Vertical AI Agents Could Be 10X Bigger Than SaaS

As AI models continue to rapidly improve and compete with one another, a new business model is coming into view - vertical AI agents. In this episode of the Lightcone, the hosts consider what effect vertical AI agents will have on incumbent SaaS companies, what use cases make the most sense, and how there could be 300 billion dollar companies in this category alone.
0:00 Coming Up
1:01 Jared is fired up about vertical AI agents
7:25 The parallels between early SaaS and LLM’s
9:09 Why didn’t the big companies go into B2B SaaS?
12:25 How employee counts might change
16:25 The argument for more vertical AI unicorns
21:31 Current examples of companies/uses
35:22 AI voice calling companies
40:04 What is the right vertical for you as a founder?
41:36 Outro

[chapterme]

Chapters (Powered by ChapterMe) -
00:00 - Coming Up
01:01 - Jared is fired up about vertical AI agents
07:25 - The parallels between early SaaS and LLM’s
09:09 - Why didn’t the big companies go into B2B SaaS?
12:25 - How employee counts might change
16:25 - The argument for more vertical AI unicorns
21:31 - Current examples of companies/uses
35:22 - AI voice calling companies
40:04 - What is the right vertical for you as a founder?
41:36 - Outro

[Oleg-Melnikov]

I loved the point by Jared about 3 major types of internet companies:
1. products/services that people obviously need: mail, search, maps, messengers, ... - dominated by the largest corps
2. not obvious consumer products - AirBnb, Doordash, Uber, Instacart, ... - born from startups that were more agile and creative
3. B2B SaaS - multiple verticals, not dominated by large corps
it kinda creates a picture for me

[judederick]

3 things to vertically differentiate
1. Vertical agent has exclusive access to high value dataset . Ex Corrosion detector agent has exclusive access to worlds largest refinery corrosion dataset

2. Vertical agent owns a key decision making node in workflow. Example : Next best action decision in equipment troubleshooting buddy sgent

3. Vertical agents own end to end workflow. Ex : Regulatory reporting agent in Pharma

Trust that helps

[MarkLewis]

I have created a ton of agents using openai, crewai & composio - how should I sell it?

[DTrackFinder]

It's wild to see how often you, the Top Startup Advisors, are shifting your perspective and advice lately—it’s a real testament to just how fast and profoundly things have been changing these past couple of years. Feels like the ground under everything is constantly shifting!

[MAureliusHiggs]

Loved this conversation! Two takeaways that stand out for me are:
1. The focus on founder-led teams at Rippling suggests that domain expertise and entrepreneurial thinking remain crucial even in an AI-driven future. Collaboration, Critical Thinking, Creativity and Communication will still be the foundation of how humanity progresses regardless of the tool.
2. The most successful AI startups might not be those with the best technology, but those with the deepest understanding of boring contextually-specific workflows.

Two questions:
1. What does these 'new rules' look like decentralized?
2.With World Coin on a mission to bring a billion+ new 'players' into the game, how might we empower these traditionally underserved global communities?

What a time to be alive!

[Penrose]

The lump of labour "fallacy" is one that the tech community loves. "Don't worry about job automation today because these are simply yesterdays jobs. Embrace tomorrows jobs today" they say. Except, no one stops to consider any potential retooling time for the economy. It might well be that there will be new jobs "tomorrow". It's just that today there are millions of people waiting for that tomorrow- and it isn't coming anytime soon. What to do in the interim, hmmm? When will people wake up to see this I wonder ?

[yousefsuliman]

Umm, I'm going to need AI model builders to take a vacation. Come on. Just let us get rich off our apps before you outdate us all a week after launching. Y'all have no chill.

[vijayrangan]

I don't follow the narrative that these YC folks are pushing about boring jobs. If people working those jobs could get something else they would. Making a case for AI removing boring jobs just so it "allows" the current employees to leave is just nonsensical

[pjm]

Yeah, but what is preventing SaaS companies from just pivoting their tech stack or integrating vAIs into their offering? Like trying to sell into these categories is extremely hard, even if the tech is cheaper, faster, better. These customers are exhausted.

[jjdorig]

Light Cone Podcast - Vertical AI Agents Discussion Timestamps

Introduction & Historical Context (0:00-4:04)
- 0:00 - Discussion of progression in AI capabilities over 3-month periods
- 0:05 - Introduction to vertical AI agents replacing enterprise teams
- 0:17 - Evolution from OpenAI dominance to competitive landscape
- 0:44 - Introduction of hosts Gary, Jared Harge, and Diana
- 1:04 - Jared's enthusiasm about vertical AI potential
- 1:28 - Prediction of $300B+ companies in vertical AI category

SaaS Industry Analysis (4:04-9:09)
- 4:04 - Historical context of SaaS evolution
- 4:44 - XML HTTP request as catalyst for SaaS boom
- 5:41 - Three categories of successful SaaS companies:
1. Obviously good mass consumer products (won by incumbents)
2. Non-obvious mass consumer ideas (won by startups)
3. B2B SaaS companies (300+ successful companies)
- 7:53 - Discussion of why incumbents didn't enter certain markets
- 8:47 - Analysis of Uber's early regulatory risks

B2B SaaS Deep Dive (9:09-12:28)
- 9:09 - Analysis of why incumbents didn't dominate B2B SaaS
- 10:34 - Discussion of bundling vs. unbundling in software
- 11:35 - Enterprise software pricing models
- 11:51 - Problems with traditional enterprise software

Employee Count Changes (12:28-16:25)
- 12:28 - How AI might change company employee scaling
- 13:02 - Current unicorn employee counts (500-2000 employees)
- 13:31 - Shifting hiring priorities with AI
- 14:43 - Historical perspective on company building
- 15:45 - Example of engineering approach to marketing

Vertical AI Examples & Analysis (16:25-24:22)
- 16:25 - Case for 300 vertical AI unicorns
- 17:01 - Discussion of enterprise uncertainty about AI needs
- 18:30 - Consumer vs. enterprise adoption patterns
- 19:50 - Why vertical AI could be bigger than SaaS
- 21:37 - Real examples of vertical AI companies
- 23:53 - Case study of QA testing automation

Customer Support & Specialization (24:22-29:01)
- 24:22 - Analysis of AI in customer support
- 25:51 - Example of developer support automation
- 26:30 - Discussion of Power Help case study
- 27:42 - Importance of vertical specialization
- 28:38 - Why specialized solutions win over general ones

Management & Scale (29:01-35:22)
- 29:01 - Discussion of Coase's theory of the firm
- 30:02 - Rippling case study
- 31:06 - AI's impact on Dunbar's number
- 32:16 - Example of AI-powered employee communication
- 33:43 - Rippling's organizational structure

Voice AI & Industry Evolution (35:22-40:04)
- 35:22 - Discussion of AI voice companies
- 36:17 - Case study of Salient in auto lending
- 37:31 - Evolution of voice AI capabilities
- 38:11 - Historical perspective on LLM applications
- 39:23 - Rate of AI progress analysis

Closing Thoughts (40:04-41:53)
- 40:04 - Advice for founders on choosing verticals
- 40:19 - Importance of finding boring, repetitive admin work
- 41:34 - Final analogies to robotics
- 41:47 - Closing remarks about targeting "butter passing jobs"

[adamsonbanda]

y'all feel how yc just drops gold everytime for those who are so far out on the other side of the world

[DarrenMegaHR]

A new OSI model has emerged, built atop the old. Agents exist at the Application Layer with LLMs/transformers as TCP and ML as IP.

[ljli]

Everyone in this discussion is sooo articulate 🤯

[ThinkAIst]

Hunting for the boring repetitive jobs should be every AI startups focus. A friend and are building a basic MSP and have locked on with a client who employs many people to perform basic repetitive services. We get to look at this clients work efforts through our AI lenses and the number of opportunities we have categorized is a bit amazing. When time permits we are building out our flavor of an industry vertical agentic solution that we feel applies to multiple vertical service types of SMBs.

[fromduskuntodawn]

Should this be titled: How to bypass complainers to replace their jobs without their approval?

[gottahide]

25:50 what's the name of company she mentioned? 😅

[fromduskuntodawn]

This does sound like a dystopian nightmare tbh.

[jacobsingh]

I started coding before AJAX, that's not what caused SaaS to happen. It was one of many improvements. The main thing was just the web and dynamic scripting languages which enabled it.

[aga]

What exactly does it mean to do" vertical " integration. Why not top-down, diagonal? Native ? 360 degree ? Genuinely asking ...