AI CAN DRAFT YOUR CONTRACT. IT CAN'T DEFEND IT IN COURT.
Why experienced legal judgment still decides how contract disputes are won and lost — a practical guide for ZS Attorneys clients.
Every week, a client asks us a version of the same question: "We ran this contract through an AI tool and it looked fine — do we still need you to look at it?" The honest answer is yes, and this briefing shows you exactly why — not in theory, but through the kind of clauses, disputes and courtroom outcomes we see in practice.
The Two Kinds of "AI Lawyer" You'll Encounter
Not all AI tools are the same, and the difference matters when you're deciding how much to trust one with a contract worth protecting.
AI chatbots (e.g. general-purpose assistants) read a clause and suggest wording. Useful for a first pass, poor at judging consequences.
Agentic AI systems follow a configured checklist across a whole contract suite. Faster and broader — but only as good as the risk framework a human built into it, and still unable to weigh commercial relationships or courtroom exposure.
A ZS Attorneys lawyer does what neither can: interrogates the clause against your business, your counterparty's leverage, and how a court will read it years from now.
How the three compare on what actually matters
Where This Plays Out in Real Disputes
These composite scenarios reflect the pattern of disputes our team sees when a contract was cleared by an AI tool without an experienced lawyer testing it against real-world consequences. Names and identifying facts are illustrative, not case citations.
Scenario 1 — The Indemnity That Wasn't
"The AI flagged the indemnity clause as present and adequate."
A supplier's contract included a clause indemnifying the client "against all third-party claims." The AI tool confirmed the indemnity existed and moved on. When a data breach caused internal remediation costs — not a third-party claim — the client discovered the indemnity never covered that loss at all. The gap surfaced only in a dispute, not at signature.
What a ZS Attorneys review adds: we test what an indemnity excludes, how its limited, whether it speaks to the risks and not just whether one exists — before it is signed, not after it is litigated.
Scenario 2 — The Channel Conflict Clause
"The AI saw no issue with the distribution terms."
A supply agreement permitted the supplier to "contract directly with end customers in exceptional circumstances." Read in isolation, the wording seemed reasonable and the AI raised no flag. In practice, it opened the door to the supplier bypassing the client entirely on its largest accounts — a commercial risk no keyword search was built to catch.
What a ZS Attorneys review adds: we read clauses against the commercial relationship they sit inside, not just the words on the page.
Scenario 3 — The Data Clause That Was Technically True
"No third parties will have access to your data" — confirmed accurate by the AI review.
The clause was factually correct in isolation, yet the client's own cloud infrastructure provider held the data throughout. The AI tool had no visibility into the client's actual vendor stack, so it could not flag the disconnect between the wording and the operational reality.
What a ZS Attorneys review adds: we ask what your business actually does — not only what the page in front of us says.
The Risk Hiding Beneath the Wording
This is the pattern behind all three scenarios above: the words on the page are only ever the visible tip of what a clause actually commits you to.
The "AI Productivity Paradox" — What It Costs You
Economists call this the productivity paradox: powerful technology that doesn't translate into measurable time or cost savings, because the errors it introduces have to be found and fixed by someone else — usually at a later, more expensive stage.
In contract work, that "someone else" is your lawyer, and the "later stage" is often after you've already relied on flawed terms commercially, or worse, after a dispute has started.
AI-only drafting can feel faster and still take longer to reach a contract you can safely rely on.
A Word on AI "Hallucinations"
AI tools can generate citations, clauses or legal principles that sound authoritative and simply do not exist or do not say what the tool claims. This is not a rare glitch — it is a structural feature of how these systems generate text, and it has already caused real, public embarrassment at government level when unverified AI-generated references made it into an official policy document without adequate human sign-off.
We flag this deliberately: if an AI-assisted draft cites a case, a statute, or a regulatory provision you haven't independently verified, treat it as a draft assumption — not a confirmed fact — until a lawyer has checked it.
The Bottom Line
AI is a genuinely useful assistant for volume and speed. It is not, and is not close to being, a substitute for a lawyer's judgment about what a clause will mean when it is tested — in a negotiation, in a dispute, or in court. Contracts are instruments of judgment, and judgment carries consequences that only an accountable, experienced human lawyer can own.
For a practical review of your current contract suite — or a second opinion on anything an AI tool has already cleared — contact: Zinta Strydom, Managing Director, ZS Attorneys
Comments