AI is everywhere at the moment. Some companies talk about it as if it will fix everything, some quietly use it without saying much at all, and others genuinely don’t touch it.
We think customers deserve a clearer answer than hype or silence. So this article explains what we use AI for at Think Promote and, arguably more importantly, what we don’t use it for.
Our general policy
Firstly, we’re not anti-AI. We do think there is a lot of noise at the moment, sometimes too much noise, but we also think large language models (LLMs), especially recent ones, are very useful.
Our company policy on AI is based on accountability and correctness. We may use AI to check text, help rewrite badly worded copy, fix small issues, discuss ideas, or improve clarity. But the important point is that Think Promote remains accountable for what we publish, what we produce, and anything else we do.
Quite frankly, we don’t care whether AI was used to help write something. We care whether it is:
correct
accessible and clear
accountable to named human authors or reviewers
in line with our standards and the way we work
our own work, or properly credited
The truth is, these are the things we cared about even before AI came along, so not much has changed in this regard.
Explore our range
Helpful next steps if you are planning a branded print project.
Ready to discuss your branded print?
Tell us what you are looking for and we’ll help you choose the right option for calendars, diaries or notebooks.
There are strong precedents for using AI in this way in other fields, such as:
spellcheckers and grammar tools, where software suggests changes but the writer remains responsible for the final text
accountancy software, where calculations and reports can be automated, but the accountant is still responsible for checking the result
legal research tools, where technology can help find relevant material, but a solicitor still has to apply judgement
design software, where templates, filters and automated corrections can speed things up, but the designer still owns the final work
manufacturing and quality control systems, where software can flag issues, but people remain responsible for production standards
Our view is that the standards have not really changed. The tools have become more powerful, so the risks of getting things wrong are higher.
Where we do not use AI
Customer service
Everyone at the senior level at Think Promote has a strong dislike of AI customer service, especially AI live chats. It is almost cultural here.
Maybe there will be a good use for AI in customer service one day, but we do not think it is time yet.
Either way, all customer service at Think Promote, whether by phone, email or web form, is handled by humans. Thank god.
Pre-print checks
Reproducing a factory floor printer’s gut instinct for whether something is right or not is surprisingly challenging. We have not figured out how to do that yet, so we just... don’t.
Everything print-related is checked by a qualified person before going to press every time. That said, there are some promising use cases for AI within checks on copy embedded in artwork, such as spelling, grammar, etc., and we’re experimenting with how that could help our customers.
Part of that is simply experience. These products have been refined over many years, and the basics matter more than people sometimes think: clear date grids, sensible advert space, strong images, and a layout that makes sense when it is actually hanging on a wall.
AI can produce interesting visual ideas, but that is not the same as understanding what makes a printed calendar useful and readable. At the moment, it has not added anything useful to that process.
Client design work
We do not use AI to design customer adverts or other bespoke customer artwork.
The main reason is IP and data handling. Bespoke design work often involves customer logos, brand assets, artwork files, photos, copy and other material we do not own. Using AI services for that work would usually mean uploading some of that material into another system.
We do not think that is the right trade off.
There is also the practical point that AI has not added enough to this kind of design work for us to justify using it. Customer artwork should stay within the normal design and production process, handled by people who understand the job and the responsibility that comes with it.
Internal systems
Take pricing calculators as an example. These rely on structured product catalogues, price sheets, quantity breaks, production rules and other details that need to be handled exactly. If a quote is built, we need to know exactly how that quote was built. We need to be able to check it, explain it, and repeat it.
That is why many of our internal systems are deterministic. In plain English, deterministic means they follow fixed rules and should produce the same result every time when given the same information.
This matters across systems like Track My Order, the Trade Portal, quote builders, pricing calculators and production scheduling. These systems are not there to sound clever. They are there to be correct.
LLMs are mainly good at understanding (and generating) language, especially when working with unstructured or loosely structured information. That can be useful in the right place, but it is not what most of these systems need. For accounts, production tracking, invoicing, pricing and scheduling, we need reliability and correctness above all else.
So, we do not use AI to run these systems.
Where we do use AI
Fixing sloppy writing
With the tools available in 2026, we think it makes sense to use AI to make written copy clearer and more accessible. So we do use AI for that purpose.
We do not do this without caution. For articles on this site, the process usually looks like this:
The author writes the article in their own voice
The author may use an LLM to polish the writing, without introducing new ideas or changing the author’s voice
The article is sent to a reviewer, who checks it for correctness, clarity, and whether it is in line with the company’s position and brand
We do this internally for all articles. Where a reviewer has had a significant role in the content, they may also be named publicly on the site.
The key point here is that AI can help clean up the writing, but the responsibility for the final copy stays with us.
Coding assistance
A lot of development work involves taking code that already works and making it cleaner, easier to maintain, or easier to extend. In the software world, this is often called refactoring.
The developers we work with may use AI to help with that kind of work. They may also use it to help draft code, spot issues, suggest fixes, or explain why something is behaving oddly.
This is not especially controversial among developers. It is fairly normal now with built-in tools such as GitHub Copilot.
Code also has safeguards that written content does not always have. For example, software can be checked with unit tests, end-to-end tests, continuous integration (CI) checks and other automated tests. These help confirm that changes have not broken the parts of the system that already worked.
That does not mean AI gets to write code unsupervised. It means developers may use AI to help with drafts, fixes, refactoring and debugging, with the final responsibility staying with the people building the system.
Brushing up marketing images
We do use AI for some image editing, but not for anything that goes to print.
A common example is taking a real photo of one of our products and using AI to improve the setting around it. That might mean changing the background colour, placing the product on a cleaner table, removing clutter, or making the image easier to use on a product page. Factories just aren’t that pretty for product shots!
The line for us is the difference between presentation and misrepresentation. We are comfortable using AI to improve how a real product photo is presented. We are not comfortable using it to change what the product is.
So the product itself still has to be representative of the real thing. The cover, shape, proportions, binding, print finish and general appearance should not become misleading just because the image looks tidier.
AI can clean up the scene around the product. It cannot be used to make the product look like something we do not actually produce.
Accessibility
This is arguably where we have been most impressed by AI’s ability to speed up our developers’ work. This is a relatively recent use case for us, so we expect our approach to develop further. You can track these changes in our accessibility statement.
We use coding agents to review code with a specific focus on accessibility. This saves a tonne of time and catches things we would ordinarily miss.
Final thoughts
We are not trying to sound more advanced than we are. AI is useful in some places and in others, it adds very little, or creates problems that were not there before (at least in our workflow).
That is kind of the line for us. Where it helps, we will use it. Where it creates unneeded risk, we won’t.
What needs to change in the promotional calendar industry
Our view on three problems holding the promotional calendar industry back: weak digitalisation, opaque pricing, and unclear positioning.
Read article
20 May 2026
Dogs in the workplace
During busy calendar season, the dogs around our factory often do a better job of lifting spirits than anything else. Here at Think Promote, they’re always welcome.
Read article
14 May 2026
Caroline's five years with Think Promote
Caroline shares a few words about her five years in the production team at Think Promote, the support she has received, and what the workplace means to her.
Read article
10 May 2026
LED curing technology and UV varnish
After two years of development, Think Promote has successfully introduced LED UV varnish curing technology into its production process, replacing older mercury arc systems.
Read article
3 May 2026
Are promotional calendars still worth it?
In today’s crowded advertising landscape, printed marketing has to justify its place more than ever. So, are promotional calendars still worth it for businesses?
