2024 MIT Digital Technology and Strategy Conference: Lightning Talk - SWIRL

SID PROBSTEIN: Good afternoon, everybody. It's great to be here. I'm Sid Probstein, MIT '90, and I'm a first-time CEO. I've been CTO for many ventures.

And I think the key thing I've loved about it is bringing together data from multiple sources. In fact, my first job was at John Hancock Financial Services, put together a database of 12 different product lines that had been separate for a hundred years. And the insights we produced were amazing.

That pattern is one that really continues. And what I'm here to tell you about is SWIRL for the enterprise, helping turn that painful data scavenger hunt where you spend time going from tab to tab, saving stuff, rewriting it, organizing it. That scavenger hunt is, AI is going to end it. We're going to get answers and insights. And the key is AI infrastructure software.

Let's take a step back. You might have heard a lot about the death of AI or seen articles about how we're in the trough of disillusionment. It's very hard to make it work. The problem is that most AI solutions want your data. They want you to copy it in.

Now, if you have just a single silo of data for your application, that's pretty straightforward. But if you have a data sprawl like the large enterprise, it's very, very impractical. There are security issues, there are compliance issues, and actually those loom a bit larger.

And there's always the question of resources. Do we really need to copy 9 petabytes of data? How expensive is that going to be? The answer is about a million and a half dollars.

SWIRL is AI infrastructure software. You install it in your environment. It's not a hosted service. You can deploy it anywhere, Edge, your own networks, any cloud provider.

It provides a secure connection between your choice of AI and your back-end data sources. It integrates with your single sign on, so there's no new permissioning of data. There's no bulk copying. And it actually gives you the power to turn an AI into an assistant.

What is a conversation with your data look like and why would you want one? The answer is, it saves a huge amount of time and it follows a very natural process of being interrupted by questions. And one question leads to another.

So I got a question recently, what do we have for cyber insurance? I asked SWIRL. And by the way, you could do this with a UI, an API, or Microsoft Teams, or Slack. One of the great things about AI is it's omnichannel.

So after I clarified some questions about my intent, I was trying to find SWIRL's policy, less than a minute later in total I got the answer. This was open AI GPT-4 via SWIRL's RAG against, as you can see, my OneDrive or SharePoint. Not only do I have the answers, but I have the citations so I can verify that this is correct.

And these insights are personal. I'm the only person in this world who can see this, because this document was sent to me and I purchased the policy. SWIRL produces personalized insights.

Our secret sauce, if you will, is SWIRL's reader large language model. You may not have heard the term. It doesn't hallucinate. It's only good at reading and evaluating text, and finding relevant portions.

A great video online. You can check it out. SWIRL reranks Google and does a better job of finding the right answers.

If you search right now, go to Google Search for "red hat acquires." The number 1 answer is all about IBM acquiring Red Hat. Why? Well, it's popular. And, you know they spent a lot right on that PR and SEO.

If you ran that through SWIRL, it will pick up results 7, 8, and 9 from Google, which are about Red Hat acquiring companies. LLMs and AI care about word order and that you're actually on topic. That's contextual accuracy.

So we all can do that same conversation or answer questions and produce insights using any of the AI on the left-hand column and any of the data platforms, enterprise applications, and web services, including structured data like ThoughtSpot, Snowflake, Postgres, new databases like Pinecone, older ones like Solar and OpenSearch, stuff I love from that world. Also, information services, whether it's OpenSanctions or something like Blockchain.

Now, who is using SWIRL? I'll give you some examples. A global life sciences firm, their supply chain group, essentially stopped using a ticket-based data science team with a two-week SLA. They self-serve using SWIRL now across Collibra, ThoughtSpot, Atlassian, Snowflake, many others. Their internal study showed a 7.5 hour per week savings. 7.5 hours per person per week.

A local university uses SWIRL to search across multiple catalogs. A global retailer is combining the ability to ask product and HR questions, like can I take a return with a tear in this area? Again, massive time savings.

So SWIRL, we're here looking for people who want enterprises, who want to do proof of values. We guarantee results. And we're happy to do POVs for anybody who wants to engage with us.

We're looking for folks with at least 5K employees, also US government agencies that may be interested in air-gapped solutions. SWIRL can be deployed anywhere. It's a great solution for that.

Last point, software and solution developers who build solutions for similar clients, love to hear from you. We're easy to embed. Thank you so much.