Past innovations in IT might pale in comparison to the global impact of AI—but only if vertical industries—like healthcare—define and deliver on the jobs to be done
It’s 1995. After a 26 year gestation, the Internet and the World Wide Web were suddenly front page news. An overnight success 26 years in the making. The first ARPANET data was transmitted on October 9, 1969.
Microsoft’s then-CEO Bill Gates saw big changes coming that would fundamentally redefine personal computing, business computing, and proprietary private networking—markets that were core to Microsoft’s industry leadership.
In a now-famous email to his executive team titled “The Internet Tidal Wave,” Gates predicted the Internet “will set the course of our industry for a long time to come.” He directed that “every product plan … try and go overboard on Internet features,” noting he could find almost no Microsoft file formats on the web after 10 hours of searching.
In what’s since been described as turning an aircraft carrier on a dime, Microsoft rapidly pivoted the company to ride the Internet tidal wave rather than be overrun by it. And of course, today the company is a leader in Internet technology, along with others such as Apple, AWS, Google, and Snowflake.
Creating the Next-Generation Infrastructure for AI
It’s 2023. After a 78 year gestation, a day doesn’t go by without another AI launch, innovation, or prediction—either gloomy or glorious—in the news. Sparked by OpenAI’s unexpectedly explosive launch of ChatGPT 3.5 in Nov. ‘22, it’s clear that generative AI and LLMs will influence the course of not just computing, but of virtually every pursuit and industry for years to come.
Companies are scrambling to create AI slideware and software. Consumers are dually clamoring for or fearing AI’s impact on their personal and work life. Companies are equally baffled by, banning, flirting with, or embracing the potential of generative AI in their organizations. The rules and guardrails have yet to be set.
The AI tidal wave is here. And it’s going to be a huge productivity booster for society at large. Personally, I think it will make all of us a lot more efficient and effective at doing what we love to do. I believe AI will allow us to do more of what we love, and less of what feels like “work.” More on that in a bit.
As the AI evolution unfolds over the next few years, we see three fundamental layers coming to the fore. The first two are already evolving at a breakneck pace. The third layer is emerging, and I believe it’s absolutely critical we get this layer right to ensure success in healthcare or any other vertical.
- The compute layer, which companies like Nvidia, Intel, AMD, and IBM have cracked (especially Nvidia). Even the most powerful traditional CPUs can’t support deep learning models. AI chips with immense parallel computing capabilities and bandwidth are required to get meaningful results as LLMs get larger and the number of parameters of their neural networks expand (GPT-4 has 1.7 trillion parameters).
- The general intelligence LLM layer, which companies like OpenAI, Anthropic, Microsoft, Google, Meta, Stability, and others are building. These Large Language Models use billions to trillions of training points (“tokens”) culled from the public Internet and World Wide Web. For industry, the practical applications for general AI will play into horizontal functions.
- The vertical intelligence LLM layer, which is where we’ll see “industry-specific jobs to be done with AI” emerging. This is the third fundamental layer and it requires domain expertise and a foundation of high-quality, industry-specific data to function. This is what we at Innovaccer are building and investing in to create the next-generation AI infrastructure for healthcare. This is the layer that starts to define what AI in healthcare will look like.
When developing industry vertical generative AI applications, the context must be specific, and the domain understanding deep. Regulated industries—such as healthcare, finance and banking, pharmaceuticals and biotech, energy and utilities, aerospace and aviation, etc.—will be particularly challenging for companies building general/horizontal AI applications (in other words, those using the general intelligence layer).
The companies that will end up delivering the best industry-specific LLMs for their customers are those that have the high-quality data infrastructure in place for their specific industry. For Innovaccer, that’s healthcare. Innovaccer has cleansed healthcare data for more than 50 million Americans—data that’s trusted by clinicians with the help of over 10 million clinical terminologies and 7,000 quality rules. I believe we’re one of the largest healthcare data infrastructures in the country.
In addition to an expansive healthcare data set, healthcare AI needs an infrastructure that ensures data quality, ownership, accuracy, security, compliance, and trustworthiness. Those are the same prerequisites for everything we do today on the Innovaccer platform—data analytics, data-driven decisions at the point of care, VBC, population health management, care management and coordination, healthcare CRM, risk adjustment, and so on.
AI infrastructure and innovation that is “ready for healthcare” is, therefore, a logical next step for us and our customers, because we have the domain knowledge, we have the domain data, and our platform is built on a scalable data infrastructure that’s well suited for vertical intelligence LLMs. Healthcare organizations can’t risk care quality or processes on general AI that’s prone to hallucinations or inaccurate output, or at risk of copyright claims. We want our customers to be able to fully leverage the power of generative AI, and also ensure it is applied in ethical and trusted ways.
A Clear Path to the Future of Healthcare AI
Healthcare AI needs an enabling platform, and we’re committed to delivering that—building the first and largest vertical AI infrastructure for healthcare. We recently announced our first four healthcare AI solutions built on this infrastructure, and we’re identifying and investing in more jobs to be done.
Our suite of four healthcare-specific models—Sara for Point of Care Assistance, Sara for Care Management, Sara for Consumer Experience, and Sara for Insights—leverage our Best in KLAS healthcare data and analytics infrastructure, the Innovaccer platform—with high-quality data from over 50 million patients to create the next-generation of integrated workflows powered by AI.
Our first four Sara “agents” will help with work stakeholders now have to do manually, tediously, after hours in their pajama time, or ask others to do for them. They mark the beginning of our journey to identify, invest in, and fill the “AI whitespace”—the jobs to be done that we think can, over the next five years or so, be assisted by AI.
We want to help everyone practice at the top of their licensure, and help reduce the “digital paperwork” that now consumes so much of their time—as well as provide easy access to the information and insights they need, whether at the point of care or as part of business operations.
We thought about “what does a physician do, and how can we reduce their administrative burden?” and created Sara for Point of Care Assistance, which is going to help doctors do what needs doing. Ambient listening documentation and coding is something that’s important for payment, but it’s not where doctors want to spend their time. If an AI could remove that from their plate, wouldn’t they be incredibly happy?
We asked, "What can we do for care coordinators?" It’s the same thing. They’re on voice calls, they need to transcribe them, they need to come up with action items, care plans, and things of that nature. Again, all jobs that need to be done, but not jobs care coordinators really love doing. What if they could have those tasks done automagically and spend more of their time talking to patients? That’s Sara for Care Management.
Similarly, for contact center representatives, does anyone enjoy having to dig through a 1000-page PDF trying to figure out where the copay is for a particular thing? That's something that AI can assist with—and does, with Sara for Consumer Experience. Contact center reps would rather spend time talking to patients and members and helping them navigate care more effectively.
As for the data analyst, do any really enjoy spelunking through hundreds of dashboards to figure out what the PMPM for a cohort and population is? What if they could just ask the question and get the answer? Sara for Insights is our AI solution for making conversational analytics a reality.
Our Sara AI solutions are just for starters. Innovaccer is known for our ability to ingest disparate data, transform it to high-quality data, and embed analytics to create smart, integrated, engaging workflows. We see many more jobs to be done with AI in healthcare beyond our first four tools, and we’ll continue to invest to fill that “whitespace” with AI solutions built specifically for healthcare, which leverage our high-quality healthcare data infrastructure.
The jobs to be done are the verticalization of AI for healthcare. If people just depend on artificial general intelligence, it's essentially putting AI to work that hasn’t gone to medical school, never practiced healthcare, and doesn’t understand the complexities of healthcare data—not to mention the legal and customer requirements around the use of healthcare data.
We need AI that’s trained on healthcare and only healthcare, and that’s going to be our focus throughout this evolution: developing and delivering proprietary models trained to understand thousands of healthcare entities, concepts, formulas, and disease conditions. Healthcare AI you can trust.
In our view, that’s how and where AI is going to create impact. And so that’s the direction we’re moving in: to develop more AI products that solve the jobs to be done, using the industry’s leading data platform for healthcare as a strong infrastructure base. We’re the first company to do this, and we’ve resolved to swing for the fences by identifying all the whitespace where healthcare AI can help stakeholders be more productive, have less busy work, greater career satisfaction, and less burnout.
We’d love to hear from you—providers, payers, and life sciences companies—about the jobs you see that need to be done, and could be assisted by a Sara AI for healthcare.
A Future Where We Do Less Work, Yet Are More Productive
If we look just five years into the future, we’re going to be living in a very different world—and I think all of us are not going to have to do a lot of “stuff” we don’t like to do today. The notion of work itself will be redefined.
In all aspects of what we now think of as “work,” the opportunity is to figure out what are things that take our stakeholders away from what they love doing—caring for patients—and ask how we can use a healthcare-trained vertical AI model to do that “work” for them. We want to help stakeholders do something that’s cognitively dissonant: be a lot more productive by doing a lot less work, to give them more time for the high-value things that bring them joy and satisfaction.
This is, I believe, the next-generation in software’s evolution, and the next-generation in the evolution of AI analytics, assistants, agents, copilots—call them what you will. In healthcare, it’s all about enhancing productivity by helping with the jobs that must be done, but which stakeholders would prefer someone else do for them—so they can spend more of their time operating at the top of their licensure, delivering quality care, bending the healthcare cost curve, and improving the financial strength of their organizations.
This is the moment which starts to define what AI in healthcare will look like. This is the future of healthcare and AI that we are being asked to, want to, and will strive to bring to the industry. Innovaccer is working to create that next-generation AI infrastructure for healthcare to deliver that future for all.
Looping back to Bill Gates, he also once famously said: “We always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten.” The effort to define and work on the jobs to be done, the imperative to establish AI’s third-layer and drive the change that will occur over the next decade—the time to define the future of work in healthcare—is now.
/1698734448492/file.png)
