Even as artificial intelligence promises to transform many facets of life, the majority of AI companies in the U.S. are still concentrated in traditional tech hubs. According to a Glassdoor analysis from November, 30 percent of open jobs that included the words “artificial intelligence, “AI,” or “deep learning” in their title were located in San Jose, with another 18 percent in San Francisco.

VentureBeat decided to take a look at what kinds of AI startups are forming in the Midwest — an area venture capitalists have traditionally overlooked — and see what problems these companies are tackling. Using data from research firms CB Insights and Crunchbase, we looked for Midwestern startups with unique machine learning or artificial intelligence platforms that have raised significant amounts of venture capital. We also spoke with two Midwestern venture capitalists about which startups are worth watching.

Chris Olsen, a partner at Columbus, Ohio VC firm Drive Capital is bullish on the Midwest’s ability to produce game-changing AI startups.

“The real benefit of artificial intelligence is the application to traditional problems and products that the world needs, and the really successful companies have that domain knowledge that they can understand how to apply this technology,” Olsen told VentureBeat in a phone interview. “We see more of those domain experts in these industries [with] massive chunks of GDP that exist here in the Midwest.”

One disadvantage AI startups in the Midwest face is that the area has less talent capable of building AI systems. Gene Munster is managing partner at Minneapolis‘ Loup Ventures — which invests in AI and VR, among other pioneering technologies — and he estimates the Midwest has access to “one tenth” the AI talent Silicon Valley does.

“I think the good news is that as AI becomes a more popular field — there are obviously good schools here, and as it becomes a more popular field over the next 10 years, I think that gap will dramatically close,” Munster told VentureBeat.

Our analysis revealed that many of the Midwest’s most promising AI startups were born of local universities or have ties to the region’s historically dominant industries, which offers a clue as to what types of startups the Midwest could give rise to in future.

Chicago

One of the fastest-growing AI startups in the Windy City right now is Uptake, the latest in a string of companies from Groupon cofounders Brad Keywell and Eric Lefkofsky. With  $264 million in venture capital funding, Uptake has developed algorithms to help its enterprise customers collect and analyze sensor data on industrial machinery and equipment.

There’s also Narrative Science, which has raised $43 million in funding. Born out of a Northwestern University student’s class project, Narrative Science has developed a Natural Language Generation platform that delivers written insights that can be used to assemble material — like client reports and marketing copy — in the tone and style of the customer’s business.

Another notable candidate is car insurance startup Clearcover, which has raised $11.5 million in funding. Clearcover has developed an algorithm that helps it advertise to potential customers during what it calls “moments that matter” — when a consumer is shopping for a car or looking at ways to save money. To time its offers, Clearcover integrates its API with insurance comparison sites, personal finance apps, and other partner websites.

Data science startup Civis Analytics has garnered national attention as it’s led by the former chief analytics officer for President Barack Obama’s 2012 reelection campaign and counts Eric Schmidt as an investor. Civis Analytics, whose clients include Airbnb, Verizon, and Robinhood, has developed a subscription cloud-based data science environment in which data scientists can consolidate data, analyze it, and build predictive models. Civis Analytics closed a $22 million round in November 2016.

Then there’s Catalytic, which has raised about $16.7 million in funding. Catalytic’s AI-powered software advertises the ability to automate over 100 common workplace tasks, such as working with spreadsheets and filling out forms. Catalytic also integrates with software like Salesforce, Dropbox, and Workday.

St. Louis

Of the St. Louis startups dabbling in machine learning, two that seem to have gained traction have ties to local research institutions. There’s Cofactor Genomics, founded by three former Human Genome Project scientists who worked out of Washington University in St. Louis. An alumnus of Y Combinator’s Summer 2015 batch, Cofactor Genomics built a platform that uses RNA biomarkers to predict drug response and monitor disease. It then uses machine learning to tie cell types, diseases, and treatments to an encyclopedia of RNA fingerprints. The company has raised about $21 million in venture capital.

Benson Hill Biosystems is based on research that came out of St. Louis’ Donald Danforth Plant Science Center. The company has developed a crop design platform called CropOS that uses machine learning to help agriculture companies identify which seeds will produce the desired traits, such as an increase in yield or in a certain nutrient. Founded in 2012, the company has raised $35 million in funding.

Another up-and-comer is Jane.ai, founded by the former CEO of Answers.com, which recently raised an $8.4 million round. Jane.ai offers a chatbot for businesses that sifts through data from cloud storage providers, emails, and other data files to help employees more easily access company information.

Ann Arbor

Ann Arbor’s University of Michigan has played an instrumental role in the growth of several AI startups in the state. Trove, which developed a platform that scans users’ emails to provide them with insights about their professional network, has worked with a pair of research teams at the University of Michigan, the company previously told VentureBeat.

Specializing in conversational AI is Clinc, founded by University of Michigan professors. One of the products Clinc has developed is a voice-activated intelligence assistant called Finie, which integrates with certain banks’ mobile applications, allowing customers to ask questions like where the nearest ATM is and receive a natural language answer.

May Mobility, which has developed its own fleet of electric self-driving vehicles, was founded by the former head of the APRIl robotics lab at the University of Michigan. A spokesperson for May Mobility also notes that about half of the company’s employees today have some sort of connection to the University of Michigan.

Trove, Clinc, and May Mobility have raised approximately $11 million, $7.8 million, and $11.6 million in funding, respectively.

Columbus

Like Clearcover, Columbus’ Root Insurance is using artificial intelligence to sell cheaper and more targeted car insurance. Root, which recently raised a $51 million series C, has developed an algorithm that uses sensor data from a person’s phone to determine things like how quickly they drive around a corner or how often they tailgate, and it then uses that data to generate a quote for a prospective customer.

Columbus is also home to CrossChx, a health care startup armed with about $40 million in funding that has developed an AI bot called Olive. The bot helps health care companies complete tedious administrative tasks. Lastly, there’s Nexosis, which has created a machine learning API for developers and has raised about $7 million in venture capital funding.

Minneapolis/St. Paul

Though his firm Loup Ventures is based in the Twin Cities, Munster said he hasn’t invested in any Minneapolis AI companies yet. But he highlighted two promising startups to VentureBeat — Equals 3 and Rambl (formerly known as Aftercode), which have raised about $7 million and $2 million in venture capital funding respectively, according to Crunchbase.

Equals 3 has developed an AI-powered marketing assistant called Lucy. Lucy scans client and industry reports to advise marketers where to spend their money. Rambl’s AI-powered assistant targets sales professionals, promoting the ability to analyze sales calls and suggesting follow-up actions — like when to schedule another meeting, or whether professionals should spend more time listening during their next call.