The Future of Gen AI, Apps, and Your Code in Flux
Sonya Huang and Pat Grady of Sequoia Capital recently published a paper on the future of generative AI. In it, they compare Gen AI to prior waves of tech disruption, specifically mobile and cloud. A nice graphic that structures their argument is here:
The implication, which I believe is correct, is that the next wave of ~$B+ revenue companies will use generative AI to build solutions above the infrastructure layers of the AI wave. Though perhaps straightforward, it is still super exciting to imagine the MANY solutions that will be created across a multitude of market segments. We are at the very beginning of this revolution, and it will be amazing to witness how it unfolds.
More interesting than that idea, however, is their notion that these AI-native solutions will deliver Service-As-A-Software – a clever turn-of-phrase of the previous generation’s SaaS (Software as a Service). Specifically, the use of generative AI in these applications will provide a level of customization and effectiveness that was only previously possible with professional services applied on top of a SaaS platform. For example, to make Salesforce truly effective you need to hire professional services to tune it to your specific needs and requirements: what your sales stages are, what criteria must be met to move to the next stage, and so on. Imagine instead a next-gen Salesforce that derives these unique requirements from your email, call transcripts, etc. It could determine based on this data what the stages of your sales process are, the status of accounts, when they go to the next stage, and more. It could even suggest improvements based on observing data over time. For example, based on witnessed behavior, a better indicator to use as a basis for creating a sales opportunity is x, not the indicator that was previously defined. In so doing, via AI, this app provides at least some of the service in its software that you would otherwise require as professional services in order for the solution to be effective.
Though I find Service-as-a-Software to be an awkward mouthful, I have to admit the idea really resonates with me. The more you think about it, and the more applications (like Sales, Marketing, or Customer Success) you apply it to, the stronger it gets as a framework for thinking about markets and opportunities. And like the very best ideas, it provides new and valuable perspective even to situations you are already intimately familiar with. Like Flux, for example 🙂
Based on feedback from our users, we’ve known for some time that one of Flux’s superpowers is that it makes it easy to get a high-level (but detailed) evaluation of your code across quality, complexity, security / privacy, and third party risk. Connect Flux to your repo, give us a few minutes, and voila! The idea of Service-as-a-Software, however, beautifully clarifies just how valuable this is.
Without Flux, evaluating code in even one pillar, like complexity, is no small thing. Doing that well typically involves using a handful of discrete tools to get sufficient coverage across your tech stack. And because the output of those tools is diverse and verbose, you will then need a small team of senior engineers to comb through it all to turn the noise into signal by de-duping, correlating, removing false positives, adding weighting, and so on. It’s frankly surprising given the enormous level of effort and complexity that organizations can do this for even one pillar, let alone multiple. But in truth, the only way you can do this well – in particular if you care about multiple pillars – is to hire consultants. For proof, look no further than the most critical moments – like a significant M&A event or large scale transformation – when expensive third party consultants are standard.
Flux’s compound AI approach – blending LLMs, static analysis, prompt engineering, and more – is the secret to achieving this Service-as-a-Software. It allows us to deliver a level of code evaluation that would only otherwise be possible with significant human effort, certainly internally and probably externally, too. In this way, by enabling Service-As-A-Software for code evaluation, Flux democratizes the outcome, bringing it to the masses and not just for the most critical moments, but for everyday use.
