Predicting the arbitrator: how a Geneva counsel ran a complex case through Lexity, and won

In our In Practice series we sit down with the lawyers who actually use Lexity day-to-day. Not to demo features, but to ask what changed in their work, where they trust the platform, and where they don't.

First up: Nilofaur Balouch, an arbitration specialist in Geneva and a Legal Tech Advisor at Bridge Legal Tech. She walked us through a real case she ran through Lexity, and what the platform got right.

Executive summary

• A multi-party arbitration with 16 years of paper documents, scanned, uploaded to Lexity, mapped in minutes instead of days.
• Lexity's preliminary success assessment came in above 50%. The final award matched both the outcome and the reasoning.
• Time saved didn't go into more cases. It went into Danish-law research and risk management. That's where the case was actually won.
• Nilofaur uses AI for finding, summarizing, and extracting. She does not use it for legal drafting, analysis, or judgment. The boundary is deliberate.
• The single task she will never do manually again: data extraction.

Meet Nilofaur

Nilofaur is a cross-border lawyer holding an LL.M in International Business Law and Innovation. She splits her time between two roles: Of Counsel at SMBC Law Firm Geneva, where she works on arbitration, mediation, and dispute resolution across trade, technology, and e-commerce; and Legal Tech Advisor at Bridge Legal Tech, where she advises law firms and legal departments on adopting AI and digital tools.

That double mandate matters. As she puts it: "I have a tech background, so I'm more optimistic about AI than maybe a normal lawyer, but I keep a pragmatic view as well." She is not a believer. She is a tester.

Before AI: a room full of paper

The case Nilofaur walked us through is the kind of arbitration that gives senior counsel a headache. The relationship started in 2010. Multiple corporate entities on each side. A chain of contracts and amendments. Applicable law in a jurisdiction the firm doesn't sit in. And a client who arrived with, in her words, "really, I'm telling you this much papers."

The traditional first move is to read everything, twice, and write notes. Nilofaur can do that. She is, in her words, used to working with documents most of the time. But the cost of that first read isn't just time. It's the strategic attention you no longer have for the parts of the case that actually matter.

A case in focus: scanned paper to a strategic map

Nilofaur scanned the paper documents, dropped them into Lexity, and ran the Preliminary Case Assessment Clickflow™. What came back was the map she needed before she could think about strategy: parties, lawyers, arbitrators, applicable law, the financial structure, the agency-agreement timeline, the procedural timeline pulled from the Terms of Reference (deadlines, hearing dates, filing dates), and the contracts-and-amendments chain.

The output was structured, but more importantly it was cited. Every data point on the map traced back to the source document with a click. "In this age of legal AI, citations are essential. Without them it's almost impossible to work."

Then she ran the Success Indication, Lexity's preliminary success indicator. The output came in above 50%, with reasoning attached: a prior similar case had resolved favourably, the respondent had effectively accepted the debt and not counter-claimed, and the procedural posture was clean. "It gave me hope for the case, and it gave me the reasoning."

The case was decided in their favour. "The award came out, and the reasoning was true, exactly how the arbitrator wrote it. The whole idea of the success was correct."

What actually changed: time becomes strategy

The headline isn't the predictive accuracy. It's what Nilofaur did with the hours she got back.

This is the message that's easy to miss in a product demo. Lexity didn't substitute for Nilofaur's judgment. It compressed the part of the work that produces the least differentiated value (mapping facts) so she could spend more time on the part that produces the most (sharpening strategy and pricing risk).

Where AI helps, where judgment stays human

Nilofaur draws the line clearly, and we asked her to draw it on camera so other lawyers can borrow her framing:

• AI is for finding, summarizing, extracting, mapping. Compliance article extraction. Timelines. Party maps. Procedural calendars from a Terms of Reference. "Finding information with AI is amazing. Summarizing is amazing."
• AI is not for legal drafting. "It gives you the structure, but you spend almost the same time restructuring. Lawyers already have their structures."
• AI is not for legal judgment. "I don't ask legal questions of AI. The legal judgment is exactly the boundary where humans can do something AI cannot. Lawyers read between the lines. AI, even with full documents, still can't."

She still always reads the underlying case law and confirms with her own reasoning. The Clickflow™ is a starting point, not a conclusion.

The thing she'll never do manually again

We asked for one task. The answer was instant.

Procedural calendars buried inside Terms of Reference. Party structures across amended contracts. Financial figures across jurisdictional schedules. Anything that's in the documents but not yet out of them.

On trust: ISO, Swiss data residency, and the noise of the category

Nilofaur started the way most lawyers start: she liked ChatGPT but wouldn't put a client document near it. "I had concerns about privacy and security, so I didn't use AI for legal purposes at all, until I found tools like Lexity I could trust."

Two specific things moved her:

1. Attorney-client privilege and professional secrecy are addressed up-front. Documents uploaded to Lexity are not ingested into a shared LLM space. "This is so important. If you don't pay attention to it, it can ruin your reputation. You can't upload your documents to any general LLM."
2. ISO certification, verifiable, not aspirational. "There are a lot of legal-tech tools out there that are just another LLM with a different surface and a very high cost. ISO certification is a real, verifiable security standard."

She's also blunt about the wider state of legal AI. There is, in her words, a "bad impression" of the category among lawyers. Too much hype, too many tools that don't deliver. We didn't push back. We think she's right. The series exists partly because we believe a lawyer's voice is the only credible response to that.

Honest critique

We asked what's still frustrating. She gave us a specific, useful answer, scoped to a particular kind of Clickflow™ rather than the platform overall:

The point lands on the long-form summarisation flows specifically, not on the platform as a whole. Output density on those flows is one of the open product problems we'd rather have flagged on a public blog than buried in a support ticket, and it's actively on the list.

What she'd tell a colleague

Three things, in this order:

1. Talk to the team first. "Lexity has many Clickflows™. If a colleague spoke to you directly, you'd help them find the right ones for their work, otherwise it can be confusing." Fair. We're working on better discovery in the new release. In the meantime, we're happy to do a 20-minute call.
2. Use it for what it's good at, not as a drafting tool. Summarization, timelines, data extraction, internal case assessment.
3. Trust the security posture. "Your data stays in Switzerland. It doesn't go abroad. There's no professional secrecy violation."

What changed about the work itself

We closed by asking Nilofaur what's actually different about her practice now. Not the features, the work. Her answer was about legal research:

That's the version of "AI for lawyers" we recognise. Not replacement. Not autopilot. A faster, cleaner path to the part of the work where a senior lawyer's judgment is the asset.