Does your freelance business need an AI usage policy?

Year-end marketing sprint starts Monday!

Marketing your freelance business can be less intimidating if you a) do a bunch of marketing all at once, and b) have a supportive group for accountability and idea-sharing. Join the Training for Translators year-end marketing sprint, and plant some good marketing seeds before the year-end holiday season sets in. We kick off on Monday, registration is $75 and includes a set of marketing resources, daily marketing nudge e-mails, and four live sessions (recordings provided). 

Next up: Turn AI into your freelance marketing assistant, with guest instructor Nora Díaz on October 16. This two-hour master class will teach you how to use AI for market research, and to create marketing materials. Registration is $75 and includes the recording. 

 

This week’s topic: Does your freelance business need an AI usage policy? 

Thanks to my friend and colleague Veronika Demichelis for suggesting this topic! It’s timely and relevant, whether or not you use AI in your business! 

In my opinion, your freelance clients need to know if, how, and when you use (or don’t use) artificial intelligence tools. I had been pondering this idea in a vague way, but I hadn’t really formulated any coherent thoughts until Veronika suggested this as a discussion topic during a T4T online course alumni question and answer session. My own ideas on this topic are evolving; definitely reply to this e-mail and let me know if you have thoughts! Veronika perceptively pointed out that an AI policy is important because some clients are really excited about the possibilities that AI offers, while others are concerned about AI-related confidentiality, or just don’t want to use it for a variety of reasons. 

Most importantly: yes, I think you need an AI usage policy, in that you need to disclose to your clients if, how, and when you use AI, and what AI tools you use.  An important caveat: in my own business, it’s honestly very rare that a direct clients asks me about AI. The majority of my translation clients never cared much about technology, and still don’t. Clients aren’t obligated to care about your technology setup, and I’m fine working with clients who just want to throw money at a problem (translation) and deal as little as possible with the process details. This is how I feel about my accountant: I pay him so that I can think about accounting as little as possible. I assume he uses an array of technology tools, but I really don’t know or care, as long as my taxes are done accurately and on time. If your clients also feel like this, I think that’s also fine. Here, I’m talking about clients who may be interested in knowing about your AI use. 

An AI usage policy could be posted on your website, or you could send it to all new clients, or something else. I haven’t created or distributed a formal policy yet, but I’m working on mine. Here are a few ideas for crafting your policy: 

1. If you use no AI tools, just say so. “In my work for clients, I do not use artificial intelligence tools for research, translation, writing, or any other tasks.” Done! Your clients know where you stand on this. 

2. If you use standalone or integrated AI tools, state which one(s) and which version(s). For example, I use ChatGPT Pro (the lowest-level paid version) and I think clients should know this so that they can make an informed decision about whether they want any of their information fed into it. If you use an AI plugin for your translation memory tool (I do not), I think you should disclose this as well, and provide a link to the plugin’s own web page so that clients can read about it if they want to. 

3. If you use AI tools, state whether clients can completely opt out of having these AI tools used on their projects. Personally, I’m fine with this. I do think that ChatGPT’s research capabilities have improved, and certainly sped up, my ability to research complex questions, and I use DeepL regularly for interpreting prep, but I translated for 20 years without AI, and I went to graduate school for interpreting before machine translation was very useful. It’s completely fine with me if clients don’t want AI tools used at all on their projects, and they don’t need to give me a reason. 

4. If you use AI tools, I think (and this is perhaps not a majority opinion) that clients deserve to know whether you use them for first drafts, or not. I’m open to being disagreed with on this one, and I think the concept itself is one that sparks a lot of discussion among translators these days: I personally really dislike the way AI writes; even when I ask it for a specific tone or style, I find the writing robotic and just kind of cringe-y. Thus, I do not use AI tools for translation first drafts. But if you do, I think that your clients deserve to know that, if it matters to them. I talk to lots of translators these days who say that they run everything through DeepL because they’d rather have a starting point than begin from zero. I think you should tell clients this, especially if you’re putting their documents into a tool (like ChatGPT) where even the paid version isn’t truly confidential. 

5. To help clients understand, give a few examples of how you use AI. Mine might include:

• I may use the paid version of DeepL to pre-translate speeches that I’m going to interpret, particularly if it seems that the speaker will read the speech from a script. When I do this, I always use the translated speech as a reference document only, while listening closely for differences between the speaker’s script and their spoken words. 
• I may use ChatGPT for research purposes: to explain words or legal concepts, to search for reference documents online, to identify the most common translation or term of art for a specific word or phrase. 
• I may use ChatGPT to suggest a translation of a sentence or a paragraph. When I do this, if the text in question contains any identifying information about a client (even if not confidential: an example of this would be the client’s name when I do not have a non-disclosure agreement with the client), I manually re-type the text instead of copying and pasting it, so that the identifying information is never entered into ChatGPT. 
• I may use ChatGPT to edit or revise parts of a translation, or to comply with specific length or style requirements, using prompts like “Make this 10% shorter, keeping the main points and key words, but eliminating redundancies and wordiness.” “Rephrase this to eliminate passive voice.” 

6. Just to be safe, assure clients that you always comply with the usage policies of the AI tools you use, and that you verify and take responsibility for any output of the AI tools you use. Don’t end up like the Philadelphia Inquirer, publishing a summer reading list of books that don’t exist, because no one could be bothered to check the AI output. 

I also find value in talking to people about AI on a human level: I’m constantly surprised by ChatGPT’s ability to answer complex research questions in a few seconds, and by how often ChatGPT just makes stuff up! 

I hope these tips are helpful if you’re contemplating your own AI usage policy!

To get our weekly posts directly in your inbox, sign up for the Training for Translators mailing list!