Jylo adds redlining

Deterministic vs Probabilistic AI in Legal Workflows

One of the biggest hesitations lawyers still have around AI is the sense that “it won’t do this as well as I can” or “I already have a tool that does this better.”

And in some cases, they’re right. Not all legal tasks are created equal, and not all should be left to probabilistic AI models.

Large language models (LLMs) excel at producing probabilistic outcomes: outputs that combine context, pattern recognition, and what feels like “creativity” to generate text, analysis, or summaries. That makes them powerful in areas where flexibility and judgment are useful, for instance, drafting a first-pass witness statement, summarising case law, or suggesting negotiation strategies.

But in other parts of the workflow, creativity is not encouraged. Think about redlining contracts for example. Here, lawyers need a deterministic tool: a system that delivers the same, predictable result every time, with no surprises.

The reality is that modern legal practice requires both. A typical matter may start with a deterministic step – think automating the production of standard forms or NDAs – before moving into a more flexible, probabilistic stage, where lawyers want options, nuance, and perspective that an LLM can provide.

The real challenge, then, is not choosing between deterministic and probabilistic AI, but understanding where each belongs in your workflow.

In comes Jylo. With our new redlining tool, we’ve designed our platform to let lawyers do exactly that: combine a deterministic, predictable redline tool with probabilistic AI. Run your documents through the redlining tool in our AI Assistant, and then ask gen AI to summarise the changes, put them in a letter format to explain the impact to your client, structure them in a table, a bulleted list, a templated report – the list is endless.

Great lawyers don’t choose between precision and creativity. So they should choose a tool that delivers both.