Q&A With Anthony Usher: Expertise at the Intersection of AI and High-Stakes Investigations

Anthony, you’re a core part of our artificial intelligence innovation group. Will you provide an overview of the work you’re doing with our teams and clients?

I collaborate globally with teams to develop and implement AI solutions, with a focus on disputes and investigations. This subject is close to my heart, as I spent most of the past two decades working in intelligence and investigation in the public sector. I joined FTI Technology from His Majesty’s Revenue & Customs, where I was part of the senior leadership team overseeing the Fraud Investigation Service. 

The Fraud Investigation Service is the part of HMRC that deals with the most serious tax fraud and economic crime cases, including criminal investigations, comprised of approximately 5,000 counter-fraud specialists and other professionals (e.g., cybercrime investigators and forensic accountants). I have firsthand experience of leading the teams focusing on serious and organised crime, who are in the field carrying out dawn raids and also tackling risks and threats. 

Having come from an environment where sifting through huge data volumes is routine and the stakes are high, I’ve been keen on finding ways to leverage generative AI to accelerate key investigative steps, in addition to using the technology as a great critical friend — e.g., deployed as a layer of quality control to mitigate the risk of something being missed or simply to test an argument or theory.

Day-to-day, I work with our in-house software engineers and our specialist e-discovery data scientists, where we collectively explore the very latest generative AI capabilities and test validation confidence for robust and defensible workflows. To date, these efforts have resulted in the use of IQ.AI on more than 300 matters to solve client challenges. 

There’s a lot of noise in the market right now, especially around AI. What’s different about FTI Technology’s approach? 

The sheer calibre of expertise here is unmatched. All our consultants are hands on and actively involved in the development of new solutions. We have delivered so many complex and high-profile matters using IQ.AI and other tools; that vast array and proof of our experience provides deep confidence to our clients. This also takes the form of countless long-running client relationships, with many of the world’s top law firms and largest corporations returning to work with us again and again. 

On a personal level, the quality of the experts was one of the many factors that drew me here. Even for individuals who are extremely knowledgeable in their respective fields, there is so much to learn from working in blended teams with wider expertise. We see this when we work alongside our clients too. The way we have developed IQ.AI Studio makes it easy for them to codify their expertise into custom AI workflows. On a recent internal investigation, the lead investigator included specific questions they knew to have been critical in previous similar fact-finding efforts in their organisation, helping to progress lines of inquiry.

Can you share more about how these differentiators play into the AI solutions?

Unlike many others, our AI offerings are much more than just the technology. We bring the most comprehensive and flexible solution with IQ.AI, which encompasses leading edge AI and generative AI, through our own proprietary solutions and numerous third-party technology providers. We’re platform agnostic, so that we can easily offer our clients a solution using the best technology for the matter at hand or to meet any specific requirements the client may have. Layered with that are the humans in the loop, which is pivotal, as our subject matter experts have advanced knowledge in the relevant field. 

Additionally, we’re building on years of tried-and-tested approaches to offer AI-augmented workflows and analytic solutions in creative and effective ways. This is important to me in the context of investigations because, in my experience, while there might be common stages and steps, there’s no one-size-fits-all approach. We will tailor our workflows to fit the scenario in question, taking all necessary steps to ensure accuracy and defensibility, including using the most appropriate measure of effectiveness for AI tasks, so we can confidently stand behind the results. We will also focus our use of AI where we can unlock the most value, For example, we used generative AI to provide detailed descriptions of photographs, analysing 50,000 images over a weekend, where there simply wasn’t the time or the resources to do it manually in this case. 

Each investigation will have its own challenges and it’s rarely a push button exercise. I have seen matters where a single piece of information can impact the whole case — either positively or negatively — such as difficult-to-decipher handwritten notes that explain fraud methodologies that might not be immediately clear or nefarious activity captured on mobile devices that wouldn’t necessarily be picked up by search terms. IQ.AI Studio has been developed to be highly customisable and the in-built capabilities enable legal teams to pick the right capabilities for the job.

I sometimes talk about the need to be fast alongside the need to be thorough. There isn’t always a smoking gun,  although we can use AI to rapidly surface insights to find key materials, like interrogating data with natural language queries to quickly pinpoint documents of interest. In an IP theft or trade secrets case, it may be sufficient to find just enough relevant documents to take protective measures and consider whether a wider investigation is warranted.

What’s an example of an interesting projects you’ve worked on recently?

We had a client that wanted to detect anomalies and identify supplementary information against a version of events upon receipt of new material in a case. With our ability to fully customise prompts and include project-specific information to provide the model with additional context and help it make the right prediction, we were able to set out a bespoke workflow for this. This is not new to us: we’ve worked on other matters where we have compared witness statements for anomalies and analysed expert testimony using generative AI to quickly make sense of incoming materials. 

I also recently consulted on an internal investigation, in which I worked with one of our data scientists and leaned on my previous experience to help ask the right questions of the material. That approach will now be the starter prompt template for future compliance investigations. 

Generally, as someone who has spent a significant part of my career trying to uncover things people are trying to hide, I enjoyed the first time our IQ.AI Studio was used to identify coded conversations and off-channel communications. Due to the model’s understanding of language, and thanks to the billions of data points the models have been trained on, they are great at identifying unusual activity in communications, including in chat data.

In your view, what are the biggest barriers facing clients right now?

I am not just saying this because my role is AI-focused, but generative AI is both exciting and disruptive. As mentioned, there is a lot of noise and people have differing experience levels, which is compounded by organisations being under pressure to use AI and achieve efficiencies. In parallel, firms are trying to put the right governance and risk management wrapper around their AI work, so that they can protect themselves and their teams. 

The pace of change is phenomenal, with new developments constantly. We want clients to know we can help with these challenges. 

Do you have any predictions for where this is all headed in the coming year?

There will continue to be interest in agentic workflows and I am particularly excited about this, as I’m currently working with our data scientists to develop and test specific agents in investigations settings. I would have loved to have been able to deploy technology like this in the past, kickstarting an investigation by being able to quickly understand the relationships between key players and surface insights. 

In the year ahead, I expect there to be an uplift in more people viewing these agents as additional team member that can carry out advanced tasks with a suitable level of supervision and report back for further instructions. Overall, the use of AI on projects will continue to increase. People have become more comfortable with some of the potential pitfalls, such as hallucinations, but are accepting certain tradeoffs given the technology’s advantages.

Related topics:

The views expressed herein are those of the author(s) and not necessarily the views of FTI Consulting, its management, its subsidiaries, its affiliates, or its other professionals.