Artificial intelligence (AI) isn’t coming. It’s here. Every day we hear of new advancements and applications of AI in the workplace. As the torrent of information accelerates, our questions, concerns, and curiosities grow.
Without a doubt, it’s important that we consider the implications of data security, regulatory factors, job displacement, ethics, and quality. However, these concerns have to be reconciled with how AI can augment our workflow, optimize efficiency, enhance overall productivity, and generate unique content. AI is not something publishers can ignore any longer.
As each of us explores AI, it can help to keep things simple––starting small, seeking to optimize workflow efficiency before pursuing more transformative uses. It’s already a challenge navigating through all the hype and flood of technological breakthroughs. Identifying simple AI projects that reduce busy work or address workflow functions that repeat themselves is a good place to start. Isolate the most productive test use cases and experiment with solutions that can be implemented quickly with minimal cost and risk. Here is a simple road map for how to think about onboarding AI into your organization.
- Discuss your readiness for AI with all internal staff. Perhaps some are already using various tools. Ultimately, it’s helpful to have key stakeholder buy-in and build consensus around goals, budget, strategy, and personnel capacity.
- Set up your team to use AI appropriately with a ChatGPT Teams or Enterprise account. This addresses the most important need: securing the input and output of data across all eligible departments and employees.
- Personalize ChatGPT for your specific publisher’s use, framing out your organization’s specific context along with specific instructions.
- Decide which department might have appropriate test use cases and conduct an audit with the team to identify these potential uses of AI. Prioritize this list and focus on the ones that offer high value to your team, are low risk, low cost, and have short implementation times.
- Create custom GPTs to accomplish the specific tasks for the test use cases and/or deploy other available AI tools.
- Train staff to use the custom GPTs or existing tools. Test and refine as needed. Evaluate ROI and plan next steps. Continue learning and experimenting.
Reflecting on Step 4, what might this look like within a publisher’s art department?
Before an art department considers viable test use cases, it can be helpful to decide how AI will not be used. A popular position taken by most designers in the publishing industry is that generative AI should not be used to produce book cover art. There are multiple fair use cases pending in the courts, and the US Copyright Office will not register copyright status for a work generated by a machine. As AI regulation takes shape, it seems prudent to avoid using AI-generated imagery on commercial goods. It can be helpful to think of AI less as the artist and more as the paintbrush. AI is a tool, one that can help designers accelerate learning, clarify strategy, build consensus, and streamline workflow.
Over the years, technological advances like the digital camera and computer have provided designers with powerful new tools, but they haven’t made designers less creative. Instead, they have expanded the scope of what designers can do and the problems they can solve. They offer new ways to express ideas, to experiment, and to push the boundaries of what’s possible in design.
While new technologies can automate certain aspects of the design process, the human element—creativity, critical thinking, aesthetic judgment, and understanding of human needs and behaviors—remains essential. The true art of design lies in how designers use these tools to bring their ideas to life.
Once basic standards are in place, use cases can be explored with more confidence. Three use cases are outlined below; however, these might not be appropriate for every publisher. For each use case below, it’s assumed all data is secure. Large language models (LLMs) are moving toward more security, not less. If you are not yet confident in the security language being offered by a particular LLM, just keep an eye out for updates. Better security options are coming. Security and privacy are paramount to resolve before moving forward with AI.
1. Manuscript Analysis
In a perfect world, book cover designers read every book they design. Alas, not every manuscript is adequately processed before the cover design process begins. Even when a manuscript is fully read, a designer will naturally filter the story, the insights, and the takeaways through their personal lens and bias. While this is certainly valuable, other stakeholders (author, agent, editor, marketer, etc.) may have other valuable input. Too often a book’s content is not sufficiently processed by a design team.
Great design compels buyers emotionally but is also meaningful and purposeful, accurately reflecting the content, the author’s brand, and the promise to the reader. The more intimate a designer is with the content, the more meaningful the solutions can become. AI tools now allow designers to instantly analyze full manuscripts and offer an objective report for all stakeholders to review.
Anthropic’s Claude LLM is a useful tool for manuscript analysis—summarizing content into an easily digestible word count, highlighting key scenes, visuals, themes, tone, objects, etc., to derive creative insights and build consensus on strategy from all stakeholders—before the design process begins.
2. Mood Boards
Sometimes the author or other stakeholders want to remain very involved in the design process. The process for cover design often involves the design team producing solutions with minimal direction on the front end and then decision-makers narrowing the strategic direction only after looking at many iterations of the cover design. In other words, designers are tasked with getting started, and once cover designs are rendered, a strategy then follows based on having something to look at and interact with.
While feedback and revisions are always part of the process, generating mood boards on the front end can align stakeholders much earlier in the process, leading to more efficient and meaningful design stages. These mood boards can help clarify such things as strategy, colors, composition, hierarchy, and tone, saving valuable time and resources later in the process. AI tools like Midjourney, Dall-e, and others can generate multiple sketch variations based on a manuscript analysis and well-composed prompts, building early consensus on direction before the actual design phase begins. These sketches are not intended to be cover solutions. Rather, they are intended to help solidify strategy earlier in the process so designers can be as intentional as possible in the actual design phase.
3. Marketing
Every publisher needs marketing assets to promote their books. The more important need, however, is for these assets to be highly customized, highly targeted, look fantastic, and be produced efficiently and cost-effectively. AI tools like Midjourney and Dall-e can easily create highly customized lifestyle backdrops that reflect scenes specific to a book’s content. The book can appear in highly relatable settings for various target audiences, leading to more meaningful positioning.
For publishers leery of uploading manuscripts, this is a way to engage AI with very limited data sharing, greatly reducing any assumed or real risk. The prompts to produce these customized backdrops can be void of any actual IP.
There is a growing need for AI solutions that not only drive workflow efficiency but also preserve the core values of the publishing industry. It’s appropriate to keep things simple at first, but get started. Engage in the conversation and learn all you can. Then seek to optimize workflow in at least one area. Evaluate the results and keep tweaking and testing until you have a new workflow that is saving you time and money. The smarter we get at publishing, the better we can serve our authors, elevating their stories and ideas with excellence and innovation.