From Curiosity to Commitment: My Journey Building the Business Case for Gen AI in Legal

When I joined the recent webinar with Sarah Green from Relativity and Susan Stone from AT&T, I was struck by how far our industry has come—and how much further we can go. As someone who’s spent 20 years at Cimplifi, overseeing everything from product development to forensics, I’ve witnessed the evolution of legal technology firsthand. But the leap to generative AI feels different: it’s not just another incremental change; it’s a genuine turning point.

The Moment We’d Been Waiting For

For years, our business revolved around compliance—staffing large teams of attorneys to review documents at speed. It was effective, but not sustainable. Since 2010, we’ve been searching for ways to move beyond sheer manpower. Technology Assisted Review (TAR) helped, but adoption was slow, and the process remained opaque. When GPT and other generative AI models emerged, it was like flipping a switch: suddenly, everyone was interested, and our long-held vision of fewer attorneys supported by smarter engines started to materialize

Building the Case: Strategy Over Scale

One of the biggest lessons I’ve learned is that building a business case for gen AI isn’t just about cost savings, it’s about being strategic. Many of our corporate clients operate under heavy budget constraints, so predictability and efficiency are paramount.  Our fixed fee AI review offering can provide that budget-oriented approach.  However, gen AI isn’t TAR; it doesn’t just give you a binary answer across a fixed population of documents. It provides rationale, citations, and feedback that empower attorneys to make better decisions in a variety of use cases across cross sections of a dataset. We often recommend an à la carte approach: using AI strategically on segments of your data, validate with human review, and demonstrate value through targeted cost-benefit analyses.

For example, we helped a client identify contracts with hidden fees—not by brute force, but by prompting the aiR for Review to find patterns and expand the population using our other AI tools to surface similar agreements. The result? Significant savings and a more effective review process.

Overcoming Resistance: The Outside Counsel Challenge

Adoption isn’t always smooth. Two years ago, law firms were hesitant to embrace these technologies. Today, many are actively seeking opportunities to use gen AI in repeatable, lower-risk scenarios. Still, some resistance remains, especially when it comes to defensibility and accuracy. Our role at Cimplifi is to guide clients and their counsel through the nuances of prompt engineering, validation workflows, and the effectiveness of rationale and considerations in AI-driven review.

Sometimes, the cost savings and efficiencies are so compelling that corporations will choose partners who are willing to innovate. As Susan noted in the recent webinar, firms that help save $100,000 on a matter aren’t just efficient, they become preferred partners.

Security, Privacy, and the AI Review Board

Security and privacy are top priorities. We advise clients to ensure their prompts, outputs, and embeddings aren’t used to train third-party models or improve external services.  Specifically, it’s important that client prompts, outputs and embeddings:

• Are not available to other clients.
• Are not available to OpenAI (or other).
• Are not used to improve OpenAI (or other) models.
• Are not used to improve any Microsoft (or other) products or services.
• Are not used for automatically improving Azure OpenAI (or other) models for your use in your resource. Models should be stateless unless fine-tuned for a client’s exclusive use.
• Your fine-tuned Azure OpenAI (or other) models are available exclusively for your use.
• The Azure OpenAI (or other) service is fully controlled by Microsoft (or other) and they host the models in their environment. The service does not interact with any services operated by OpenAI (or other).

Opting out of abuse monitoring is also critical to maintaining confidentiality. These aren’t just technical details, they’re essential for building trust and passing the increasingly rigorous AI review boards that organizations are establishing.

Lessons Learned and Looking Ahead

What excites me most is the creativity gen AI unlocks. It’s not just about up-or-down decisions; it’s about using prompts to surface insights, validate results, and empower attorneys to focus on strategy rather than rote review. The technology is evolving rapidly, and so are the use cases—from QC and internal investigations to extraction and targeted document analysis.

My advice? Get involved. Try out the technology, experiment with prompts, and start small using datasets with which you are familiar. The engines powering your everyday searches are the same ones driving innovation in legal. The more comfortable you get, the more value you’ll unlock.

Final Thoughts

The journey from curiosity to commitment isn’t linear. It’s filled with headwinds—security reviews, stakeholder buy-in, and the challenge of scaling adoption. But with the right mindset, strategic approach, and willingness to learn, gen AI can transform legal workflows and deliver real, measurable impact.

Thanks to Sarah Green and Susan Stone for a fantastic discussion, and to everyone who’s pushing the boundaries of what’s possible in legal tech.

Ari Perlstein is chief technology officer at Cimplifi where he leads the design and deployment of secure, scalable systems that power complex legal workflows. With over two decades of experience, Ari specializes in software development, cybersecurity, and data analytics, and has played a pivotal role in advancing Cimplifi infrastructure and AI capabilities. He holds a B.S. in Computer Science from SUNY Binghamton and is a Relativity Master.