Last month a Malaysia-based services firm used DocuSign's AI to draft a client engagement contract. The LLM generated clean liability caps—but omitted the entire indemnification clause and invented a renewal pricing term that never existed in the original statement of work. The lawyer caught it on day 6. The contract sat unsigned for another ten days while gaps were manually fixed. AI contract drafting is fast. It is not safe. LLMs hallucinate terms, flatten nuance into templates, and miss jurisdiction-specific rules that regulators and courts expect to see. This is not a theoretical risk; it is a live problem across DocuSign's AI, PandaDoc, and other platforms that promise to save time. The cost of missing one clause can exceed the hours saved by months. Here are the seven gaps LLMs consistently leave behind—what your lawyer will catch if you do not. Gap 1: Invented or Misaligned Pricing Terms LLMs are trained on thousands of contract templates. When drafting renewal or escalation clauses, they often blend terms from multiple sources and generate pricing logic that does not match your actual commercial agreement. We tested DocuSign's AI on a simple service retainer with tiered hours. The AI-generated contract included a clause that said "pricing shall increase 5% annually unless otherwise agreed in writing." The statement of work said "pricing fixed for 24 months." The two contradict. A lawyer will flag this. Your client may refuse to sign. You now need days of back-and-forth to clarify which term governs. PandaDoc's AI performed similarly on a project-based contract: it synthesized a price per milestone but added language suggesting hourly overage billing—a term that never existed in the source. The human who reviewed it had to rewrite the section entirely. Watch for: Any numerical term (price, cap, minimum) that does not appear in your original statement of work or brief. Fix: Strip auto-generated pricing clauses and manually port them from your signed SOW. Use field variables, not generative text. Gap 2: Jurisdiction and Governing Law Flattened to Default Malaysia, Singapore, and Indonesia have distinct contract law frameworks. Liability caps, indemnification scope, and dispute resolution are not interchangeable. An AI drafter trained primarily on US and UK templates will default to English law and London arbitration unless it has explicit instructions to use Malaysian law. We tested this: DocuSign's AI on a Malaysia-based SaaS contract generated a clause reading "This Agreement shall be governed by the laws of England and Wales." The user had not specified any jurisdiction. When you enforce a contract with the wrong governing law, you lose leverage. A Malaysian court may refuse to apply English law without explicit written agreement; you may have to refile in a foreign jurisdiction and pay international counsel. The client's lawyer will spot this on the second read. Watch for: Any reference to jurisdiction, arbitration, or dispute resolution that you did not explicitly request. Fix: Prompt AI drafters with a jurisdiction clause as a required input. For SE Asia contracts, lock in Malaysian, Singapore, or Indonesia law before generation. Gap 3: Missing or Incomplete Indemnification Scope Indemnification clauses protect you when the other party's breach causes third-party harm. LLMs often generate skeletal versions that omit critical conditions: caps on indemnification liability, notice periods, sole-remedy language, and carve-outs for the indemnifying party's own negligence. PandaDoc's AI generated a clause for a software consulting contract that said "Client shall indemnify Vendor for all third-party claims." No cap. No notice requirement. No carve-out for Vendor-caused harm. If a Client data breach triggered third-party litigation, the Vendor could claim unlimited indemnification recovery—even if the Vendor's own poor API design contributed. A lawyer would rewrite this to include a ₹50L cap, 30-day notice, and a negligence exclusion. The gap cost ten business days of redline and negotiation. Watch for: Indemnification clauses without caps, notice periods, or carve-outs. Fix: Provide a reference indemnification template to the AI, or disable auto-generation and write these clauses manually. Gap 4: Liability Caps That Do Not Match Your Risk Appetite LLMs often generate liability caps tied to contract value (e.g., "liability capped at 12 months of fees" ). This is a template default, not a business choice. For a ₹50L retainer, a 12-month cap means ₹50L exposure—which may be far higher or lower than your actual risk tolerance or insurance coverage. We tested all three platforms: DocuSign, PandaDoc, and Orin's contract drafting engine. DocuSign and PandaDoc generated liability caps tied to total contract value without asking. Orin prompted for explicit cap inputs before generation. The difference was material. On a ₹30L SaaS services contract, DocuSign auto-capped at ₹30L; the client's insurer would only co