AUTHOR: Kritin Sundaram
DATE: 11 December 2025
I’ve spent the last few months speaking with many Partners, General Counsels, Innovation teams and senior Investigators across the industry. Whether we meet over a coffee or catch up at an event, the conversation inevitably pivots to the same topic; namely, the immense pressure to adopt AI.
The mandate doesn’t just come from C-Suite personnel, but from pressure across the industry to remain present and keep with the times – whatever that might mean. It is clear, however, that there is a big opportunity to use AI in investigations to optimise efficiencies in data handling, workflow automation and timescale.
In these conversations, I’ve noticed a similar theme. There is a fundamental misunderstanding of where the inefficiency actually lives, as applying generic “AI” to different workflows does not necessarily give you a smarter, more efficient system. There is a rush to buy tools which can be used for many use cases, but a hesitation to focus on what is actually needed to fix the plumbing – and this hesitation, this focus on the gadget over the groundwork, is the LegalTech Trap.
To understand how to get a real ROI from LegalTech platforms, I often discuss with these two distinct ways of operating their matter workflows: Model A (Disconnected Processes) and Model B (The Singular Process).
Model A: The Disconnected Reality
Model A represents a lot of legal and compliance teams I see today. The “disconnect” here is actually two-fold.
First, there is the departmental disconnect. The reality is that workflows and requirements across in-house legal and compliance teams, and in private practice law firms are fundamentally different. A Corporate team racing through acquisition due diligence has entirely different drivers and requirements to an IP team managing a patent portfolio, a Real Estate team handling lease renewals, or a Corporate Crime & Investigations team conducting an urgent and highly sensitive investigation into improper conduct.
From an Investigations standpoint in particular, there is a workflow disconnect. When a complex investigation is triggered, the workflow itself is broken. An ethics report or regulatory notice received doesn’t flow smoothly into an effective scoping mechanism, which in turn doesn’t speak to a data analysis engine, which doesn’t sync with other incredibly important material from a holistic perspective (such as historic cases, witness testimony, corporate policies/procedures). Add to this the complexity of having multiple individuals managing the multiple workflows, being handled in isolated siloes, there is a significant chance that something doesn’t work right or something fundamental is missed.
Here is the hard truth: these inefficiencies have nothing to do with the lack of AI. They are structural. Applying AI on a piecemeal basis here often creates more inefficiency, not less.
If you simply bolt an AI powered document summary tool or an SPA comparison platform onto one of these components – say, the Corporate Department’s due diligence workflow – you haven’t necessarily helped the Ethics & Compliance team as their workflow is totally different, they can’t use that tool to spot fraud or analyse millions of emails to identify evidence of improper conduct.
If you inject AI into just one part of a workflow, you fracture the process and force teams to duplicate data transfers or constantly context-switch between environments. You don’t have the ability to take context from one workflow into another, and you’re just moving the bottleneck somewhere else. Perhaps you add 10% efficiency to a specific task, but you add 20% in friction to the overall workflow.
Model B: The Power of Centralised Investigations
Now, look at Model B. In this model, the priority is centralisation, and finding a point solution which tackles a specific need.
Model B acknowledges that processes should not have different data truths. By moving to a single interface environment where investigation workflows are integrated, you allow different workflows to exist on top of a unified data foundation and resolve a massive chunk of inefficiencies before you even start.
Once you have a centralised data environment, AI transforms from a simple productivity gadget into a strategic powerhouse. Why? Because the AI can now “see” the full picture of the investigation. It isn’t just analysing a static batch of PDFs or emails; it is drawing on a unified data source that includes initial ethics reports, raw data, forensic financial review results, source enquiries, witness testimony, historic case data – the list goes on.
To achieve a truly transformative ROI, we need to stop thinking of AI adoption as merely buying a new tool. It must be viewed as a two-fold process: 1) re-engineering investigation workflows to be integrated and centralised on one platform as transformation is most effective where end-to-end requirements are addressed; and 2) deploying focused tools (whether powered by AI or not) on top of that unified foundation to cover the remaining inefficiencies and uncover insights hidden by volume. If you skip step one, you are merely automating bureaucracy.
The Risks
While standard businesses measure failure in lost pounds or dollars, legal and compliance teams measure it in regulatory fines, sanctions, criminal liability, damages claims, and reputational issues.
The stakes of staying in the disconnected “Model A” are incredibly high. A vast amount of missed red flags during due diligence, failures to spot internal fraud, or spoliation of evidence occurs not because of malice, but because crucial information got lost in bureaucratic handoffs.
Let’s picture an old school investigation method. An investigator downloads crucial evidence onto a local laptop to analyse it because their central tools are too slow or disconnected. That laptop breaks. They migrate to a new machine, but don’t realise they missed one specific file—a crucial piece of exculpatory evidence or a smoking gun email—which failed to transfer back to the record. In a disconnected process, that file is gone. The chain of custody is muddy. The workflow broke. These are real legal risks that arise constantly when you are not integrating workflows in a centralised manner.
So, what’s the verdict?
A single interface solution designed for a specific workflow and process is the safety net for compliance; AI is the accelerator for discovery. We need to stop treating AI as a magic wand that works in isolation.
To get real value, you must deploy smart solutions within specific, end-to-end workflows. Whether you are running an M&A transaction or a regulatory investigation, the AI needs to live where the work happens, by connecting the dots from inception to reporting.
Don’t just buy a “smart” tool. Build a smart process. Integrate your workflows so that when you do switch on the AI, it empowers the entire team, not just a single silo.








