Units
1. Contract Review and Summary 2. Risk Substance Identification and Playbook 3. Contract Drafting Support 4. Commercial Contract Types and Type-Sensitive Review 5. Legal Research and Jurisprudence: The Obligation of Verification 6. Case and Petition Support Drafts 7. Compliance Checklists and Gap Analysis 8. Policy and Procedure Writing 9. KVKK, Personal Data and Data Processing 10. Intellectual Property, Copyright and Ownership of AI Output 11. Confidential and Privileged Document Rules 12. Legal Limits of AI Output and Professional Liability
Unit 3 / 12

Contract Drafting Support

Gains:

  • Ability to produce a draft contract adapted to the scenario, based on an approved template
  • Ability to design reusable templates with placeholders and fallbacks
  • Ability to mark boundaries and assumptions while preparing the draft for legal review.

If reviewing a contract is “reading,” drafting is “writing”; And writing often takes more time and carries more risk. Artificial intelligence (AI) eliminates the fear of the blank page: produces a workable draft in minutes. But here lies one of the module's most dangerous traps. If you tell the AI ​​"write me a confidentiality agreement", you will get a text that looks fluent and professional, but does not meet your institution's standards, does not comply with the applicable law, and even contains fabricated (hallucinatory) clauses. There is a single principle at the heart of this unit: starting from an approved template. The approved template serves as an “anchor” that anchors the AI ​​to the agency’s tried-and-true, lawyer-approved text.

Let's clarify the terms. A template is a standard contract text approved by the institution and used repeatedly. The placeholder is the space in the draft that will be filled according to the scenario; e.g. [PARTY NAME], [FEED]. Fallback options are alternative texts to be presented in negotiations for an article. Redline is a traceable marking of changes made to a text. Assumption is an assumption that is temporarily accepted instead of unknown information in the draft and must be verified.

Generating a Draft from an Approved Template

The correct workflow is not "create from scratch" but "adapt approved template to scenario". You give the AI ​​both the template and the scenario information, just asking it to fill in the placeholders and add the necessary items according to the scenario. Step by step:

  1. Give the approved template. The institution's standard text is the skeleton of the draft.
  2. Provide structured scenario information. Parties, subject, price, duration, special conditions.
  3. Put a placeholder and assumption discipline. Any unknown information should be marked with [TO BE FILLED] and ASSUMPTION: tag.
  4. Mark the boundaries. Let the AI ​​point out any place where it strays from the pattern or requires a legal decision.
  5. Ask for redline and justification. Let him report what he changed according to the template and why he changed it.
  6. Prepare for legal review. The draft is submitted with the status "pending human approval".

Your role: a legal assistant drafting corporate agreements. Take the APPROVED TEMPLATE below as a basis and adapt it to the scenario. Rules:- MAINTAIN the structure and standard clauses of the template; invent a new legal article on your own.- Leave any information you do not know as [TO BE FILLED: ...]; do not guess.- Mark each assumption you remove from the scenario with the "ASSUMPTION:" tag.- Mark the places where you go outside the template or require a legal decision with the note "HUMAN REVIEW REQUIRED".<template>[approved template text]</stencil><scenario>Parties: [...] ; Subject: [...] ; Price: [...] ; Duration: [...] ;Special conditions: [...]</scenario>

Hint: The instruction "Don't make up new legal clauses on your own; leave out what you don't know [to be filled in]" curbs the AI's most dangerous behavior (adding clauses that seem reasonable but do not comply with applicable law). Leaving a space is safer than filling it incorrectly.

Designing a Reusable Template

AI doesn't just produce drafts; It also helps you design good templates. A good template is text with clear placeholders, fallback options, and clear usage notes.

Convert the following service agreement clause into a reusable template clause: - Show variables with [SQUARE BRACES] placeholders. - Provide 3 fallbacks for the liability ceiling: (A) our ideal position, (B) reasonable middle, (C) final acceptance limit. Label each separately. - Add a short "usage note": which fallback to choose in which situation. Prepare the text as a draft for a lawyer to approve; giving definitive legal advice.

There is also the aspect of language simplification. Complex, intertwined contractual clauses make both negotiation and implementation difficult. AI proposes simplification by trying to preserve legal meaning; but due to the risk of semantic change this is always done comparatively and human-verified.

Write the following article in a simpler and more understandable manner WITHOUT CHANGING its legal meaning. Give the output in two columns: the original on the left, the stripped-down version on the right. Also mark any simplifications that might affect the meaning as "CAUTION: semantic check". When in doubt, keep the original.

A common task when drafting is to merge two parties' texts: reconciling your template with the changes sent by the other party. AI is quick to make differences visible and mark which side favors each difference — but it's still up to the human to decide which difference to accept.

Below, on the left is our approved template item, on the right is the version recommended by the other party. Compare the two:| Difference | Our text | Opposite party text | In whose favor | Suggestion (acceptance/negotiation/rejection) | Mark each difference that changes the meaning with a "meaning effect" note. I will make the decision; you just show the differences and their effects.

Weak Prompt / Strong Prompt

Weak prompt:Write me a service contract.

Result: A text derived from the internet average, which does not meet your institution's standards, where you do not know which article is real and which is fake, and which is dangerous to use directly.

Powerful prompt: [approved template + structured script + [TO BE FILLED]/ASSUMPTION discipline + "make new clause" limit + "HUMAN REVIEW REQUIRED" flags + redline and justification]

Result: A draft that is faithful to the institution's standard, has gaps honestly marked, shows where human judgment is required, and is ready for legal review.

Draft Maturity Levels

Level

What does it contain

Who approves

Is it available

Raw draft (AI)

Template + filled placeholders

—

No, only inner work

Marked draft

Assumptions and review notes added

contract specialist

no

Reviewed draft

Lawyer crossed redline

lawyer

Yes to negotiation

final text

Ready for signature, approved

authorized signature

Yes

Three Mini Cases

Case 1 — Fabricated substance. When a startup asked for a “startup-friendly draft investment agreement,” AI added an “automatic conversion” mechanism that didn’t actually exist, like a legal template. The founder almost sent this to the investor. The team changed the workflow and went into production only from the approved SAFE/share agreement template; There were no fabricated articles in the next 12 drafts, and legal review time decreased from an average of 3 hours to 1 hour per draft.

Case 2 — Placeholder discipline. A legal team was preparing franchise agreements to be adapted individually for 40 dealers. They put placeholders [DEALER NAME], [REGION], [ROYAL %], [TERM] in the approved template and gave each dealer data in a tabular form to the AI. 40 drafts were produced in half a day; No contracts were submitted with incomplete information, as a single unfilled field in each remained visible as [TO BE FILLED]. This job took approximately 5 days manually.

Case 3 — Simplification gain. An institution's membership agreement for consumers was written in strong language, which was the source of the complaints. Two-column simplification was done with AI; The legal team approved the 7 articles marked as having changed meaning one by one, rejected 2 of them and returned to the original. With the new text, "I did not understand the contract" type calls to customer services decreased by 28% in three months. AI suggested simplification; Man confirmed that the legal meaning was preserved.

Common mistakes

  • Asking for a draft from scratch. Production without a template produces text that does not meet the institution's standards and has a high risk of fabricated material.
  • Prediction rather than placeholder. Instead of having the AI ​​fill in the blanks, ask it to leave [TO BE FILLED].
  • Leaving assumptions unchecked. Every unverified acceptance is a risk silently embedded in the contract.
  • Redline and not asking for justification. It's hard to review a draft without seeing what's changed and why.
  • Mistaking the AI ​​draft for the final text. No AI draft should go to negotiation/signature without legal review.
  • Overlooking the semantic shift in simplification. Simplification should always be comparative and human-approved.

In summary

The power of AI in drafting is to quickly fill the blank page; The danger is that he will make up items that seem reasonable but are wrong. The way to manage the two is through a single discipline: starting from the approved template. Work with placeholders, marked assumptions, and “human review required” notes; Make simplification comparative and do not consider any draft final without legal approval. AI produces a workable outline; The authorized professional decides whether the draft is legally correct and suitable for signature.

Application task

Select an approved contract template from your institution (or sample). (1) Define a scenario and have it produced via the template-based draft prompt; Verify that all [TO BE FILLED] and ASSUMPTION flags appear. (2) Turn an item into a reusable template with 3 fallbacks. (3) Simplify a complex entry into two columns and examine the meaning check marks. (4) Close the draft by adding a review note with the status "pending legal review."

checklist

  • [ ] Was the manuscript produced from an approved template (not from scratch)?
  • [ ] Unknown information left as [TO FILL]?
  • [ ] Are the assumptions marked with the ASSUMPTION: tag?
  • [ ] Are areas requiring legal decision marked with "HUMAN REVIEW REQUIRED"?
  • [ ] Have the redline and justification been received according to the template?
  • [ ] Have changes in meaning in simplifications been checked separately?
  • [ ] Was the draft submitted without legal approval before it was considered "final"?
Previous Next