🗞️ What Makes a Software Company "AI-Proof"?

Happy Sunday! 👋

This week we’re taking a look at the software businesses that have proved the most resilient in the face of AI, and more importantly, the underlying characteristics they have in common.

Let’s dive in!

What Makes a Software Company “AI-Proof”?

It’s no secret that AI has changed the barriers to software development.

Software projects that used to take days or weeks can now be built in minutes with a single prompt.

So in a world where almost any front-end can be replicated, what makes a software business truly defensible? What specific traits prohibit an incumbent from losing business to internal solutions and new competitors?

Our research team has spent the last couple weeks trying to answer this question and we’ve identified 4 shared characteristics.

1. A Network Effect

A network effect refers to when the value of a service is dependent on the total number of other users consuming the same service.

For example, as more users join and share content on Instagram (owned by Meta), the platform becomes more compelling to both existing and potential new users. Once the platform reaches a certain size, it becomes extremely difficult to displace.

After a certain point, the network almost becomes the product itself, and as long as that platform adapts to changing trends (stories, short-form video, etc.), they can resist a lot of competition.

Think about it from a pure product perspective. Replicating Instagram’s design and interface is certainly achievable, but getting 3 billion people to use and post on the platform regularly, that’s virtually impossible.

While Meta’s an obvious consumer-focused example, there are also network effects embedded within many B2B software providers. Slack (owned by Salesforce), Figma, and Autodesk are all great examples. Each of them have become industry standards where cross-company collaboration happens by default.

Take Autodesk for example. Within the architecture space, BIM (Building Information Modeling) data is primarily shared in .rvt files. This is a proprietary file format native to Autodesk’s Revit application. And since competing software can’t reliably open or edit .rvt files without breaking the underlying data, every architect, engineer, and contractor on a project is effectively forced into the Revit ecosystem.

2. Proprietary Data

The 2nd characteristic that can insulate software businesses from AI-powered competition is proprietary data.

While interfaces and designs can easily be copied, data is much tougher. Especially when that data is compiled through human-dependent processes.

One example of this is CoStar Group. While they’ve diversified over time into different sectors, CoStar’s core business is collecting information and analytics on commercial real estate (CRE) properties.

That collection process is still largely manual. CoStar employs thousands of photographers, videographers, and field researchers that travel from site to site collecting visuals and quantitative data on various buildings.

In the case of CoStar, data basically is the product. The same can be said for businesses like Verisk Analytics (actuarial data) and Fair Isaac Corporation (consumer credit).

But there are also instances where software providers can develop a data advantage over time by become the system-of-record for clients. While this isn’t quite as defensible (because the data is often transferrable), clients will usually build workflows around that specific system which makes it more painful to switch.

Some companies that serve as systems-of-record include Salesforce, ServiceNow, and Procore.

3. A Physical Element

Any time a company has a physical layer to its business, it’s going to be less prone to disruption strictly from AI, especially if they’re able to tie the hardware and software together into a unified product.

One example of a company that has done this well is Toast. Can someone use Claude to replicate the menu management or order processing software? Yes, probably.

Can someone use Claude to manufacture a spill-resistant tablet and place it across 189,000 bars and restaurants? That’s a lot tougher.

Another great example of a unified hardware/software ecosystem is Axon Enterprise. Axon got its start selling Tasers to law enforcement, but over time, has expanded into body cameras and ancillary software services like Evidence.com (their digital evidence management platform).

Footage from Axon's body cams now automatically upload to Evidence.com, which creates a deeper dependence on both the hardware and the software. An AI startup could easily build a video storage tool, but it can't easily replace the physical cameras already strapped to hundreds of thousands of police officers worldwide.

4. Reputational Advantage/Liability Shielding

When the potential cost of software failure is catastrophic, companies tend to go with established, reputable vendors.

The most obvious example here is cybersecurity. If a company gets breached while using the “cheap” security tool, the executive who selected that vendor is likely going to lose their job. But if you’re using the most reputable providers like CrowdStrike or Palo Alto Networks, in a way, you’re shielded from liability.

When a client issues an RFP (Request for Proposal), it weighs the potential cost savings of a new tool against the massive risk of an unproven vendor. For most enterprises, saving 30% on a software contract isn’t worth risking a regulatory fine or an operational shutdown.

Beyond cybersecurity, this dynamic also plays out across almost every high-stakes industry. Autodesk for architecture, Dassault Systems for mechanical engineering, and Veeva Systems for drug development are a few obvious examples.

When a single software glitch can delay an FDA approval or compromise a building’s structural integrity, companies tend to pay up for accountability.

That’s all for this week.

If there are any characteristics that we might have missed, let us know! Or if you have any questions about Fiscal.ai, simply reply to this email!