more than passive input boxes, and browsers have been single player for too long.’ -?LinkGrep

‘The core of GitHub hasn’t changed in over 10 years, and during this time Figma, Notion, and Linear have modernized every other piece of the product development toolkit.’ -?Pierre

‘Zendesk built a $10 billion business on email tickets, and email tickets are going away.’ -?Unthread

‘2 out of 3 Figma users aren’t designers, i.e. they are using the wrong tool.’ -?Noya Software

‘New database platforms such as PlanetScale and Neon offer game-changing features such as change management, branching, real-time events, and instant rollbacks; however, the only way to use these platforms is to migrate your entire databases.’ -?Inquery

‘Previously B2B companies could onboard customers by connecting to Segment, but now they can’t since event data has moved to the data warehouse.’ -?Fabra

‘Quickbooks was built 40 years ago for small businesses that go through transactions one by one and for bookkeeping firms that price high but hire cheap.’ -?Porter

‘We will never, ever, ever use CoPilot or anything else that sends our code to a third party.’ - customer conversation with?CodeComplete

‘OpenAI models are slow because they are generalized. When trying to solve a specific problem, you can train smaller models that perform just as well but perform dramatically faster.’ -?Meru

‘The locked-down LLMs are more concerned with AI safety than with developer experience.’ -?Anarchy

‘As the data format of ML, embeddings are critical to AI applications, but there are no tools to help developers manage them.’ -?Metal

‘While developers don’t yet want AI tools writing their most important code, they do want AI tools that can quickly build tabestakes features.’ -?Second

‘Software architects at large companies are changing from ivory tower decision makers to collaborators, and generic whiteboarding tools just don’t cut it for the software system complexity.’ -?IcePanel

‘Traditional RPA tools have very low adoption rates. At one accounting firm each employee is given a $6k RPA license that less than 5% use.’ -?Squack

‘MobileEye, a single autonomous vehicle developer, manages nearly as much data as all of Snowflake’s customers combined.’ -?SensorSurf

‘B2B sales teams spend 50% of their time on pre-sales processes.’ -?Intently

‘Personalized sales emails take 10 minutes to do, and most sales reps do not do it.’ -?Coldreach

‘A $10 million construction project that takes 18-24 months involves 18,000 invoices and 1.8 million invoice line items.’ -?inBuild

‘Average home service business misses 62% of inbound phone calls.’ -?Sameday

‘Every day SMBs miss 50% of their calls and most of the times the customer calls a competitor.’ -?Avoca

‘Most tech companies have no visibility into their per-customer or per-product costs, and the need is greater than ever given the move to costly LLMs and usage-based pricing.’ -?Rosettic

‘Biotech companies use dozens of software tools to manage trials but they don’t walk to each other.’ -?Miracle

‘The current software stack for opticians and optical labs is stuck in the 90s.’ -?SpecCheck

‘Nobody ever thinks about car wash software.’ -?FlexWash

2) Key Themes and Company Descriptions

Generative AI tools

Large language models provide a robust opportunity to rethink the tech stack, and YC W23 startups included platforms and tools across LLM infrastructure layers.

LLM infrastructure.?Anarchy ?is developer-first LLM infrastructure to provide companies with their own LLMs and move away from LLMs with locked-down APIs.?Chima ?provides infrastructure for enterprises to customize and scale generative AI.?Baseplate ?is backend-as-a-service for LLM apps.?Stack AI ?provides a platform to compose and fine tune LLM applications, combining a drag-and-drop visual interface with a backend that connects vector databases, language models, and document loaders.?Vellum ?is the developer platform for production LLM apps.?

Model training.?Rubbrband ?provides open source software that removes friction in fine tuning generative AI models.?

Custom models.?Meru ?trains task-specific, custom-built platforms that work as well as ChatGPT but run as much as 40x faster.?

Observability.?Helicone ?is an open source observability platform, helping generative AI companies log requests and track costs, result quality, and latencies for providers like OpenAI.

Automation.?Lasso ?is an RPA tool for Chrome that uses ChatGPT and computer vision, enabling operations managers to create automations using plain text or recorded video.

Personalization.?BerriAI ?provides SaaS businesses an API to create ChatGPT apps that are personalized to users.?

Data assistant.?Defog.ai , the ChatGPT for data, is an AI data assistant that is embedded in apps and lets users ask questions about their data.

Production.?Wild Moose ?helps on-call developers solve production incidents with generative AI trained on observability data from their environments.?Traceloop ?trains LLMs on OpenTelemetry production system data to generate tests and detect anomalies in production.?

Generative AI applications

LLM-based startups sought to reimagine a plethora of horizontal and vertical applications.?

Marketing.?Type ?is an AI-first document editor and workflow creation platform for marketing teams.?SpeedyBrand ?is a generative AI powered tool for marketing content for SMBs.

Gaming.?Iliad ?is a creative tool for generating concept art, 2D, and 3D game assets, letting designers step through and edit evolution of assets to have fine control of final output.

3D design.?Dream3D ?is an AI-powered 3D design tool, enabling designers and artists to create, render, and refine 3D images and videos by natural language text.

Market research.?AiFlow ?uses LLMs to automate market research, starting with customer reviews and news.?

Workflows.?Waveline ?automates workflows using LLMs, classifying, extracting, and handling specific information from unstructured data such as emails or documents.

Search.?Gloo ?provides custom embeddings to enable semantic search for domain-specific enterprise data.?

Knowledge base.?Tennr ?turns a company’s knowledge base into a custom LLM, that can answer nuance questions and produce briefings and reports for employees.?

Photography.?Booth AI ?is a generative AI photographer for eCommerce.?

Call centers.?Flair Labs ?engests unstructured call center voice data with LLMs to uncover customer insights.?

Accounting.?Squack ?gives accounting firms an RPA tool that uses LLMs to create complex automations in natural language.

Insurance.?Kyber ?is automating the insurance industry with LLMs, focusing on tedious jobs such as customer support questions, underwriting information review, and claims intake management.

CoPilot for X

Inspired by GitHub CoPilot, numerous YC W23 startups aimed to build the “Copilot” for their target industry, business function, or business process.

Workflows.?Extend ?is a developer tool for building internal Chrome extensions that are custom copilots for teams’ internal workflows.?

Code.?CodeComplete , the “GitHub CoPilot for the enterprise”, provides AI-powered code completion that is self-hosted to offset enterprise security concerns.

Browser.?LinkGrep ?is an AI browser copilot for teams, providing automated content suggestions from teams’ knowledge bases and enabling collaboration through notes and chat.

Sales.?Persana AI ?provides a sales copilot powered by fine tuned models, identifying high-converting prospects, generating actionable insights, and enabling messaging at scale.

Healthcare.?Fairway Health ?is a “Copilot for health insurers” to authorize patient treatments, analyzing long messy medical records to determine treatment approval.?

Hedge funds.?AlphaWatch AI ?provides an AI copilot for hedge funds, solving LLM hallucinations by integrating trusted live data and internal docs.

Retail.?Luca , the “Pricing Copilot for Retailers”, provides an AI-based pricing system for e-commerce businesses.

Sales and marketing

Sales and marketing remained a major theme in YC W23.?

Pre-sales.?Intently ?provides an AI tool to help sales teams identify the best targets and craft perfect messages that match the future customers’ intent.?

CRM.?Lightski ?provides an AI-powered sales assistant to automate tedious sales processes and update CRMs from call records and Slacks.?Paperplane ?updates Salesforce automatically after sales calls and enables sales leaders to instantly understand what’s going on in their pipelines.?

Personalization.?Coldreach ?provides AI-powered sales email personalization at scale.?

Sourcing.?Sail , “sales on autopilot”, uses generative AI to replace SDRs and to find and contact high quality sales leads.?

Demos.?Mirio , the “Webflow for sales”, provides a no-code editor for sales teams to create personalized demos without needing to request engineers’ time.

Call intelligence.?Zenfetch ?is a real-time sales call assistant, accessing insights and providing talking points across GTM silos such as internal docs and prior calls when customers ask questions.

Emails.?Rift , the “Rippling for Go-To-Market”, help companies send cold outbound emails without landing in spam.

Reselling.?Bluebirds ?helps sales teams to sell to past customers who switched jobs.

Influencer marketing.?1stCollab , the “DoubleClick for Influencer Marketing”, helps brands instantly work with hundreds of relevant, cost-effective micro-influencers.

Customer support.?Yuma ?is the “ChatGPT for customer support”, targeting Shopify merchants.?Unthread , the “Zendesk for Slack”, is a customer support platform natively built for Slack, Teams, and Discord.

Cloud marketplace.?Suger ?helps B2B companies sell through the cloud marketplaces at AWS, Azure, and GCP.

Product

YC W23 innovations for product teams included designs, features, integrations, collaborations, and user interface components.

Product design.?Noya Software ?is a product design tool for non-designers, letting everyone design like a designer.?Mirrorful ?is an open source design system framework to manage apps’ building blocks.

Product features.?Cardinal ?enables product teams to continuously prioritize and build product features, combining signals like CRM records or BI/analytics success metrics.?

Product integrations.?Nango ?is an open source platform for product integrations, with 60+ supported APIs currently.

Engineering collaboration.?IcePanel ?is a collaborative diagramming tool that helps engineering and product teams align on technical decisions, creating a full context map of complex software systems.

UI.?Blocksfabrik ?automatically generates user interface components such as colors, buttons, and inputs using companies’ design guidelines.

Voice

Numerous YC W23 startups focused on voice–with the goal to provide real-time AI bots that users can actually talk to. Applications included phone receptionists, customer support representatives, personal assistants, sales assistants, and tutors, among others.

LLM applications.?Vocode , the “Langchain for voice-based LLM apps”, provides an open source voice AI platform on top of LLMs, enabling real-time AI that you can actually talk to for applications such as phone support, personal assistants, sales assistants, and tutors.?

Home services.?Sameday ?is an “AI that is trained to sell” and automatically answers calls and books appointments for home services businesses such as plumbers and roofers.?

Voice generation.?Play.ht ?generates or clones voices from as little as 30 seconds of training data and turns any text into human-like speech–with the biggest applications in media production, gaming, and eLearning.

Fast food.?OfOne ?is an AI-powered order taker for fast food drive-thrus leveraging conversational AI.

SMBs.?Avoca ?provides an AI receptionist for SMBs.?

Data

Data parallels in the YC W23 batch included the “Snowflake for real-time data and for sensor data”, the “Fivetran for LLMs and for log data”, and the “GitHub for data teams.”

Modern features.?Inquery ?is an open source platform that provides superpowers for users’ existing SQL databases, starting with collaborative data changes and real-time events.

Data pipelines.?Arroyo , the “Snowflake for real-time data”, is a serverless platform for real-time data pipelines in SQL, making streaming as easy as using batch data warehouses.?Turntable ?enables less technical analysts to build their own data pipelines.

Customer data.?Fabra ?enables B2B companies to build integrations with and ingest data from customers’ data warehouses.

Interface.?Outerbase ?is the interface for databases, enabling users to view, query, and edit data and to generate dashboards without writing SQL.

Shared data.?Sherloq , the “Github for data teams”, is a source of truth for SQL queries to ensure answer consistency for different analysts.?Honeydew ?is the shared source of truth for data teams, to make sure that key business metrics such as revenue, churn, and active users are always consistent.

Sensor data.?SensorSurf , “the Snowflake for sensor data”, manages petabytes of data from cameras and LIDAR in cars, robots, and IoT devices.?

Homomorphic encryption.?Blyss ?is a leakproof data warehouse that lets companies store and analyze sensitive data without decrypting it.

ETL.?Sidekick , the “Fivetran for LLMs”, makes it easy for companies to sync data from SaaS tools such as Confluence to vector databases.?Vaero , the “Fivetran for log data”, enabling enterprises to store logs in their own S3 accounts rather than expensive specialized log storage.

Integrations.?Lume ?is an AI tool to generate and maintain custom data integrations.

ML tools

Many YC W23 companies innovated in machine learning tools for developers, across semiconductor, software, and edge platforms.

Model deployment.?PoplarML ?provides deployment infrastructure for ML models, enabling production to scale with serverless GPUs in one command. Chart, the “GitLab for ML”, makes it easy for companies to deploy high performant ML models in their own infrastructure.

Hardware-accelerated APIs.?Texel.ai ?enables developers to run AI pipelines 10x faster with the latest NVIDIA GPUs and Google TPUs without having to become hardware experts.

Federated learning.?Flower ?provides an open source platform for training AI on distributed data through federated learning.

Monitoring.?UpTrail , the “DataDog for ML”, is an open source tool to monitor ML models in production and identify failure cases to improve upon.?BabylonAI , the “DataDog for ML on edge devices”, extracts the most valuable data from edge devices and estimates model performance.

ETL.?Dagworks ?provides an open source ML ETL platform for data science teams, enabling junior data scientists to write and maintain production code.

AI features.?Pyq ?makes it easy for developers to build features powered by AI, identifying and providing APIs for specific engineering tasks that AI solves well.

Embeddings.?Metal ?provides ML embeddings as a service.

Developer tools

Developer tools, even beyond the newer LLM and ML developer tools discussed above, was a major theme for YC W23.

Automations.?Trigger.dev ?is the developer-first open source alternative to Zapier to create workflows easily in code.

APIs.?SuperAPI ?provides a programmable API gateway that makes APIs cacheable, reducing load times by at least 10x.?Fern ?provides API automation to build APIs twice as fast.

Developer bots.?Second ?provides autopilot developer bots that can build web application features.?

Background processing.?Defer ?provides a zero infrastructure background processing platform, replacing home-made solutions and starting with Node.js market.

Alert management.?Keep ?provides open source alert alert management for developers, using AI to surface alerts that matter.

Web apps.?Pynecone ?allows developers to build Web apps in pure Python and deploy with a single command.?Atri Labs , the “Vercel for Python developers”, provides an open source Web framework for Python developers to develop production applications.

Backends.?Fastgen , the “Webflow for backends”, provides a low-code tool to build APIs, workflows, and complex backends.?Keyval ?provides an open source codeless monitoring tool.

Cloud ops.?Rosettic ?breaks down AI and cloud costs by customer and feature, providing the financial infrastructure for AI.?Cloudchipr , the “Zapier for CloudOps”, provides an automation platform for cloud operations.

Emails.?Resend , the “Stripe of email”, provides an API for developers to build, test, and send transactional emails at scale.

Observability.?Middleware ?is an AI-based cloud-native and microservices observability platform.

Debugging.?Nebula ?provides an observability and debugging platform for large enterprises, enabling developers to see how any variable or function is used across the entire codebase.

Search.?Buildt ?provides an AI devtool to search and understand large codebases, enabling developers to ask questions in plain English through an IDE extension.

Review.?Planar ?is the “Superhuman for code review” and addresses the one-third of engineering time spent on code review.?

Security

There were a few YC W23 startups in cybersecurity, including both developer and enterprise markets.

Secrets management.?Infisical ?provides open source, encrypted secrets manager for developers.

APIs.?Escape ?provides GraphQL API security testing for developers and security teams.

Passwordless.?0pass ?provides an enterprise platform that replaces passwords and two-factor login with convenient factors like face ID, touch ID, and YubiKey.?

SIEMs.?Matano ?is an open source cybersecurity platform that replaces SIEMs like Splunk with data lakes, analyzing security logs and detecting threats in real-time.

Healthcare

Healthcare remains one of the largest industry verticals in YC batches. YC W23 startups addressed a wide range of opportunities across the healthcare ecosystem.

Independent clinics.?Pledge Health ?automates patient billing for independent healthcare clinics, enabling patient charges at time of service instead of after insurance processing.

Mental health.?Orchid ?provides an AI-powered EHR for independent mental health professionals.?Vitalize Care ?provides mental health coaching and peer support for healthcare professionals.?

Digital health clinics.?Untether Labs ?provides scheduling software for digital health clinics.?

Diagnosis.?Glass Health ?provides doctors with AI-assisted diagnosis and clinical decision-making.?Scanbase ?is the API for diagnostic test analysis, enabling medical companies to analyze photographs of rapid diagnostic tests and return results.

Trials.?Miracle ?helps biotech companies speed up clinical trials by connecting trial data sources into one view.

Drug formulation.?Lavo Life Sciences ?uses AI to simulate drug molecules and predict the drugs’ stability and manufacturing conditions.?Persist AI ?builds long-lasting microsphere formulations and optimizes drug releases using automation and machine learning.

Optical industry.?SpecCheck ?is an ordering, billing, and customer management system for eye doctors and optical labs.?

Hospitals.?Latent ?uses language models to automate medical paperwork for hospitals, beginning with prescription approvals.?

Autism.?Finni Health ?is a platform for autism care providers to go independent and start and grow their practices.

Construction

Construction was a significant industry vertical among the YC W23 batch, and companies included the “Rippling for construction” and the “GitHub for construction.”

HR.?Hammr , the “Rippling for construction”, is purpose-built for project-based payroll, distributed workforce, and compliance.

Blueprints.?Pirros , the “GitHub for construction”, manages all blueprints and enables architects and engineers to reuse past blueprints rather than design each time from scratch.

Finance workflow.?inBuild ?automates invoices for construction companies, verifying and reconciling invoices’ line items against contracts and budgets.

Telecom.?Clad ?provides telecom companies contractor management software to find, manage, and pay subcontractors for projects such as laying fiber or building cell towers.

Accounting and tax

Accounting and tax was another theme at YC W23, driven by the lack of innovation in accounting for decades and by new regulations for country tax reporting and state sales tax collecting.

Accounting.?Porter ?is the modern accounting software, specifically built for startups to easily integrate and automate bookkeeping accurately.?Truewind ?provides AI powered bookkeeping and finance software with an AI bot that automates transaction labeling.?

Taxes.?Invopop ?helps global software companies comply with local country tax reporting.?Numeral ?is ecommerce sales tax on autopilot, helping merchants to collect local state taxes.

Content creators

Several companies in the YC W23 batch addressed needs for content creators, across analytics, audio, video, and NFTs.?

Predictive analytics.?CreatorML ?provides ML-powered predictive analytics to YouTube creators to guide topics, titles, and thumbnails for their videos.?

Dubbing.?EzDubs ?provides a real-time AI dubbing tool for content creators and has a vision to enable any two people who speak different languages to speak together in real-time.

Videos.?Decoherence ?is a tool for content creators, musicians, and studio artists to create and edit generative videos.?

NFTs.?Members.Land ?streamlines launching NFTs and managing communities for content creators.

Other vertical applications

Other verticals addressed by the YC W23 batch included manufacturing, real estate, financial services, and car wash industries.

Manufacturing.?Serial ?is bringing big tech software automation to traditional manufacturers, who are still using legacy systems, spreadsheets, or paper forms.

Real estate.?Homeflow ?is a digital insurance brokerage for large real estate investors to purchase and manage insurance portfolios in a single place.?Propify , the “Plaid for commercial real estate”, provides a unified API for commercial real estate property management.?BidSight ?is modernizing project management for real estate developers.

Financial services.?Hadrius ?is the “AI SEC Compliance Autopilot” for financial firms, automating compliance program and flag compliance failures for review.

Car wash.?FlexWash , the “operating system for the car wash industry”, helps owners manage customer data, track operations, and grow revenues.?

Productivity and other

Other YC W23 startups focused on productivity, legal, and management functions for workers or companies.

Slide decks.?Rollstack ?allows teams to automatically create and update slide decks from BI tools or data warehouses.?

Workflows.?Layup ?is an AI-powered workflow creation tool for small to medium enterprises, enabling employees to complete entire workflows with one command.?Versori ?provides enterprise integration and workflow automation, starting with retail and e-commerce.

Contracts.?Common Paper , the “SAFE for commercial contracts”, provides standard open source contracts for tech startups.

Coworking.?FlexDesk ?helps companies manage their spend on coworking spaces, by replacing underutilized office leases with pay per lease program.

Goals.?Stellar ?is the system of record for company-wide goals, synchronizing goals across departments and notifying executives about risks.

Conclusion

According to Y Combinator, 4% of companies in recent YC batches become $1B+ companies, implying that there are 11 possible unicorns in the YC W23 batch of 265 companies. It will be exciting to see which companies continue to gain momentum and scale over the coming quarters and years. Many congratulations again from Wing to the entire YC W23 cohort.