Generative AI can process and surface information at remarkable speed, but its real power emerges when it’s guided by humans’ investigative expertise.
At Control Risks, as our team has begun deploying Relativity aiR for Review across investigations globally, this theme has become clear. On every matter, it’s the judgment, context, and intuition of seasoned investigators that transform the AI’s results into meaningful insights.
Artificial intelligence is already transforming how investigations are conducted. With it, we can accelerate document review, uncover insights faster, and surface connections that might have taken weeks to find manually.
As AI has exploded onto the scene, we’ve talked about the importance of keeping a human in the loop—someone to guide, test, and validate the work of AI systems. But in the world of complex investigations, that phrase doesn’t go far enough. What’s needed isn’t just a human; it’s an investigative expert in the loop.
“You wouldn’t embark on an investigation without the guidance of a subject matter expert, and the advent of AI doesn’t change that equation. AI tools are most effective with expert guidance from the start, ensuring better prompts and more reliable outcomes.”
— Michele Wiener, Partner, Control Risks
From Human Oversight to Investigative Insight
AI models are only as effective as the context and direction they’re given. When developing prompts for an investigative matter using aiR for Review, the nuances matter immensely.
For example, a whistleblower allegation about revenue recognition requires different lenses than a corruption probe in a foreign subsidiary. Building the right AI prompts, workflows, and quality controls demands more than technical skill—it demands investigative judgment.
That judgment comes from people who:
- Read the whistleblower complaint and understand its implications
- Review internal audit findings and understand internal controls
- Interview accounting or compliance personnel
- Trace transactions and recognize red flags buried in ledgers or emails
In other words, those who have lived and breathed investigations—attorneys, CPAs, CFEs, CFAs, forensic accountants, and former prosecutors—bring an intuition that helps AI focus on what truly matters.
“Based on years of conducting fraud investigations, we've developed a sense of language, behaviors, and other indicators to look for. aiR helps us scale that intuition so we can test more hypotheses, faster.”
— Ben Cohen, Principal, Control Risks, Americas
Crafting Effective Prompts: The Difference Between Good and Not-So-Good
When it comes to leveraging AI in investigations, the quality of your prompt can make or break the outcome. A well-constructed prompt doesn’t just ask the AI to “find relevant documents”—it frames the request with investigative precision, context, and hypotheses.
Poor prompts tend to be vague, overly broad, or disconnected from the substance of the matter, leading to noise instead of insight.
Tip #1: When drafting an effective AI prompt, think of it as creating a knowledge transfer document for a colleague joining the matter.
Structure the prompt to deliver essential context and explain why the information matters. Additionally, specify what tone and behaviors should be identified in the documents under investigation. This helps the AI focus not only on content, but also on the underlying attitudes, professionalism, or intent reflected in the materials.
Tip #2: Think like an investigator drafting an interview question.
You want your prompt to be clear, purposeful, and designed to elicit meaningful information.
The best prompts reflect the substance of the allegation and the investigative theory you’re testing—just like your strategy for interviewing someone involved in the matter.
Take a Closer Look: What makes a good investigative prompt?
These tips can be distilled into a few key components for any high-quality investigative prompt:
- Specificity: Anchored in the allegation or issue at hand.
- Contextual cues: Includes relevant timeframes, entities, and transaction types.
- Investigative intent: Signals what you’re trying to uncover (e.g., patterns of behavior, language indicating concealment).
- Adaptability: Designed to evolve as new facts emerge.
Let’s illustrate these points with a couple of example prompts:
- Good Prompt:
“Identify emails between procurement staff and Vendor X from January to March that reference non-standard discounts, rebates, or unusual payment terms. Key documents will include conversations related to preferential pricing arrangements and side agreements.”- Why it works: It’s targeted, time-bound, and tied to a hypothesis about irregular financial arrangements.
- Not-So-Good Prompt:
“Find all emails about procurement.”- Why it fails: It’s too broad, lacks investigative focus, and will return an overwhelming amount of irrelevant data.
AI + Investigative Expertise = Amplified Outcomes
When investigative experts work hand-in-hand with AI, the results are transformative. AI can surface patterns at scale; investigative experts can interpret those patterns and refine the search logic. Together, they create an iterative loop of discovery, validation, and insight.
This partnership pushes the envelope of what’s possible. It enables investigation teams to:
- Develop targeted prompts that reflect the substance of allegations
- Design review workflows that prioritize investigative relevance, not just keyword hits
- Rapidly validate and refine results based on evolving hypotheses
“In one recent investigation into procurement irregularities, aiR helped us isolate relevant conversations in multiple languages in a fraction of the usual time. It meant we could focus immediately on analyzing intent and motive rather than wading through irrelevant data.”
— TJ Tan, Principal, Control Risks, Asia Pacific
The outcome isn’t just faster reviews; it’s smarter investigations, where AI doesn’t serve as a replacement for human expertise, but a force multiplier for it.
The Next Evolution of Discovery and Investigations
As AI becomes more embedded in our investigative toolkit, the role of discovery professionals is evolving. Project managers, analysts, and investigators are now expected to collaborate across disciplines—combining technology fluency with investigative acumen. Ideal outcomes come when teams bridge these worlds seamlessly.
“What’s exciting about aiR is that it’s teaching our teams to think like investigators and technologists at once. The prompts, the iterations, the validation—they all mirror the way good investigative work has always been done, just faster and with more precision.”
— Adam Garside, Principal, Control Risks, EMEA
The future of investigations lies not in choosing between AI and expertise, but in integrating them. When we put an investigative expert in the loop, we go beyond overseeing AI and toward elevating it. And in doing so, we deliver greater value, insight, and impact for our clients around the world.
From financial misconduct in Mexico to corruption inquiries in Indonesia, bribery investigations in Brazil, or conflicts of interest reviews in China, the combination of local investigative expertise and global technology platforms like aiR for Review is redefining what’s possible.
Wherever the facts lead, the power of pairing AI with experienced investigators ensures we get there faster—and with greater clarity—than ever before.
Graphics for this article were created by Kael Rose.
