Disrupting systems of record: The generative AI opportunity

This current wave of generative AI is both exciting and noisy for founders and investors alike. Exciting in the sense that a new general purpose capability has been unlocked, and noisy in the sense that now it’s up to everyone to figure out what to do with it - what use cases does it enable that previously couldn’t be tackled? 

For Square Peg, we have the benefit of seeing across multiple geographies that we invest in (Israel, ANZ, SEA) and the use cases founders are exploring in different regions are truly fascinating. One particular area of interest to us is founders that are leveraging this AI capability to disrupt systems of record.

If you’re building in this space or know anyone who is, we’d love to hear from you. Reach out at lucy@squarepeg.vc. 

What is a system of record and why is it so lucrative?

A system of record is the key place where mission-critical data is stored for a company. There can be multiple systems of record in a company for different purposes. For example, the sales team will orient around their CRM and the HR team will have an HRIS. 

Because of how mission-critical they are, they tend to be very sticky products. The ones that successfully make it into the enterprise segment become enormously valuable companies - today, Salesforce is worth ~$200B and Workday is worth ~$60B (as at 1 September 2023).

What happened to systems of record in the cloud and mobile era?

In previous platform shifts (cloud and mobile), we saw the emergence of new systems of record that subsequently dominated their horizontal or vertical category. For example, Workday for HRIS, Salesforce for CRM, Procore for construction, Wisetech Global for logistics.

Taking Workday as an example, it was founded in 2005 when the shift from on-prem to cloud was beginning to gain momentum. At the time, the HRIS/ERP market was dominated by a few legacy players like Oracle. But many of those players had stopped innovating, banking on the stickiness of their on-prem products. Even if they wanted to innovate, they were handcuffed by the complexity created by growing through bolt-on acquisitions over the years.

Workday saw this opportunity to build native in the cloud, embracing an architecture that allowed them to move fast and at low cost. As a result, they were able to pass $1B in revenue in their first 10 years. 

Systems of record also exist within particular industry verticals and these too have changed. Procore’s rise in construction was largely enabled by the shift to mobile. It was founded in 2002 but its growth really began in 2012 as mobile devices and Wi-Fi proliferated on construction sites.

We’re excited by this new platform shift to AI for two reasons:

1. It opens up new opportunities to create a system of record where one couldn’t exist before. For example, the 1.9T automotive parts market is incredibly opaque - you may not know that the headlight for your Ford might also fit a Jaguar. But our portfolio company Partly is able to use AI to map all the parts in the world so you can always find the right part for the best price.
2. It creates vectors of attack for AI native startups to displace incumbent systems of record.

Of course, incumbents have strong advantages in distribution and data - many are not resting on their laurels as they have quickly embedded generative AI features into their products. For example, Salesforce’s Einstein GPT which can do copywriting and sales emails for you. However, we believe that many will be unable to produce entirely new product experiences enabled by generative AI from the ground up as doing so would mean cannibalising their existing business - the classic innovator’s dilemma.‍

How will AI-native startups disrupt systems of record?

We have spoken to a few founders in this space and some consistent themes emerged.

1. Find a vector of attack that is hard and valuable‍

We often see AI pitches focused on problems that are relatively easy to automate and proximate to almost everyone. But these startups pitching ‘Slack summarisers’ or ‘LinkedIn comment and post writing’ are then faced with existential risk in the space of a few months when the incumbent platforms themselves release it as a feature. 

When it comes to disrupting systems of record, founders need to look for stubborn workflows at the core of heavy administrative burdens. One sign that you’re in the right place is that there was no way for customers to solve the problem before without enormous amounts of admin. Another might be that the entire workflow doesn’t neatly fit into a quick automation (yet) - perhaps AI agents can carve off 80% and leave the human to do quality assurance on the remaining 20%.  

2. Find a vector of attack that gives you access to important data

If you have access to the unit of data that is most valuable to the customer and you’re able to store that information in your product, that gives you a strong starting position from which to challenge incumbents. 

While it seems that many existing systems of record already house all the critical data, many industries still operate via emails, PDFs, spreadsheets getting passed around and a human pulling out the key information manually and inputting it into the system. Similarly with CRMs, it often involves a human manually trawling through the web to prospect for leads to input into the system. These are all user experiences that generative AI is tailor-made to disrupt. 

3. Start with an under-resourced, underserved segment.

Although Workday started and stayed in the enterprise segment, replacement cycles for enterprise are long and you might not be in a period where buying decisions are being made.

Being ready for enterprise also requires product maturity in areas like cybersecurity and features that enable mass distribution. For example, if you click a button, the feature needs to work for thousands of employees as opposed to tens or hundreds. It requires a go-to-market arm that understands how to navigate complex procurement processes.

Many, like Procore and Wisetech Global, didn’t start there and instead chose to target the mid-market or below, then move upwards. The advantages of doing so are 1) faster buying decisions 2) a less complex migration (or no migration at all) 3) more time to iterate towards product maturity 4) a deeper need for automation as unlike enterprises, they’re too under-resourced to throw more bodies at the problem.

Opportunities abound in this era of the AI platform shift. While some value will accrue to incumbents, we are bullish that a sizeable amount of the value will also accrue to startups that are unburdened by existing tech stacks and workflows.  

We welcome your thoughts on the above and if you’re building in this area - come talk to us about what you’re doing!