AI startup Swa Technology today released a multi-model AI orchestration platform that incorporates models from open source and commercial providers to meet engineers where they are at and cut down on shadow AI.
While many organizations try to cut down on shadow IT and AI by offering users limited choices of tools, Swa’s platform provides secure, sanctioned access to all of the popular and most-used AI models currently available.
“If employees aren’t given the tools they want, they’ll go and get them,” Mike Sirchuk, founder and CEO of Swa, told SD Times. “Basically, we are trying to solve the question of which AI do I choose? And you don’t have to choose one. You just get them all.”
What Swa does is, it matching a user’s prompt with the model best suited for the question. “Some models perform better than others, depending upon what the question is,” he explained. “Some models you you can’t ask math questions because it will get them horribly wrong. Other models can perform math questions very well. Some models are more conversational and great for marketing, and others are very much more logical and are great for code creation workflows and designs.” Also, the platform can detect is a model is down, so it will shift the request to a different model.
The platform separates a context layer from the customer’s information and brings AI agents into such productivity tools as Slack, Zoom and text messaging, for example, allowing companies to build the security layer around it so they can manage who has access to what model, what data each model has and how much data is being stored. Sirchuk noted that nothing has ever been trained in Swa’s systems, which gives the platform the ability to switch and choose from any AI provider at any point in time.
Businesses can upload into Swa’s administration portal their proprietary information, such as marketing information, product information, company information, PTO policies, expense report policies and the like. “So then a new employee or an existing employee, instead of having to talk to somebody, they just ask, ‘Swa, what’s our PTO policy?’ They’ll get a response directly from the policy. ‘Swa, I have an expense report. Here’s a receipt. Can you file this expense report for me?’ Yes, it can extract the data from the receipt to this expense report policy and send it to the necessary policy stakeholders.” This, Sirchuk said, removes a lot of layers of IT management from a lot of different software applications.
Swa, a Sanskrit word meaning ‘one’s own,’ defines the company’s all-in-one approach to AI. The company was formed 18 months ago by engineers who were laid off from Nike, who took their knowledge of how to build bots. Sirchuk said Nike used the bots to detect when a system goes down and then spin it back up. “Well, you throw generative AI on those bots, and they can do a lot more than just reset an instance. They can detect things that they weren’t previously programmed to detect. They can perform actions that they weren’t necessarily specifically told to do. So after we got together, and we were kind of ideating on what could we do with this technology. The first problem that I had was I was using two or three different models, and the cost was just just too much. I mean, I’m spending 60 bucks per person just on three models. And so we said, let’s first build a bot that can talk to all the models for us. And that’s kind of how Swa was born. We reduced those costs from, I think it was $300 or $400 a month for our small team down, to $40 a month for our small team.
“And that’s when we’re like, Yeah, this is something. Wow.”
