AI impersonation scams target Minnesota's job market: fraud could reach $40B in 2027

By GovTech / Star Tribune · June 24, 2026.

Workers at Doherty Staffing Solutions, a Minneapolis-based employment agency, began receiving puzzling calls from people responding to recruitment text messages the company had never sent. The subsequent investigation revealed that scammers were impersonating the firm to send fake job offers to job seekers, aiming to obtain their Social Security numbers and banking details in exchange for jobs that did not exist.

The Doherty case illustrates an accelerating trend: impersonation scams—where criminals pose as legitimate companies or government officials—are proliferating at an unprecedented rate, driven by generative artificial intelligence. According to data from the Federal Trade Commission (FTC), this type of fraud cost consumers $3.5 billion last year, a figure that has nearly tripled in five years. Roughly $1 billion of that came from the impersonation of legitimate businesses, frequently banks or government officials.

What makes the Doherty case especially sophisticated is the use of AI-generated images. The fraudulent messages included high-quality photographs of smiling women wearing blazers, all identified by the name 'Emma Smith,' accompanied by the company's real logo. An outside consultant carried out a computer forensic analysis that confirmed the images had been produced with artificial intelligence. The fraudulent offers promised remote positions that required U.S. residency and a valid Social Security number—an especially valuable data-extraction vector.

Beyond the images, what Billy Doherty, the firm's president, highlights is the personalization of the scams: over the past month, the fraudsters tailored their messages to each individual's specific job interests, using public information gathered from the internet. 'On its face, this looks fantastic,' Doherty said. 'It looks like something we did with deliberate effort to put in front of you.' This level of personalization marks a qualitative leap from the mass campaigns linking to a fake site that characterized earlier scams.

Professor Manjeet Rege, of the University of St. Thomas and director of its Center for Applied AI, sums up the problem clearly: 'Generative AI has reduced the cost of producing a convincing deception. Even a website can be put together in a matter of minutes.' Rege notes that the traditional advice of 'seeing is believing' no longer holds: 'I would now say, don't believe it even if you see it, even if it looks very, very authentic.'

In Minnesota specifically, the FTC recorded 33,204 fraud complaints in total last year, with losses of approximately $168 million and a median loss of $340. Impersonation scams accounted for about 37% of all reports. The state's labor market also suffered an additional impact in the wake of the federal government's immigration enforcement campaign last winter, which generated greater vulnerability among workers actively seeking employment.

Temporary staffing agencies are a strategic target: in 2024, according to the American Staffing Association, these firms employed approximately 247,000 workers in Minnesota in temporary jobs in manufacturing, transportation and building maintenance. Stephen Dwyer, the association's CEO, warns that the sharp rise in impersonations over the past year could cause lasting reputational damage to the sector: 'They believe it's a fruitful way to scam innocent people.'

The national scale of the problem is significant. The FBI received nearly 22,000 complaints about AI-enabled scams last year, with losses in that category of $893 million nationwide, within a total of nearly $21 billion in fraud attributed to computer crimes. The agency highlighted cybercriminals' growing ability to clone voices or craft 'official-looking emails,' making the scams increasingly difficult to detect. Christopher Mufarrige, director of the FTC's Bureau of Consumer Protection, called impersonation scams 'one of the most pernicious forms of fraud.'

Looking ahead, the consulting firm Deloitte estimates that deepfakes and other advanced technologies could push the total annual cost of consumer fraud in the U.S. up to $40 billion by 2027. This projection underscores the speed at which AI is transforming the digital fraud ecosystem, shifting from an occasional tool to a structural engine of online organized crime.

From the perspective of the psychology of fraud, researcher Marti DeLiema, an assistant professor of social work at the University of Minnesota who studies why people fall victim to scams, introduces the concept of 'unmet need' as a vulnerability factor. According to DeLiema, the conversation about susceptibility to fraud usually centers on age, cognitive decline or technological know-how, but unmet need is equally decisive: 'A person who has an unmet need, such as the need for employment, might be more open to interacting with potential offers they see online.' This perspective is relevant to understanding why the labor market is such an effective attack vector: it combines economic urgency with the routine use of digital platforms for job searches.

Rege, the St. Thomas professor, further warns that looking for visual cues to identify AI-generated content—blurry logos, odd phrasing, spelling errors—is a losing strategy in the long run, since those indicators are disappearing with newer models. He defines the situation as 'a cat-and-mouse game': 'By the time a detection method becomes reliable, the generation tools have already advanced to something it can't capture.' This gap between detection tools and generative models is one of the most pressing technical challenges in the field of digital security.

For local and state governments, the Minnesota case has direct public-policy implications. Public agencies are also a frequent target of impersonation—37% of complaints in the state involve this type of fraud—and the proliferation of fake job offers invoking government agencies or staffing-sector companies can erode institutional trust. AI's ability to generate communications indistinguishable from legitimate ones raises questions about what technical and regulatory safeguards should be implemented in official channels for communicating with citizens.

In general, as sector context, the phenomenon described in Minnesota is not an isolated case but part of a global trend: the use of language and image models to automate and personalize fraud at scale. What once required criminal operations with significant resources—graphic design, convincing writing, managing multiple identities—is now accessible with commercially available, and in some cases open-source, AI tools. The barrier to entry for mounting a sophisticated fraud campaign has plummeted, while detection becomes progressively harder for end users.

Ultimately, the case of Doherty Staffing Solutions in Minneapolis condenses several of the most worrying vectors of misapplied agentic AI: automatic generation of fake identities (images, names, logos), mass personalization based on public data, and at-scale deployment of deception campaigns targeting vulnerable populations. The institutional response—public awareness, coordination among affected companies, complaints to the FBI and FTC—remains reactive against a threat that evolves faster than the available detection and regulatory frameworks.