Authored by Venky Balasubramanian, CEO and Co-Founder, Plivo. Plivo were finalists in the Best Enterprise-Level SaaS Product, and Best SaaS Product for Communication, Collaboration or Conferencing categories at the 2023 SaaS Awards.
Like most technology, SaaS followed the Gartner Hype Cycle, soaring to the Peak of Inflated Expectations and wallowing in the Trough of Disillusionment before gaining the Plateau of Productivity where we find it today.
In 2024 generative AI is at the same stage of development as SaaS was 25 years ago, when Salesforce launched, but things don’t feel the same. It’s hard to remember what life was like in 1999, but we live in a much more connected world today. The Web, GPS, and smartphones keep us all online all the time — and that’s where generative AI tools live. It’s no heavy lift to consider using them, lowering their barrier of entry into organizations and accelerating their adoption.
SaaS succeeded in spite of corporate managers and buyers’ attitudes that could be expressed as “but we’ve always done it this way.” People don’t like change, but they will change if you give them compelling reasons. SaaS offered more accessibility, faster time to market, and greater flexibility and scalability than the on-premises alternatives that were the norm at the time, winning over the skeptics. In fact, one of the key milestones in the founding of my company, Plivo, was a potential customer saying, “Put it on the cloud and we’ll pay for it.”
The factors driving generative AI are similar enough to let us learn some lessons from SaaS, but different enough to require caution in drawing easy conclusions.
Parallels between SaaS and Generative AI
Both SaaS and generative AI have characteristics that have both helped and hindered their adoption in surprisingly similar ways. On the plus side, both offer …
A new model. SaaS offered a new delivery model for software — via the internet rather than locally installed applications. Generative AI offers a new model for content — from cloud applications and APIs rather than from humans. Both require broadband connectivity.
Promise of increased productivity. SaaS touted improved workflows via online software. Generative AI aims to augment human creativity and output. Both technologies can boost organizational productivity.
More flexibility. SaaS is easier to deploy and faster to scale than on-premises software. Generative AI can do a lot of different things pretty well and pretty quickly — things that it might take multiple humans more time to accomplish. Both make organizations more efficient and more agile.
Subscription pricing. With SaaS you don’t pay an annual license fee. Instead, you either pay by the transaction or pay a monthly fee that might vary depending on how many transactions you want to incur — pay as you go, in other words. Generative AI services build on the same consumption-based payment model — they’re cloud services, after all.
(Having said that, tools exist for running large language models (LLM) on local hardware. Wouldn’t it be ironic if generative AI spearheaded a return to local hardware?)
Other similar characteristics of the two technologies are more negative in nature. Both come with …
Data concerns. SaaS raised new questions about whether data stored in the cloud and access over the internet could be kept secure and private. In the case of generative AI, the concern is with both data in and data out. First, is the data used to train a model accurate, complete, and objective? Also, when companies provide data to be used to generate answers, how secure and private is that data? Finally, when the AI platform delivers its answers, is the data accurate?
Learning curve. SaaS required new cloud administration skills, even if they were generally simpler than managing on-premises software. Generative AI demands new techniques to craft useful prompts. Both SaaS and generative AI require rethinking organizational processes and potentially reassigning some employees to new roles (or even letting them go).
One key parallel is that neither are inherently positive nor negative. In the case of both SaaS and generative AI, APIs are critical tools and will continue to drive adoption and integration. Both the future SaaS landscape and future generative AI platforms will be characterized by greater interconnectivity, so — other things being equal — platforms with capable APIs that are easy for developers to use to access data are likely to have a competitive advantage.
It’s also worth noting that neither SaaS nor AI is a panacea. Some applications are too complex or specialized to be covered by a general-purpose service, and therefore have to be addressed by custom-coded applications. This is just so for some problems and situations, that are too complex or esoteric for AI to offer meaningful input. That’s why the machines keep us humans around.
Not everything is parallel
One area where SaaS and AI seem to diverge is speed of adoption. SaaS saw gradual enterprise adoption on the part of hesitant IT departments. By contrast, generative AI has made incredible inroads faster than any technology in our lifetime, including cell phones and the internet. That’s likely to mean that businesses will experience all of the promise and the pains of new AI platforms in a more compressed timespan than they saw with the gradual adoption of SaaS.
In the end, both technologies are disruptive, but the speed at which businesses are rushing into generative AI is accelerating the journey through the Gartner Hype Cycle. If businesses’ experience with SaaS is a valid “role model,” the end result is likely to be positive, but organizations are going to suffer through some growing pains to get there.
Still, we predict those difficulties will be less painful than the disruption SaaS engendered. Generative AI’s promise is, essentially, “do what you do now, only faster and using fewer human hours.” That’s a pretty compelling argument, and completely sidesteps organizations’ resistance to changing strategies. Changing processes is still an issue, but it’s an easier one to overcome. We’ve found that in our organization, where virtually every department is using ChatGPT to do what they were already doing, only faster. For example, we built an AI-powered chatbot for Slack that saves our users time when searching internal documents.
Of course nothing comes without cost. You know the adage: “Better, faster, cheaper — pick any two.” Generative AI may be faster and cheaper — but it’s not undeniably better. Virtually every ChatGPT deliverable — whether it’s content, code, or art — requires multiple rounds of prompt engineering to get a draft on par with what a human can deliver (albeit more slowly). Using output from AI platforms without running it past a human editor is risky business, as some professionals have discovered.