AI & design.

The short version is that when you pay Onetonne, you’re paying for human creativity.

This business is incentivised to see you succeed - if what we do together delivers positive results then you stay as clients.

Our goals overlap as it’s in our best interest to see your business succeed. Most of our clients are long-term and we get to see them grow and succeed over time.

AI companies don’t care about any of that because they’re not actually concerned about your business being successful. They care mostly that you use AI as much as possible. That’s all. That’s the business they are in. (That business model is looking quite wobbly also).

AI Policy

This is tricky stuff to write (Hey! That’s how you know a human wrote it).

While there are uses for AI & Large Language Models (LLMs) and Machine Learning (ML), there are significant ethical, environmental and financial reasons to think carefully about how your business uses AI.

Every bit of culture accessible on the web (and rare books) has been stripmined by the AI companies for the purpose of training AI models. Most of this has not been paid for. Even if you “opt-out” your content is not safe.

Generative AI is the word we use to describe when AI creates new content (words, images, video & music).

The thing you need to know as a business owner is that AI generated artwork and copy doesn’t sell. People don’t like it.

Accuracy of AI generated information is a huge issue. We use the word “hallucinate” to describe when an AI makes up information which is a hell of a marketing glow up for the words “lie to you”.

When you use Generative AI tools to present your business, the authenticity of your whole offer comes into question.

Our AI Rules:

  1. I don’t talk to robots, I talk to you, the client.

  2. Robots are good at robot work, but creative work isn’t robot work.

  3. I don’t use generative AI to make what you’re paying for. 

  4. I don’t accept AI generated project briefs.

  5. I don’t accept AI generated project feedback.

  6. I don’t accept AI generated mockups, visualisations or “inspiration.” 

  7. Don’t upload anything I’ve made for you into any AI tool.

  8. AI generates bad creative work that risks your business.

Humans do better creative work than robots because robots cannot truly create.

AI “thinks” by averaging out all the inputs and guessing what you want to see. Word by word, pixel by pixel. But it doesn’t really know. It guesses.

What we use AI for.

AI is in almost every digital tool. Even a simple Google Search now returns AI results above the actual search results.

There are uses for AI where it is the best solution. Robots are best suited for robot work, but if, and when that happens, we never outsource our critical thinking.

We use AI:

  1. As little as possible. AI is in most software tools and the reality is that it is now hard to avoid. If AI is going to give a truly better result we’ll consider it. But any AI use weighed up very carefully at Onetonne.

  2. Sorting data. Robots are good at sorting data (but are untrustworthy when it comes to summarising, providing insights or taking action on those results).

  3. Searching for content. Robots are very good at indexing information - especially when the right word is on the tip of your tongue!

  4. Image upscaling - when supplied images are not high resolution for the intended use and sourcing a replacement image is not possible. (We will generally try and find a better solution instead though).

  5. Image restoration or background extension - an example task might be adding more sky to an image to better centre the overall composition. Adobe asserts their Photoshop AI tools have been trained on ethically sourced / licensed stock photos. We used to do this manually and AI is often quite good at this task.

  6. Coding assistance - AI is good at summarising content like manuals and helping contextualise the example in the manual to a specific project. (AI is not a reliable substitute for an experienced coder and if used, output must be carefully checked.)

  7. Some repetitive tasks - some repeatable tasks are a good match for AI. But tasks that repeat perfectly are often better (and more reliably) handled with a code based solution that doesn’t involve AI at all.

  8. Minimal, targeted changes to wording. AI writes terribly generic marketing copy, but for grammar clean up and review it can be useful. (We can help with and also recommend copywriters who will make your words sizzle and sell. )

AI tools always cheerfully say “Yes!” to each request. But the truth of what they can do, and do properly is much more of a grey area than the AI marketing implies.

There are so many things the AI industry would rather you not know. The moment AI loses context (it runs out of memory), or it just happens to guess the wrong word, accuracy goes out the window.

Even “robot work” is not always safe to hand off to AI. Sometimes it is far better to use a traditional code tool (such as GREP) to clean up data than trust a LLM with “robot work”.

As a simple example, you might want to clean up your online shop’s product database, make some changes to each product and add in some new detail to your product descriptions.

Seems like a robot job, right? Nope. It’s a big database and the AI keeps inserting forbidden characters and writing information into the wrong fields which completely breaks the product database.

When upload this now corrupted data source into your website’s backend it causes enormous headaches. Because the content won’t marry up with the existing database the system duplicates each mismatched product and now you’re cleaning up 500 duplicate products ...

Let’s pretend you just said these things - here’s the response you’ll get:

  • No. Please do not send project briefs, inspiration, visualisations or any other output generated by an AI tool.

    AI generated briefs often look good, but a full of holes and issues. As a professional weeding through the incorrect assumptions and fluff makes it HARDER, not easier to develop a great response for your project.

    AI tools are extremely good at saying yes to everything you ask for, but they are very bad at local context, they are very bad at responding with nuance, they are very bad at doing work that is outside the norm (they are, after all, a tool for averaging out inputs - that means you get average work in response).

    “But why not?” Because AI briefs miss the point, hide the real issues, and fill lists of requirements that are often irrelevant. Most importantly they AVERAGE out the brief to give an average product.

    If you want average go for your life, but it’s not a service we provide.

  • You don’t have to. Let’s sit down over a cuppa, have a talk and figure out what you actually need. We can get more done with a face to face chat because that’s where creatives can ask questions, and uncover the whole picture.

    Good designers don’t worry about what a project should look like until after we’ve figured out all the important stuff like “how should it work?” and “are we asking the right question?” and “Is this actually going to work before we spend the money?”.

    What it looks like is actually much further down the list than you may think.

  • Absolutely not.

    We ask our clients to never use AI to respond to the work we’ve made for you.

    Why not? When you do that, not only does the feedback suck, you’ve uploaded our work without our consent to further train AI tools.

    We actually need feedback from your unique perspective. We don’t need to hear the amalgamated, averaged out response of a data centre on the other side of the planet. Your business is unique and your context is unique and so are your customers.

    AI tools write really shitty generic feedback. Usually pushing you business towards the middle of the pack.

  • If you choose to upload our work to any AI tool, you are doing so without our consent and we reserve the right to cancel any project where we can see that has been the case (within reasonable doubt).

    A copyright breach fee of up to 50% of the project value may also be charged.

    Work done to date will be invoiced up to the nearest milestone.

    Handoff files can be created for other creatives/agencies to use, but these will be charged at our standard hourly rate.

  • Yes, AI is often very fast.

    The lawn mower is probably also faster at chopping up lettuce, but it doesn’t make a good salad.

    If you want to use AI tools in your business workflow, that’s your choice, but the growing consensus is that clients don’t like AI artwork, they don’t respond to AI copy and return on investment with AI is extremely low.

    AI replaces quality for speed and that’s not a choice we’d recommend.

  • AI tools are made to be agreeable virtually no matter what you say.

    AI is a math machine - it guesses what the next word will be based on the content it’s been fed. It’s complicated but ultimately that’s how it works.

    AI is not a reliable source - it averages out a huge amount of (stolen) data and then gives it’s best guess what the answer should be. Even better, AI is now polluting the pool of human knowledge as AI generated content is now filling the internet. This means for every mistake AI makes that gets published on the open internet, it further compounds the problem.

    Because AI can’t truly think - it will ingest that wrong data and deliver worse and worse results over time. This is a thing called Model Collapse if you’d like to learn more in a bit over minute.

  • You can say that - but what AI helps you do is paint by colours. It helps you say a lot of words, or make something instantly. It often looks good when you first look at it, but the closer you look the more dodgy AI output is.

    The reality is that heavy users of AI might start with the best of intentions (“it’s just a tool” or “it helps me work faster”) but when we outsource our thinking cognitive decline happens.

    Here’s a quick test: Ask AI about a topic you REALLY know about. Your area of special expertise - and you’ll see it get it partly right, or almost right. Almost right isn’t good enough. Almost successful brain surgery is non-successful. An almost successful business is not a business any more.

  • Chances are it doesn’t.

    It might look ok to you, and if that’s all you need, then that’s ok. (I mean it’s not ideal considering the environmental and social cost of megascale AI).

    But if you’re trying advertise your business then you might be risking more than you gain with using AI and not even realise it.

  • An overwhelmingly large amount of the training data (as it pertains to creative work) has been stolen without permission by the large tech companies. I would be absurd to say the companies that are trying to sell the end of knowledge work are on the side of disabled and neurodivergent people.

    AI and other machine learning tools can be very useful for disabled, neurodivergent and marginalised people. AI can be useful when used responsibly, but Generative AI has been a net negative for the creative class and the quality of the commons (the spaces we share online and in the real world).

  • By the theft machine?

    Or by the people who can’t do what I can do who need to pay a robot to do it? The robot that does a worse job?

  • There are absolutely good uses for this technology. But there is a distinction between generative AI and tools like this.

    But it’s absurd to ignore the environmental and social implications of AI.

    AI that saves human suffering? Here for it.

    AI that protects people from harm? Here for it.

    AI that takes monotonous repetitive work and makes that easier for the human doing the job? Here for it.

    AI that helps people learn, grow and build capacity? Here for it.

  • Of the theft machine? No I hate AI because it’s bad (culturally, socially & environmentally) , and it’s taken the world’s knowledge base and and actively made it worse.

    If it can truly do what I can do and better than me, then it’s time for me to find a different job. But that’s not the case.

  • Surgeons are not known for their bedside manner.

    I think I’m beginning to understand why. The commercial creative arts has historically been a consultative one and one that incentivised creatives to “get along to get along”.

    (After all, nobody wants to deal with a pain in the arse.)

    But would you rather a surgeon that went along with everything you said (no matter the risk to your body) or one that diagnosed you accurately and gave you what you needed?

  • When you decide to use generative AI because you’re under time pressure you transfer that time debt to the first person who knows what they are doing. When they read your AI generated brief, feedback or website content they know.

    If you are under time pressure and that time pressure causes extra work, then that extra work needs to be billed for.

  • If we ignore the environmental, social and ethical implications and say “this' is just a tool” it would be quite different to any of the powertools in my shed before we get started.

    Let’s look at like a powertool. A large circular saw is the right tool for quite a few shed jobs.

    But if you use that tool without knowing how it behaves, it can very quickly adjust your ability to count to ten using your fingers.

    When businesses use AI tools without fundamentally understanding the limitations and risks you’re not risking your fingers (hopefully) but you may well be risking your business.

  • My job is to help my clients communicate ideas and sell things. AI is NOT GOOD at so many things.

    If you genuinely can’t tell the difference and won’t trust a professional (in any area of expertise) then please allow me to unprofessionally show you the door.

    It is entirely reasonable for anyone concerned with quality and accuracy to have a strong stance on generative AI.

  • Let’s see shall we?

    Local models, running AI doing things that are trained on your own data is one thing (local models tend to be weaker in many areas) but mega scale AI is yet to cost the consumer what it costs to run and as the cost of tokens increases, what looked like a no brainer (ironic) might look quite a bit different in the next 12 months.

    What we currently have is is very artificial intelligence. We’re close to the limit of what LLM based AI can do. A significant breakthrough in technology is required to break that ceiling.

    If you’re happy with what the robots generate and you’re not interested in quality or accuracy of work, then nobody can stop you.

    Go all in if you wish, but please look carefully at what has been happening to AI led businesses around the world.