Rolling up your sleeves with AI

Satya Nadella famously proclaimed that Agents and AI would collapse SaaS. Specifically he claimed that since SaaS is something like a database plus a layer of business logic, the true value is in the data (which is owned by companies), and the rest will be managed and orchestrated by AI agents. Others have expanded on this idea (both worth a read!).

Of course SaaS companies do more than provide a shiny UI on top of your data, but I’m pretty confident this is the direction the industry is going and it will become more of a reality (e.g. AI bills replacing SaaS bills) over the next 2-3 years.

I also think that - right now - companies are taking the baby step in this direction by rebuilding (commoditized) SaaS apps internally as AI drives down the cost of development. It’s not quite the agentic future but its a step in that direction.

🤌 The longer version

Just to make sure we’re talking about the same thing, here is a simplified conceptual diagram of a typical software application:

Example B2B SaaS apps include:

• Salesforce for managing your leads and prospects and deals

• Deel for managing your global payroll

• MailChimp for marketing automation

• Asana for project management

• and the list goes on, I think technically, forever

To reiterate, other than the data layer, which represents your unique company or customer data², the rest is undifferentiated software that you buy off the shelf. My contention is that AI is making it so much easier (e.g. cheaper) to build these internally, companies will opt to build vs buy³.

Let’s expand on that!

🤔 Remind me what I’m paying for?

Simply put, SaaS is any service that you pay a monthly or annual fee for, rather than building it yourself⁴.

To break it down further, most SaaS apps are essentially databases with some automations on top. When I say databases, I mean (to steal Benn’s language) lists - of people or things you care about. You as a person don’t care much about how these lists are managed or organized, you just want to make sure they are accurate and you can use them.

Beyond the database, these apps have automations and features like customer 360 management, email outreach, calendar notifications, and other key business operations.

At its core, what you're paying for is the ability to extract value from your data, whether that’s managing employee payroll, so people get paid the right amount, or personalizing messaging to your customers to improve their conversation rates⁵, or a plethora of use cases, all centered around your data.

☁️ An example to help illustrate

Let's say you lead the fundraising team of a climate non-profit, and your goal is to increase monthly donations. You believe that with personalized outreach, donors will be more likely to donate more than if they receive a generic message (or, even worse, a message that’s inappropriate to them). To do so, you need to make sure you have the right data points on each individual (or cohort) such as their demographics, economic status, past donations, etc.

To do this, you'll need to collect and store donor info in a database (lists!), integrate it with your email outreach system, and make sure the other folks on your team have a simple UI to navigate, and you want these systems to work accurately and be fast.⁶ 

If you're a non-profit focusing on climate, it's probably a better idea for you to spend your resources strategically where your expertise is (e.g. climate initiatives, fundraising events) rather than on creating and managing software.⁷ 

So, you go out and buy Salesforce. They do this for 1000s of companies, and have been doing it for 25+ years. They are the industry leader, so yeah, it's a no brainer. The cost of standing up your own software is much higher than buying Salesforce off the shelf, even though your all in cost is still more than the SaaS fee.⁸

🤖 How does AI change this picture?

We sort of assumed it was self-evident why it doesn't make sense to build the software yourself. A little more specifically - you don't have the software development expertise to make this worth the cost. Hiring people, training them, investing in the resources they need to be successful, and doing this year after year is a huge expense.⁹ Beyond the strategic reasons why this doesn't make sense, the cost is just too high to justify it, especially when you're trying to build commoditized capabilities like sending targeted emails. Up to now, why bother?

Here's the rub - in an age where software development is getting faster and cheaper, the bet is that you can hire fewer engineers, who can do more with AI, to make the cost of building a solution internally worth it.¹⁰ 

After all, these applications, are databases (you still need to collect and store data) and some automations (which is sort of what Agentic AI is going to be used for). If the cost curve is shifted enough to the left, and now your team can rebuild Salesforce, but this time, you can rebuild it without all the premium features you never use¹¹, AND you can build custom features (e.g. specific fundraising lead scoring, or integrations with a random tool you really like, or custom designed outreach not currently possible with your SaaS tool), it makes the build vs. buy conversation worth re-interrogating.

🌊 Breaking of the dam?

Maybe not for every single SaaS app, but definitely for some of them. And when more and more companies start rethinking this, there will be more open source solutions (lots of Salesforce copies, built using AI), and the new generation of Software / AI Engineers will be more comfortable rebuilding this software (only better). Will SaaS companies respond by lowering prices to remain competitive? Or, more likely, pivot completely to offer a compelling reason to still use their (now Agentic) services?

I'm not sure how its going to play out. Some early movers are likely going to get burned. But they’ll learn and iterate, and others will follow, IMO. As the pace of innovation accelerates, and as companies look to cut costs (nothing will get your CFO's attention like giving her a plan to cut the annual SaaS bill by 50%), there's no doubt in my mind that this is a space ripe for (wait for it) disruption.¹²

These companies aren't sitting idly by, so grab your popcorn and enjoy the show!

¹ The thing that still matters is your data. Everything on top of it is a means to unlocking the value in your data.

² This is also one of the reasons why you sometimes hear people railing about why data is the most important thing (and it’s not because you can be "data driven" - unfortunately we've been losing that battle for years). With AI leveling the play field in many respects, all you have left is your data.

³ Paradoxically, I believe the solutions that are most commoditized are the ones most at risk of being rebuilt internally, and those are the ones that paved the way for the massive SaaS industry in the first place. Simply put, they got too big and bloated, and no one seems happy about it (except their shareholders!).

⁴ SaaS can be very generic, but oftentimes we (I?) think about B2B SaaS, which is the software your company buys for you to use and become very efficient.

⁵ The key assumption here (which underlies most of the internet, e.g. targeted ads) is that the better the personalization, the better the result. But how to measure the impact of personalization? That's a tough one!

⁶ and reliable, and secure, and compliant, etc. No one said software was easy.

⁷ Unless software is your competitive advantage which is a stretch for most non-tech companies, despite the valuation multiples that are dangled in front of you.

⁸ You'll need to hire consultants to integrate it into your system, and then train people internally on how to use it, and even then it won't be exactly what you want because, at the core, all software products need to be generalized enough to sell to the masses - this is the ultimate double edged sword of software. All software converges to that 80% good enough spot. Getting to 99% (there is no 100%) requires consultants and lowering expectations - a topic for another day!

⁹ Going the agency / off shore route is becoming more popular and helps, but you still need to manage it - sorry, there’s no free lunch here

¹⁰ As mentioned, it's not just a cost thing - you need expertise of tech regulations, security folks, UI people, etc but if the cost is low enough, you can start to justify the other stuff.

¹¹ But absolutely pay for in your monthly SaaS bill.

¹² Ugh, how cringe.