MIT's 95% of AI Fails statistic makes one thing clear: LegalTech Must Go Workflow-Specific like Emma

The MIT study: The GenAI Divide STATE OF AI IN BUSINESS 2025 from Project NANDA, found that 95% of AI projects fail to deliver measurable ROI. Let’s consider what that means for AI in LegalTech and more specifically, for Legal Due Diligence software in M&A.

In law, and especially in M&A, a failed AI pilot isn’t just a financial setback; it can jeopardize deals, overlook red flags, or expose hidden liabilities. In short, failure is not an option. It’s tempting, then, to ask: why take a one-in-twenty bet that your LegalTech investment will pay off? But that may be the wrong question. The real issue is why most initiatives stall and how you can be part of the minority that truly succeeds.This article explores how, with strong adoption and the right tools, AI can be genuinely transformative in legal and M&A work.

The GenAI divide: Adoption is high. Transformation is low.

Interestingly, among the top three sectors poised to benefit most from GenAI, professional services, which includes legal work, stands out. However, unlike Technology and Media, which occupy the first and second spots on that list, professional services face a different challenge.

In Tech and Media, GenAI is primarily used for marketing-focused tasks, content generation, customer engagement, and creative production. In contrast, professional services aim to use AI to reduce administrative, repetitive, and low–value-added work, which is far more complex to automate.

This helps explain why the professional services sector has been slower to adopt AI yet is expected to see a deeper, more transformative impact once adoption takes hold.

When companies attempt to scale AI into their core processes, most pilots stall. Why? Because generic tools often fail to integrate with existing workflows, lack the ability to retain critical context, and can’t handle the real-world complexity of professional work.

What the 5% are doing differently

The 5% of AI projects that succeed share a pattern and interestingly, this is exactly what Legal Due Diligence software like Emma does.

• Workflow-specific systems: Tools designed for specific niches like due diligence.
• Integration first: AI embedded into existing processes, not bolted on as an experiment.
• Learning loops: Systems that adapt from use, improving precision and relevance over time.
• ROI defined in outcomes: Treat AI initiatives like business services (similar to consulting/BPO) rather than off-the‐shelf SaaS. Hold vendors accountable to business outcomes (P&L, cost-savings) rather than just model benchmarks.

Interestingly, external partnerships (vendors): are twice as likely to succeed versus purely internal builds. The AI winners thus choose vendors with real mileage in a particular area and a vertical focus, not just generic AI capability.

How Emma Legal bridges the divide

Emma Legal was built specifically for legal workflows in M&A. From day one, the focus has been deal certainty, not vanity demos.

• Workflow-specific systems: Emma is designed for Legal Due Diligence only
• Integration first: We connect with other tools that you use already in M&A. Think of Data Rooms, Document Management Systems and Closing Lists software.
• Learning loops: With each Legal Due Diligence that you run through Emma, your firm gets smarter. 
• ROI defined in outcomes: Deal-based pricing (per data room connected) and ability to allocate precious time to higher value dealwork that clients are willing to pay for.

We breathe Legal Due Diligence and bring expertise from a wide range of firms that are struggling with exactly the same problems as your firm.

Being on the right side of AI history

The MIT study is a wake-up call. AI is not automatically transformative. Success in law and M&A isn’t about adopting AI first but about adopting it right.

That means choosing tools with proven fit, demanding ROI tied to outcomes, and training teams to use AI as an accelerator of judgment, not a replacement for it.

Emma Legal exists to make sure firms are on the right side of that divide: among the 5% of AI initiatives that actually succeed, not the 95% that don’t.

Because in M&A, there is no margin for “good enough.”

Get in touch to explore Emma's powerful AI.