SaaS Is Not Dead, But Its Shape Will Change

The rumours of the death of SaaS are much exaggerated. AI will kill SaaS. AI will enable a new age of hyper-personalisation. We have seen large crashes in the share price of SaaS across the board. While the future may look different, SaaS is not dead. Rather than hyper-personalisation, a reduction in oversized integration costs will favour more modular, best of breed vendors over bloated all-in-one platforms.

In the world of big business, the moat around SaaS is the extreme cost of integration. Big SaaS vendors have large teams to assist with this integration and free credits and consulting rains down to assist with the pain of onboarding. In this world, it can take months for a simple proof of concept (POC) to provide comfort that a solution works. Beyond the POC are multi-year projects where the implementation costs are 5-10x the annual license fee. The acquisition and onboarding cost is high on both sides. For the vendor, although the CAC is enormous, they can be confident that their customer cannot leave, providing plenty of time to turn a profit. This lock-in works like a contract. (It’s actually better without a contract. When the contract expires you can pump the price with abandon knowing you have them trapped.) For the business, it necessitates the do-it-all solution — it’s too costly to integrate multiple software vendors, so a vendor needs to be able to do everything. This often leads to the door of the same handful of large SaaS vendors. Not world-beating in any category, but able to provide a solution in most if not all. At times, it can mean that a business buys vapourware — a desperate vendor salesman commits his team to deliver new functionality to get the contract signed.

This is where AI can change the structure. AI excels when given clear specifications. These vendors have detailed integration documents. Examples in the data space, where data models need to be tweaked to align inputs and outputs — AI can consume the incoming and outgoing data models, spitting out a clean ETL. Closed loop integrations in this space can easily be tested to provide comfort. This will reduce the cost and accelerate integration. Now the POC can take days or less, the cost of experimentation collapses. This reduces the risk for the business and shrinks CAC for the vendor. The vendor can become more focused. No longer needing to be a jack-of-all trades. They can specialise.

The cost and scale of integration costs have held B2B back in this area. Yes, buying 5 or 6 best of breed products would be better than a single average jack-of-all trades — but 6 integration points (6 points of failure) would be enough to sink a business. While the outcome may be better, no executive is willing to take that risk with their career. If AI collapses the cost of integration, these small boutique vendors have a chance.

A short segue into AI-assisted development. At this point, people start to imagine that each business should in-house build their entire technology stack. Why pay 5 or 6 vendors when I can build software explicitly for me? This is a mistake. A bank shouldn’t write HR software nor a telco write accounting software. I think people have the mistaken idea that they are (or should be) different everywhere. A business must focus on the area where they add value. Is a highly customised in-house HR system going to help a bank sell more mortgages? But if you view HR as an off-the-shelf solution, doesn’t that diminish the function? Absolutely not. You need to recruit the best, but it’s not the software that does that. The software should get out of the way. If the head of HR needs to worry about maintaining and improving HR software while at the same time trying to recruit, that distraction will dilute the function.

Businesses shouldn’t build themselves to reduce costs either. If cost was the key consideration, everyone would use free open-source options. When in fact, large corporates pay a license fee to use a managed version of open-source software. The open-source movement relies on this premise — some enforce a license fee for large users, others offer a fee to provide a managed service with support. This makes sense. If I am a bank, should I be working out how to implement Docker? No. I would need to hire someone, manage them. It’s not a skill I have, nor do I know how to hire and manage for it. I outsource this to a vendor. By the same token, if I build software myself with AI, the risk is worse. At least with open-source there is a community maintaining the product. Building myself, I need to manage the product and its implementation.

Now we can imagine a business using multiple best-of-breed vendors. If any falters or a new solution appears, they can swap or augment — confident that integration costs will not erode any benefits. Then we can see B2B SaaS vendors moving in the same direction as consumer products. Over the past 20 years, big brands — once the primary signifier of quality — have fallen. For consumers, social media allowed the masses to pronounce quality. Small speciality producers could get exposure and sell their product without needing alignment with a large branding house. For SaaS, a collapse in integration costs means a large enterprise funding integration credits and free consultants is no longer required. Those that move quickly to adjust their business model will grow. They can have a leaner integration workforce and give away fewer credits while focusing on the business of writing software.

SaaS won’t die. Integration friction will no longer provide protection. Smaller, more focused vendors with the best product will thrive — not those with the biggest integration army.