As I reflect on our application layer investments made over the past few years, I’ve come to believe there are three broads ways you can build an AI business today.
1. Rebuild and rethink the core system of record
AI unlocks an opportunity to rebuild Systems of Record (SOR) in record time. And more than simply rebuilding like-for-like modules to get to feature parity, there is an opportunity to completely rethink the value propositions of certain modules altogether. For example, a customer service module could be completely automated with AI, turning a cost center into a cost saver.
The value of AI coding
This approach to company-building is newly possible because of AI coding. We’re seeing seed stage companies build complex SORs in <6 months with just a handful of (AI-fluent) engineers.
Data migrations
AI also allows for rapid migrations off of incumbent SORs, in a way that wasn’t previously possible. It has resulted in meaningful unlocks when it comes to data mapping, unstructured data parsing and analysis of data dumps - all of which means that you can get customers up and running on your system much more quickly.
Rethink the value prop with AI
This is best illustrated through an example. Our portfolio company Solea is building an AI-first ServiceTitan (SOR in the home services space). They rebuilt the core modules (like CRM, payments, dispatch, routing etc) but also completely rethought some modules. For example:
- AI customer service — using AI to answer every customer service query, across every channel, with 24/7 availability
- AI collections — automating the team chasing up unpaid invoices
- AI route optimization — allowing the technicians to do more routes per day, thus increasing revenues
Even a small home services company has a huge back-office team (across customer service reps, schedulers, invoicing, QA). Solea has rethought many of these roles with AI, saving even SMB companies hundreds of thousands of dollars.
Counterposition
Instead of competing head-on with ServiceTitan (who are focused on enterprise customers and concentrated in particular verticals), Solea is picking off SMB customers who were typically overlooked due to being uneconomical to serve. With Solea’s labor reduction value proposition, those SMB customers are suddenly able to pay substantially more than before, and are now viable.
The playbook: rebuild the SOR, rethink the value prop with AI, counterposition
2. Build an invisible layer on top of the system of record
In some categories, SORs are deeply entrenched, have huge surface area and would require substantial change management in order to get people to move off them. A good example is Workday, which is a core HRIS system that is used by 65% of the F500.
In this case, it often makes more sense to build an AI intelligence layer on top of the SOR and integrate deeply with them. The AI layer becomes the interface that the end user interacts with, and can also end up generating its own valuable data. Moreover, this approach can also unlock co-selling and partnerships opportunities with the SOR which can be useful for distribution.
Our portfolio company Alex, an AI interviewer, has taken this approach. When a candidate applies to a job, Workday is notified and pings Alex, who can instantly screen their resume, interview the candidate and then score them. Alex can also read up on existing employees within Workday and decide if it makes sense to interview them for the same role.
In this way, Alex ends up building an invisible, but highly valuable layer of candidate, interview and employee data.
3. Become an AI-native services firm
The newest approach to company building in the AI of age is to build an AI-first services firm. A few years ago you could have called this tech enabled services (a category VCs hated), but now it’s in vogue. And arguably for good reason, because AI has proven it can automate a bunch of manual workflows (like summarizing documents, sending emails, reviewing files, responding to calls and doing research).
As a result, if there is scope for substantial automation of a workflow, it may make more sense to become a service provider rather than sell software to incumbent service providers. For example, to become an AI-first law firm, tax advisory firm, insurance brokerage, recruiting agency etc.
From a VC perspective, this approach only makes sense if you can substantially increase revenues per employee versus a regular services firm (think 5-10x uplift). Otherwise, you are simply running a services firm, which will always require scaling headcount in order to scale revenues.
Moreover, it requires running a services firm and a technology company fused into one. Culturally, this is a unique beast which many tech-only founders would not be well equipped to run.
An interesting example from our portfolio is Clera, which is an AI-first staffing agency. They use AI to automate outreach, screening and matchmaking - and can simultaneously have tens of thousands of conversations open with candidates, with only very minimal human-in-the-loop. They charge placement fees like a regular staffing agency but have a low operating headcount which shouldn’t need to scale much even if they 10x’d the candidate side of their marketplace.
If this approach works, it will be hugely disruptive to existing staffing agencies because of their superior cost structure (which can itself unlock the ability to undercut on placement fees, which can end up being a very unique value proposition in the market).