How to Start Harnessing the Power of Generative AI in Your Digital Printing Business

Talking about artificial intelligence and wanting to get down to the nitty-gritty, how can we start using this formidable tool today to do things that can help us in our digital printing or customisation business?

Challenges of AI in the Printing Industry

For those in the printing industry, AI presents two challenges. The first is understanding how it works. I've touched on this in my previous post and promise to provide something more detailed yet digestible soon.

The other challenge we will discuss in this article is understanding where and how to use it in our daily operations.

AI is Already Here

Let's clarify: we already use software tools incorporating AI elements. For example, suppose we use software that monitors our devices, indicates necessary maintenance, forecasts consumables, or calculates our return on investment (like Roland DG Connect or similar tools). In that case, we are talking about predictive AI.

Unlike generative AI, predictive AI does not create anything new but uses data from devices or other systems to calculate potential future actions—precisely the task for which it was made.

Generative AI in Graphics Software

Shifting towards the graphic side, generative AI is slowly integrating into graphic software, from Adobe suites or Corel plug-ins to more commercial software like Canva or Microsoft's new Designer.

We will likely see future integrations of image-generative models like MidJourney or DALL-E into other platforms, or these will evolve, creating alternative platforms. These generative image AI platforms, called Diffusion Models, work from user prompts to create images based on the provided input (refer to my previous article for more on prompts).

The quality of these images is incredible, making them perfect for presentations, mock-ups, or the web. However, the resolution still needs to be improved for printing purposes due to the resources and costs of generating such images on the servers, even if we can try to adapt them for large-format printing purposes.

Large Language Models (LLMs)

The other side of generative AI I want to discuss in this article relates to LLMs (Large Language Models), generative AI models that work with natural language, like ChatGPT, Gemini, Mistral or Claude.

This type of AI can be immediately applied to our daily business activities. It is truly a matter of getting hands-on to understand how it works and what benefits we can derive from it.

Getting Hands-On with AI

I will refer to ChatGPT, but the same observations apply to other LLMs we might use.

First and foremost, I recommend purchasing the paid versions for their power and response capacity. You can have fun with the free versions, but the number of parameters (adjustments and tweaks for optimisation) is much higher and more performant in the paid versions. The monthly cost is accessible (around $20 per month on average) and will prevent us from forming judgments based on less effective versions.

Configuration

When creating your account, you can immediately configure ChatGPT by telling it who you are, what you do, and your goals. This way, you won't need to repeat your requirements with each prompt, and ChatGPT will know precisely who you are, how to address you, and what you want to achieve. In the settings menu, you can define:

• What you want ChatGPT to know about you to tailor its responses (role, job, company mission, goals, challenges);
• How you want ChatGPT to respond (professionally, friendly, role it should assume, how to address you, whether to provide opinions or be creative in responses).

Here's an example:

In the first box, you could write something like, "I am John, owner of a digital printing company. I produce advertising graphics, promotional items, customised clothing, vehicle graphics, interior graphics, and signage. I work in (put your region here). My goal is to stay at the forefront of offering my products to customers, explore new markets related to digital printing, and cultivate relationships with colleagues, clients, and industry experts. I have eight employees and am responsible for final decisions on purchases and pricing."

In the second box, you might tell ChatGPT, "Write your responses informally but professionally. Always be clear and precise. Use lists and numbered bullet points. Act like an expert in small, medium, and large-format digital printing, with many years of experience and knowledge of trends and graphic applications. Show me data sources when necessary. You can address me as John. Offer different perspectives; don't stop at the standard."

These two small and non-exhaustive examples will help configure our model. Unless you want otherwise, remember to turn off the sharing of your data, which OpenAI would otherwise use to train its models (settings, data control).

Using Generative AI in Our Printing Business

How can we apply generative AI in our daily work? A myth to dispel is that AI simplifies everything and applies everywhere. These two things are true if we identify the process we use it (and we'll see this shortly). If writing a prompt takes more time than getting a response, it might be better to reconsider using standard approaches.

In the initial phase, I suggest using ChatGPT (or other models) as much as possible, whether responding to an email, researching, asking for ideas, or analysing an Excel sheet. This will make everything more intuitive, helping us master the tool better.

Prompts should always follow a structure to work well, which usually is:

1. Provide context, telling the AI who it should impersonate;
2. Define the task it should perform;
3. Give instructions on how to do it and how to replicate it (tone, approach, sentiment);
4. Define the audience to whom the result will "speak";
5. Output format (language, length);
6. If necessary, define limits (do not say X or Y);
7. Give examples to guide the AI in better responding (this technique is called few-shot prompting).

You can also ask the model to review the prompt and suggest missing information. You can do this even 3 or 4 times, telling the machine to read what it produces and improve itself progressively by generating increasingly accurate prompts. We will dedicate a future article to prompts. In the meantime, you can also explore prompt collections for various occasions.

Practical Applications for Printers

Let's now take some examples of using ChatGPT to help us in our daily business. Let's use interior decoration as an example, assuming the role of someone wanting to approach this business. You can replace interior decoration with any other market (merchandising, t-shirts, vehicle graphics, etc.). Let's develop prompts to help us set up our strategy from here.

Example of Interior Decoration Market Approach

Market Analysis

So, our goal is to enter the retail and business interior decoration market. We want to know more about this market and compare what we already know. We could start by asking ChatGPT:

"You are an expert in digital printing for interior decoration with over 20 years of experience. Your task is to evaluate the market size, growth trend, and most requested applications. Your assessment should help me set up my business and plan my investment. At this link, www.example.com, you will find a business already offering this service. Always cite the sources of the reports you use for your statistics. Before proceeding, read the prompts and check if you have all the necessary information. Suggest any improvements before executing the prompt."

Depending on the response, we might ask ChatGPT to explore some aspects further or explain others better. I recommend saving this project as a PDF so you can present it to ChatGPT for subsequent requests.

Creating Our Buyer Personas

Who are we targeting? Having done the above, we also ask for the following additional information: "Describe the buyer personas for this market and divide them into influencers (architects and designers) and users (shops, public spaces, private homes). Create a table where each persona has a brief description, needs, benefits, and critical points. At these links, you will find some businesses similar to mine. Use the sites to enrich your personas. Before proceeding, read the prompts and check if you have all the necessary information. Suggest any improvements before executing the prompt."

If you continue in the same window, the above prompt might suffice. You can attach the PDF with the developed plan to reinforce the prompt further. If you do it from a new window, uploading the PDF and the prompt will be necessary to guide the machine better.

These personas can be saved in a separate document or (recommended) included in the above PDF so you can work with a single document.

Brainstorming with Colleagues on the New Market

Now, wanting to discuss this entirely and get more ideas from our collaborators, we could ask ChatGPT, "Create a brainstorming structure based on the following question – How to enter the interior decoration market with digital printing – use creative approaches and suggest ideas to position the company and our printed products. The structure should include market analysis, positioning strategy, ideas to propose our products, and possible marketing and sales strategies. We will use the results for an internal discussion. Use the attached PDF as a work base. Before proceeding, read the prompts and check if you have all the necessary information. Suggest any improvements before executing the prompt."

Creating a Web Page for Interior Decoration

From here, we could now ask, "Based on what has been discussed so far, create a web page structure to advertise our service for interior decoration based on personas X, Y, and Z, highlighting the advantages of using digital printing and the quality of our service. Include slogans and texts with a professional but friendly tone and add a call-to-action for conversions. Identify the keywords used by these personas and include them in the page text. Use the attached PDF as a work base. Before proceeding, read the prompts and check if you have all the necessary information. Suggest any improvements before executing the prompt."

If you have sketched something on a paper sheet during the brainstorming, you can upload it and ask ChatGPT to refer to it. ChatGPT will read the sketch and apply the structure to the final output.

Blog Articles for Our Promotion

If we have a blog and social media pages, we could ask our model:

"You are an expert in digital printing for interior decoration with over 20 years of experience in the field. Prepare three articles for our blog that discuss the benefits of using digital decoration for interiors. The first article focuses on applications in the retail world. The second is on how the creativity of architects and designers can be amplified with digital printing. The third is on wall decoration with customised wallpapers or wall graphics. Highlight the benefits and how the critical points of the various personas are addressed by the solutions we offer. Include the most used keywords for each article. Here are some examples of similar pages developed by competitors - www.abc.com, www.def.com, www.ghi.com - Be creative but maintain a friendly yet professional tone and always conclude by inviting contact with our company. Use the attached PDF as a work base. Before proceeding, read the prompts and check if you have all the necessary information. Suggest any improvements before executing the prompt."

Social Media Posts

Having done this, we will ask it to create five posts for our social channels:

"You are an expert in digital printing for interior decoration with over 20 years of experience in the field. Create five posts about digital printing for interior decoration for Facebook, Instagram, and LinkedIn. Use the typical tones for each social network. Use the description of personas in the attachment to better target the audience. Use the attached PDF as a work base. Before proceeding, read the prompts and check if you have all the necessary information. Suggest any improvements before executing the prompt."

We might attach the blog articles created to give more context to the model (always in PDF format).

We could also ask Chat GPY to prepare the pictures to accompany such posts in the format requested by the platform we will use.

In Conclusion

This is a small and non-exhaustive example of how a platform like ChatGPT can help us optimise work and have material that can be finalised quickly, respecting our needs and the audience we are targeting. This material can be used internally if we manage communication channels ourselves or given to an external agency as a brief.

Don't take everything for granted. Remember that what an LLM model creates should be reviewed and enriched or corrected by your field experience. Ask the model for sources when necessary and personally verify the results. Refrain from settling for the first response; iterate the prompt if necessary.

ChatGPT (or its equivalents) does the heavy lifting, but the professional touch that makes everything accurate and conveys expertise can only be added by you. Human (always) first!