Umesh Sachdev is the CEO & Co-Founder of Uniphore, a leader in Conversational AI and Automation. He’s based in Silicon Valley, California.
Artificial intelligence today is spreading like wildfire. New and more powerful applications are emerging almost daily, and people are interacting with AI more in their everyday lives.
All of this is very exciting for modern enterprises, which have reinvented themselves as digital powerhouses post-pandemic. Still, digital transformations don’t happen overnight, and tech companies—and SaaS models in particular—were the saving grace of businesses sprinting to align their customer experiences with new consumer expectations and data-driven capabilities.
But AI is a different animal. Not only does it require huge sums of data, but it also requires a new approach to how that data is obtained and managed within the enterprise. Businesses expecting a simple “build it and leave it” experience with AI are in for a shock, and many will undoubtedly ask, “Can SaaS deliver AI at speed and scale?”
The answer is not yet. But that’s about to change, and soon. To understand how, we need to look at the current state of the AI/SaaS relationship and explore how today’s brightest minds are creating new strategies for bringing these two seemingly incompatible concepts together—and how it will change the enterprise experience forever.
AI is powerful—and power takes time to develop.
To be truly effective, AI takes time to develop and deploy because it requires large amounts of data to train on and fine-tune. This is even more pertinent for large language models (LLMs), which are mostly trained on massive data stores. For example, GPT-3 is trained on 175 billion parameters, while GPT-4 is rumored to be trained on even more.
While news-breaking innovations, like those around generative AI, make the industry seem like it’s moving at light speed, in reality, those innovations have been in development for years (and not just by headlining companies). That includes generative AI models for the modern enterprise.
For B2B enterprises, it isn’t enough for AI to be trained on publicly available data. It must marry that data with internal enterprise data. This includes established product and service data as well as contextual customer interaction data.
The problem is that enterprise data comes from many sources, many of which are siloed (and some even owned by third-party vendors). This data must first be extracted before it can be of use to an AI application.
However, data extraction is just the first hurdle. Once extracted, much of this data is unstructured (i.e., free-flowing documents, knowledge base data, recorded calls, etc.). To effectively train AI software, this unstructured data must be structured. Both hurdles add time to the development and deployment process, which runs counter to the speed associated with SaaS.
SaaS is all about speed.
Over the past decade, enterprises have come to appreciate the SaaS business model, which is built on one thing: speed. The ideal SaaS solution is fast to implement, fast to consume, “pay as you go,” and “pay only for what you use.”
That’s because the underlying feature of the SaaS business model is velocity. What makes SaaS so appealing to B2B businesses is its velocity of selling, velocity of implementation, velocity of usage and—most importantly—velocity of receiving value.
And SaaS delivers, at scale, thanks to its efficiency: build once, host on a multi-tenant cloud and share the infrastructure with multiple customers. This streamlined approach allows SaaS to achieve a high gross margin, provided that additional solutions can be developed, deployed and deliver ROI quickly.
Because SaaS customers expect fast, measurable returns on their investments, the SaaS model hinges on existing customers buying more of the same product or buying more modules. When done right, this creates a compound effect on net revenue retention (NRR), which makes SaaS selling very efficient.
However, when SaaS companies overpromise on a solution—like AI—or fast-track software development at the expense of quality, it can quickly drive customers away in frustration and confusion.
Can AI ever be SaaS-ified?
Yes. In fact, I believe SaaS is the next logical step in the evolution of enterprise AI for two reasons: First, business demands it. In today’s AI-enabled world, we have all seen the immense benefits of technology on operational efficiency and revenue generation. And with recent advances like generative AI, the sky is the limit in terms of optimizing processes and unlocking new capabilities.
Second, AI developers and their customers are realizing that to reap the benefits of AI at speed and scale, enterprises themselves must be structured around data. Yes, businesses must change to unleash the full power of this game-changing technology. The status quo cannot support what AI is capable of. But making this change is well worth it.
What exactly does that change look like? For most enterprises, it means creating new business operation units devoted to data engineering, data preparation and data operations (i.e., data ops or machine learning ops).
Equally important is extracting and preparing siloed data for AI applications and converting unstructured data to structured data quickly. Together, these human and machine factors form the critical operational component for making data available at runtime to any application where it needs it and when it needs it.
It’s closer than you think.
So, as you can see, it is possible to merge something as time intensive as AI with a high-velocity business model like SaaS. The world is looking to us as an industry to deliver this. The time is now.
With the right strategy and tools, we can SaaS-ify AI for modern enterprises. In AI, we have the means to synthesize enormous sums of data—both publicly available and enterprise-gated. We also have the tools to extract, structure, organize and manage siloed data from multiple sources.
In SaaS, we have a business model that’s structured around efficiency and is highly capable of adapting solutions at scale, pooling data from an ocean of enterprise and industry sources.
By merging research with data engineering, we’re a lot closer to SaaS-ifying AI than you think.
Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Read the full article here
