DEVOPSdigest

SaaS tools have been the dominant way to deliver technology to users fast and efficiently for over twenty years now. So many of the basic functionalities that companies rely on day to day — a CRM, CMS, project management — are powered by SaaS giants, many of which have been dominant players in their space for at least a decade, if not more.

But can AI give newcomers a chance to unseat these SaaS behemoths?

While SaaS giants have been quick to proclaim new AI features along their existing products, are these new capabilities actually going to provide value?

The AI experiences they're promising need to fit into their already complex user experience. New entrants, however, have the flexibility to think differently, and the opportunity to build differently, with serverless tools at the helm. Here are three concepts you should be integrating into your roadmap if you're aiming to disrupt the SaaS space with AI:

1. Let serverless tools do the heavy lifting on infrastructure

Why bother with infrastructure decisions when your main goal is getting from zero to a product?

While SaaS giants may have the funds to hire fast and invest in new tech, AI is still a new engine that needs to warm up. From conversations with larger enterprises that are looking to build with AI, it's clear they are often starting with an AI infrastructure team. Some of the largest companies snatched up upwards of 50,000 to 150,000 GPUs, and are likely pulling productive engineers and leaders from other teams, in order to plan, build, and manage the cumbersome infrastructure needed from scratch — whether that's an internal need for an AI chatbot, or managing the hardware somewhere that makes those future hypothetical AI tools available to customers around the world.

Smaller companies can reap benefits from more recent innovations with developer tools to improve developer productivity; worrying about infrastructure decisions before you have a product can be a time suck and a major roadblock. Smaller teams should favor serverless platforms that offer scalability, cost efficiency, increased resilience, and most importantly, can reduce the tedious overhead like provisioning and maintenance, in order to ensure a faster time to value. Serverless affords you more agility down the road, if you grow faster than expected, or have users in regions you didn't predict.

2. Dream big and design the user interface of the future

SaaS software can be clunky and complicated; take a quick glance at LinkedIn and you can find countless resumes that tout the expertise and certifications required to use certain SaaS products. The first thing SaaS incumbents will do is bolt on an AI chatbot to proclaim their product is already AI-powered. However, chatbots are still too human-prompted; they require too much from the user up-front and often just lengthen the time a task takes. And most importantly, a chat bot does not necessarily mean an application is "AI-first."

Startups have a unique opportunity to completely rethink and redefine what the interfaces of the future can look like. Clunky CRMs and wizards should be long gone, and AI does not have to be a "destination" but a camouflaged tool. Where do users actually want to see AI? No one has ever asked to interact with another chatbot, but we can envision AI becoming seamlessly woven into where the user already is in real life–in their email, text, command line, dashboards, calendars, and more.

3. Don't get LLM FOMO; choose a model optimized for your needs

The foundational model space is moving rapidly, it seems like there is a new model or new version of a popular model out every week. This puts a massive burden on organizations to keep up with what the best performing models are. Constantly evaluating these models manually is a burden and takes up a lot of a team's build time.

While many companies with large budgets will gravitate to the largest, most costly, or most recognized models, that does not mean those are the best models for an organization. Aim to find an optimization point where cost, accuracy, performance, and security are all taken into account. Take into consideration the tasks you aim to execute, the scale of your project, and the opportunity for fine-tuning. Incremental improvements in accuracy will likely not make a major difference in the customer experience, but will drastically affect budget and the runway to experiment in other development areas.

Additionally, tools that help teams to compare different models through live experimentation, and kick off different fine tunes in parallel, will keep teams moving fast and prevent getting held up for the elusive "best" model available that week.

Remember the blockchain hype? IoT? VR? All of these were fast at the get-go, and then tapered off, providing little value at the end of the day. Time will tell if AI is just another hype cycle, but AI holds new potential for disruptors, and it is truly just the beginning. As new challenges emerge, from regulatory compliance to the first major AI breach, AI development platforms that enable speed, security, scalability, and affordability, will be the shovels that power the AI gold rush.