AI Redlines, AI Slop, and the Missing Policy Layer in Legal Tech

The AI redline problem

Counterparties are running MSAs and other commercial contracts through AI before sending redlines back, and the AI touches almost every clause in the document, even the parts that were never in question. A legal team that used to get a handful of substantive comments on a template now gets a document marked up top to bottom, most of it wording changes that don't move risk or liability at all. Someone still has to read every line to find the two or three edits that actually matter. A 24-hour SLA for legal review starts looking unrealistic when the volume of markup goes up tenfold and the actual signal in it stays the same.

The AI slop problem

At the same time, a different flavor of the same problem shows up on the drafting side. Ask a generic AI tool to write a contract clause, a policy memo, or a slide, and you get a first draft that reads fine on the surface and says very little underneath. People have started calling this AI slop, and legal and knowledge teams both know the routine: read the output, strip out the padding, go back to first principles, and rebuild the actual point the document was supposed to make. It looks like progress. It creates more work than it replaces.

One root cause

These look like two separate complaints, but they trace back to the same gap. Nobody told the AI what actually matters.

An AI model reading a contract for the first time has no idea which liability cap your company will actually accept, or that a certain piece of boilerplate is something you'd strike without a second thought. So it treats every clause with the same weight, and from a reviewer's side, that looks like noise. An AI model asked to draft something from a blank prompt has no house position and no sense of which two sentences are actually the point. So it fills the space with something generic and safe.

Both problems need the same fix. A policy layer has to sit between the AI and the document, and it has to know your company's standards before the AI touches a word.

How Lexnus builds that layer

Lexnus builds exactly this layer, starting from the contracts you already have. Instead of asking your legal team to write a playbook from scratch, it reads the Word templates and signed contracts already sitting in your files and pulls the standards out of them, like which liability caps you've actually accepted before and which clauses you never negotiate away. That playbook becomes the rule set every contract gets checked against, whether it's one your company sends out, one a counterparty sends in, or one an AI tool drafted first.

On the redline side, this changes what a lawyer actually spends time on. A rewritten clause gets checked against the playbook instead of read cold. If it changes something the playbook actually governs, like an indemnity cap or a governing law clause, it gets flagged as a violation and routed for approval. If it's a wording change that doesn't touch anything the playbook cares about, it doesn't need a lawyer's attention at all. The total volume of redlines coming in doesn't shrink. What shrinks is how much of that pile a person has to read line by line.

On the drafting side, Lexnus doesn't generate contract language from a blank prompt. Every clause an outbound contract uses comes from your approved clause library, the same one the playbook is built from. There's no wall of generic text to cut down to what matters. Someone already decided what mattered, back when the clause got approved for the library.

Reading a contract and deciding whether it complies with your standards are two different jobs. The AI reads the clause and works out what it actually says. The playbook takes that reading and checks it against rules your legal team wrote, then decides pass or fail. That second part has to come from your team's own standards. An AI model has no standards of its own to fall back on.

When a legal team spends more time policing AI output than they used to spend reviewing contracts directly, that's a sign the AI is running with no supervision at all. Give it a policy to work inside, and the redlines and the first drafts both start looking like something a person actually decided on purpose.

Lexnus builds that policy layer from the contracts you already have, live in under an hour, and applies it to every contract your company touches after that, whether it started inside Lexnus or came from somewhere else, a counterparty's redline or a draft an AI tool generated on its own. If your legal team is buried in AI-generated redlines or rewriting AI-generated drafts every week, that's the exact problem this was built to solve.