Model contracts: from template to benchmark
Even lawyers do not always start from scratch when drafting contracts. They regularly use their own or third-party model contracts, or use artificial intelligence (AI) to draft contracts (see below). With AI, model contracts have become practically obsolete as a starting point, at least for non-lawyers: a chatbot drafts a contract on your facts in seconds, whereas a template describes someone else’s case and must be adapted, and a layperson rarely sees which clause does not fit. As a lawyer, I still use model contracts, but only as a benchmark: my own from earlier mandates, and those of reputable providers, to check whether my draft has forgotten anything.
Reputable providers I know: the ICT model contracts of Swico and swissICT, the associations of the Swiss ICT industry, revised in 2026 by the legal commissions of both associations, balanced between customer and supplier, available in English, German and French, fee-based. The Swiss Confederation publishes its general terms and conditions for IT services (also in English) and, for construction and planning, the KBOB model contracts (in German); its IT model contracts are no longer public. The German counterpart are the EVB-IT of the German federal administration, completely revised in March 2026, free of charge, with an interview-guided tool that builds the contract from your answers. It should be noted that the model contracts of public administrations are certainly fair, but also definitely «tuned» in favour of the administration. Finally, WEKA (in German) offers probably the largest collection of model contracts in the German-speaking countries, by subscription; the templates are written by many different authors, so treat each of them with the same care as an AI draft.
Drafting contracts with AI
In my opinion, it no longer makes much sense to use model contracts as a basis for drafting contracts, especially for legal laypeople. Today, contracts can be drafted with Claude, ChatGPT & Co. that are tailored to the specific facts. The better the chatbot is instructed, the better the result: describe the parties, the performance, the money, the term and, above all, what you want to achieve and what you fear. Not every legal document is suited to this. From my practice: a client once sent me a brief for court proceedings that he had generated with AI; that was of no use, because a brief is legal argument in a procedure with deadlines, burdens of proof and an opponent, and I have to build it from the ground up with the AI myself. A contract is different: it describes what the client wants, and nobody knows that better than the client. A draft the client has developed with AI is therefore a good starting point for me: it forces the client to think about content and risks, and I can see at once what he wants.
A smart procedure for non-lawyers
A smart procedure for non-lawyers
1. Let the chatbot interview you. Prompt: «I want to conclude a [type of contract]. Ask me, one question at a time, everything you need to know to draft it. Do not draft before you have all the facts.» Answer the questions; you will notice which points you have not yet decided.
2. Worst cases. Ask: «What can go wrong in this contract, for me and for the other party?» Decide which risks must be regulated (see the exercise in Chapter 03.06).
3. Draft. Let the chatbot draft the contract from the facts it has collected: short, in plain language, from your side of the table (see the exercise below).
4. Compare. Ask a second chatbot for a checklist for this type of contract and, if there is one, compare your draft with a model contract of a reputable provider. Close the gaps.
5. Hand over. Send the draft to your lawyer with half a page: what I want, where I am unsure. You will pay less and get a better contract, because the lawyer reviews instead of guessing.
Exercise
Exercise: drafting a software licence agreement
1. Sides. In class, we first assign the sides: half of the groups act for the licensor (X AG), the other half for the licensee. Tell your chatbot (Claude, ChatGPT & Co.) at the start whom you represent, for example: «You are advising the licensor. Everything you draft should be balanced and fair, but tuned in favour of the licensor.» Each side fills in the points to be regulated (3b) in its own interest: the licensor will typically want a long minimum term, a low liability cap and a right to audit the number of users; the licensee a short term, full liability for data loss and no audit. Do not copy the other side’s wishes into your draft.
2. Caution. Never enter the real names of the parties; anonymise them and, if necessary, alter the facts (see Chapter 01.02: keep company data confidential). Adapt the chatbot’s draft to the specific facts of the case afterwards.
3a. Facts (the same for both sides). X AG, CH-Zug (licensor), licenses a software for customer management to the customer (licensee), a company in your country. Functions: recording of customer data; allocation of documents, in particular correspondence from Microsoft 365, to the recorded customers; management of contracts with customers and suppliers; invoicing. The software is provided as Software as a Service (SaaS) on the licensor’s cloud; the licensee needs only a browser and internet access. The licence is a simple (non-exclusive) licence for the licensee’s own business. The licence fee is charged per user and month; the parties have agreed on CHF 40 per user and month for an expected 50 users. Start: 1 January of next year.
3b. Points to be regulated (deliberately left open). Term and termination (minimum term? notice period?); counting of users and control (who counts, may the licensor audit?); payment terms and default (30 days? interest?); updates, maintenance and availability (guaranteed uptime?); data security and responsibility for data (encryption, backups, who bears the loss of data?); liability (unlimited, or capped, and if so at what?); confidentiality; whether the licensor’s general terms and conditions apply; applicable law and place of jurisdiction (Switzerland or your country?).
4. Overview. To get an overview of the contract type and its possible content, ask for a checklist and a sample contract: «Please make me a checklist for a software licence agreement as SaaS» and «Please create a sample contract for a software licence agreement as SaaS. Fill in the contract with your facts».
5. Draft. Now give the chatbot the facts (3a), the points to be regulated (3b) and your side’s positions, and ask it to create a draft based on them: remind the chatbot whom it is advising and that the draft should be balanced and fair, but tuned in favour of your side. Keep the draft short: a maximum of three A4 pages, so that you can still oversee it. Prompt the facts together with the instruction; otherwise the chatbot may produce another sample contract independent of the facts: «Please create a draft for a comprehensive licence agreement based on the following facts, which I will give you in a separate post. Copy the facts very precisely. Supplement the facts with any additional, necessary and/or recommendable provisions. Limit the draft to a maximum of three A4 pages (about 1,200 words) and keep the clauses short. Please do not use paragraph signs, but numbers. The parties’ signatures must include the place and date of signature. [Facts of the case].»
6. Check. Read the draft, check it against the facts and your side’s interests, and ask the chatbot to adapt it where necessary. To improve the quality, paste the draft into a second chatbot, explain that you have drafted a contract for your side, and ask for suggestions.
7. Compare and discuss. In class, the groups compare their drafts with the other side, negotiate the three contested clauses (limitation of liability, term, fees) and then reflect on their own draft: what would you change? The lecturer moves from group to group, joins the discussion and reviews the results.