When you click on a sponsoring school or program advertised on our site, or fill out a form to request information from a sponsoring school, we may earn a commission. View our advertising disclosure for more details.
Don’t think of your skills as irrelevant because AI can do it. That would be the biggest mistake. Those skills are now much more important.”
Anurag Jain, Financial Forensic Investigator and Co-creator of RiskPulse.ai
In December 2024, the Department of Justice released a landmark report on artificial intelligence and the criminal justice system, commissioned under a presidential executive order and informed by a year of roundtable conversations with law enforcement agencies, civil society organizations, AI developers, and academic researchers. Its conclusion was straightforward: AI use is rapidly transforming the criminal justice system and has the potential to make it more effective, equitable, and efficient, but also has the potential to cause harm, amplify disparities, and misdirect resources.
Students considering a career in forensic investigation are entering a landscape that has changed drastically in the past few years. While AI in criminal justice has historically relied on conventional statistical analysis, the accelerating pace of AI innovation is now driving greater use of computer vision, natural language processing, and generative AI across the systems.
More and more organizations are deploying these tools in active investigations. The policy and technology choices being made right now, the report warned, will affect millions of Americans and set the trajectory for how AI is used in criminal justice for years to come.
“Over the last two to three years, I’ve been working at the intersection of financial forensic investigations and machine learning and AI,” says Anuran Jain, financial forensic investigator and co-creator of RiskPulse.ai. “Investigation is unlike any other field, where you’ve got to look at the evidence, you’ve got to look at the analysis, you’ve got to contextualize a lot of things before you can arrive at something. What AI is doing is helping us contextualize a lot of information that was previously not possible. And that’s what makes this technology very interesting, particularly for investigations.”
Keep reading to learn more from Mr. Jain on how AI is presently being used, the danger of hallucinations, how it may hold up in court, and what forensic education may look like going forward.
Meet the Expert: Anurag Jain
Anurag Jain is a financial forensic investigator and entrepreneur with more than a decade of experience in financial crime compliance, fraud risk management, and anti-money laundering investigations. He is the founder and CEO of RiskPulse.ai, an agentic AI platform built for financial crime investigations, KYC/AML operations, and compliance functions at banks and financial institutions.
Jain is also the CEO of EntityVector, an advisory and analytics firm serving banks, fintechs, and regulatory bodies. He also serves as a Principal and Advisory Board Member at Global Economics Group and as an advisor on financial fraud investigations at Van Dermyden Makus Law Corporation. He holds MBAs from the Schulich School of Business at York University and the Indian Institute of Management Bangalore.
What AI Actually Does in an Investigation Today
To understand how AI is reshaping forensic work, it helps to start with a distinction Jain makes right away: the difference between an AI tool and an AI agent: “There’s a difference between AI and an agent,” he explains. “A simple tool is where you enter the text and your prompt, and you get a response. The agent is one step further. I can do stuff for you. It can make a call to something, pull the data out, and respond to a certain thing.” The platform he has built, RiskPulse.ai, operates as an agent, meaning it does not just answer questions; it acts on them.
That distinction matters enormously in practice. Jain describes how an agent can be directed to retrieve documents, conduct research, send requests for additional information, and then synthesize everything it finds into a structured report, handling what was once an entire investigation workflow. “If you take a step back and ask what investigation really entails, there are four or five key elements,” he says.
“You’ve got to understand the core requirements of the case, do a complete analysis in terms of what information you need, what documents you have, what additional documents you need, and where you’re going to source them. Sometimes it requires additional research. This step can be done very effectively by the agent. And then it takes the results from those steps and builds a report.” The payoff, he adds, is hard to overstate: “This whole thing, which used to take weeks, is now done almost instantaneously.”
Accuracy, Hallucinations, and Bias
Accuracy is what matters in forensic work. Work done quickly doesn’t matter if it isn’t reliable. The good news, Jain says, is that the technology has improved dramatically in a short time. “It’s getting better. Over the last 12 months, it has improved significantly. My clients are amazed at the kind of output the tools are producing now, but it was not the case before. I had to do a lot of work to curb hallucinations. I had to really enforce that it be very fact-oriented, only work on the data that I provide.” Today, he says, the feedback he receives puts accuracy around 95 percent.
That remaining 5 percent, however, is still significant. In forensic investigations, errors have consequences for cases, clients, and potentially the people whose lives those cases affect. “It’s not 100 percent, and I don’t think it’s ever going to be 100 percent,” Jain says. “That’s where the human in the loop is so important.”
Bias is an equally important concern, and one that requires investigators to bring genuine expertise to their review of AI-generated work. AI systems are trained on data, and that data carries its own assumptions and blind spots. “AI has its own biases,” Jain explains. “You’ve got to look at it and understand where those biases are coming in, and then try to figure out how to resolve them. At the end of the day, these models are optimized for probability. They may be very good at producing output, but they’re not necessarily optimized for the right judgment.” That distinction, he argues, is exactly why skilled human judgment remains irreplaceable at the center of any investigation.
Will This Hold Up in Court?
The courtroom is the ultimate test for all forensic investigations. And right now, AI-generated analysis is still working its way toward that standard. Jain sees the path forward as much an oversight question as a technology one. “The key question is more about governance,” he says. “How are the systems being managed? Where are the sources of information coming from? Can you explain the assessment and analysis the agents are producing? And can the human investigator step in, make changes, or remove findings entirely? As long as that is there and it’s being managed and governed properly, I think the courts will begin to accept it.”
Legal and law enforcement institutions are keen on this new technology, but acceptance will take time. “The interest is there, and everybody is using it, but it is going to take some time for acceptance from the courts and law enforcement agencies to come through,” Jain says. In the meantime, one principle holds firm regardless of which tools an investigator uses: “Humans are always accountable, no matter what AI agent they use or not,” Jain warns.
The Mindset Shift Students Need to Make
To excel in this field, students will approach their work and studies differently from their predecessors. “You need to think differently. AI is not going to go away, for sure, so stop resisting it and start adopting it in ways that will help you in your profession. Start bringing it into your day-to-day investigation work. Start educating your clients on how you’re using AI to improve the investigation process. The students who approach this with the mindset that it’s going to take their job away are going to find it very difficult to survive. The ones who ask, ‘How can I make my job better with the use of AI?’ They are going to succeed,” predicts Jain.
“The skills don’t change. Critical reasoning, evidence analysis, ethical judgment, documentation, cross-domain expertise, those core skills will not go away,” he continues. “In fact, I would say those skills are now required more than they were in my time. Because earlier, if I did something, only my boss or manager would look at it. Today, a machine is producing something that might be better than my own investigation skills, and I have to judge it. I have to make sure there are no biases. I have to make sure that the evidence is completely contextualized and the sources are transparent. You have to be 10 times better, because you’re going to analyze and assess the quality of the output.”
What’s Happening to Entry-Level Jobs
The disappearance of traditional junior roles is one of the more uncomfortable realities of the AI era, and Jain does not sidestep it. “Junior-level jobs are almost non-existent right now,” he says. But he does not see that as the end of the story.
“I think there will be a tipping point where organizations need more resources to work around and manage the AI better. What’s happening today is that every organization has given its employees access to ChatGPT and co-pilots. As a result, everybody is creating their own prompts and producing their own outputs, which is creating a kind of chaos. Everybody is moving in their own direction, depending on the prompt and context they’re writing. That creates a unique requirement for junior people to come in and help establish that management principle, and then move forward to apply those skills in the day-to-day job.”
Big employers are already hiring for this new version of entry-level work. “Some of the biggest consulting firms today, I won’t name any, have changed their interview process,” Jain says. “They give the ChatGPTs of the world to their applicants and ask them to write prompts. They are hiring people who are best at writing prompts. Maybe that’s the skill set needed.”
What Forensic Education Should Look Like Going Forward
For students evaluating programs, Jain has a clear benchmark: look for curricula that put AI tools directly in your hands. “In my time, we used to take a day, sometimes hours, to analyze investigation cases on our own. But gone are those days,” he says. “Those skills are still required, and we need to do some of that in a classroom setting. But more importantly, I would say, give students the AI tools and ask them to use them to do the investigation. Integrate that into the curriculum. We’ve got to train our students to use AI ethically. A very strong sense of ethical judgment is critical.”
The Bottom Line for Future Investigators
Jain’s message to students is part reality check, part encouragement. “Students need to look beyond the hype and into the depth of what AI actually is,” he says. “These models are optimized for probability, not necessarily for the right judgment. That sort of drives the point that you’ve got to look at AI differently than how the world is looking at it. Don’t be scared that you’re going to lose your job. You have to develop the skills to work around it, to work with it.”
He continues, “The skills you are learning in college right now, don’t think of them as irrelevant because AI can do it. That would be the biggest mistake. Those skills are now much more important because you need people to validate what the AI is producing, to determine whether it is right, wrong, biased, or ethically correct. There will always be a need for people who are exceptionally skilled in those basic concepts to use AI effectively.”