AI has arrived in publishing as every major shift has: with promise, pressure, and a lot of confusion. Independent publishers have seen this before. I have been on the front lines of all the digital changes from the web to social media to e-books and digital transformation. Each wave brought real opportunity, and just as much uncertainty. The difference this time is the speed of change. AI is changing every week. Just as you learn one tool, another one comes along with even more features.

I have spoken to many people in publishing, and the overall take is that there is a lot of pressure to “do something with AI,” but they’re not sure what that means in practice. They worry about brand integrity, author trust, and overburdening already stretched teams.

A widely cited MIT-associated study found that most corporate AI pilots fail to deliver the expected ROI. Another insight from Harvard Business Review is even more telling: AI doesn’t necessarily reduce work; it often intensifies it. Both of these insights match my experience.

Most AI rollouts fail for three simple reasons. First, teams treat AI like software. It’s not. You can’t install it, train people once, and expect consistent results. AI requires personal interaction, judgment, iteration, and ongoing refinement. It’s not a one-size-fits-all. Many people love Claude, but for me, it does not work as well as ChatGPT or Gemini. That does not mean I know something they don’t. AI tools respond to each of us and to our work/needs differently, depending on how we train them.

Second, leadership underestimates the emotional labor. Editors, marketers, and publicists don’t just produce content; they make a million nuanced decisions about voice, tone, positioning, and trust without even thinking. When AI enters the workflow, those decisions don’t disappear. They become more complex. I’ve been using AI every day for almost four years, and my conclusion is that humans are freaking brilliant. AI is amazing, and I love it, but it cannot do what humans can.

Third, there is no time for experimentation. All of us are already busy. We need simple solutions that work all the time. Software gives us that. You learn it once, and you go about your work. AI is different because it is not predictable. Just today, I was working on a timeline for a client’s book launch. A week ago, Gemini created a fabulous sheet with my workflow. Today, when I tried to duplicate that process for another client, it created a Google Doc even though I asked for a Google Sheet. I didn’t have time to figure out what was different. I thought I had followed the same prompt. But honestly, that’s just the nature of AI. It is not predictable. Even with a Custom GPT, I can get different results.

Does any of this sound familiar? It can be frustrating.

I believe AI adoption should be centered around humans, not around AI tools. Investing in humans will, in the long run, benefit us more than replacing them with AI. AI is not creative; it is just regurgitating what’s already out there. Only humans can create something brand new, one of a kind, like a book.

To make AI work better for my team and me, I came up with a simple framework for using it effectively: Align, Invest, Amplify.

1. Align: Start with Values, Policies, and Workflows

Before you introduce AI into your organization, you need a clear policy. You need very clear rules about the use of AI. Clear enough that everyone, including your authors, knows exactly where and when AI can be used and where it is not used. I read that Brooke Warner at She Writes Press will not use AI for cover design. That is clear and aligned with their values. No confusion.

When I speak to most people in publishing, they have not discussed a policy, which means use is inconsistent across users. Almost everyone is using AI in one way or another, but most people don’t declare it, which only adds to the confusion and critical mistakes.

Here are the areas to discuss:

• Where is AI appropriate in our work?
• Where should it not be used?
• Who is responsible for human review?
• Who is accountable if something goes wrong?

Early on, a mistake was made in a client’s proposal. We used AI. AI made a mistake (as it is prone to do). We didn’t review it thoroughly, and I was utterly embarrassed. We lost that project because we lost the client’s trust. That was an important lesson and informed a lot of my work with AI going forward.

It’s important to know when to use AI and when not to. Using AI to draft metadata or summarize internal documents may be low-risk and high-value. Using AI to shape an author’s voice or interpret sensitive content requires much more care and should often remain in human hands.

Alignment also means mapping your actual workflows. Where are the bottlenecks? Where are people spending time on repetitive or low-leverage tasks? That’s where AI can help and make a big difference.

Without mapping workflows, we all just run random experiments, which are cool but often yield inconsistent results and growing skepticism. Mapping workflows is not sexy, but it is the most important step.

2. Invest: Build AI Literacy, Not Tool Expertise

I wish we could all just learn AI the way they learned in The Matrix. Just plug us in, and the knowledge would just download into our brains, but that is not our reality.

We need to invest in training and experimentation. Both take time and money. So, before you can use AI in a truly beneficial way, we need to try things out. Most training programs focus on tools. That’s a mistake, as AI tools are changing too fast to be the focus. We need to change our mindset and learn for ourselves where AI makes sense and reduces work, and where it adds more work and emotional labor.

AI literacy includes two things:

Understanding what’s possible – Learning what’s possible as the technology changes. Understanding how prompts work, which tools actually help, and which waste time. I see a giddiness when people give a task to AI that would have taken them hours. It’s so satisfying. However, that is not AI adoption. That is AI experimentation. It is a very important step in figuring out what’s possible, which is why we need to set time aside for experimentation.

Recognizing limitations – AI is confident but not always correct. Hallucinations are more common than anyone would like. We need a system for verifying the information. One idea is to highlight everything generated by AI in orange and not remove it until a human has reviewed and verified it. This is especially important in publishing, where credibility is everything, as we all learned from the Shy Girl controversy.

Smaller publishers actually have an advantage. You don’t need massive training programs. You can start with small, focused learning moments tied directly to real work.

3. Amplify: Create a Culture of Shared Learning

I recently attended a party with other authors. Each of us was sharing our giddiness at finding good ways to use AI. It made me realize how important that is. None of us can learn everything we need to learn about AI on our own. We need support inside and outside our companies.

Even when teams experiment with AI, they do so in isolation. One person finds a workflow that saves hours, but no one else knows about it. Another tries something that fails, but the lesson isn’t shared. Over time, this slows adoption and reinforces the idea that AI isn’t working.

Instead, we need AI learnings to compound. You can set up a brainstorming session where people share their wins and failures with AI. This doesn’t need to be formal. A short weekly check-in, or a shared Google Sheet with ideas, tools, and prompts. Encourage stories of failure, because we often learn more from those. You want AI adoption to be organic and fun, not forced and tortured.

AI adoption is not a tool decision. It’s a leadership decision. Leaders set the tone for how AI is introduced, discussed, and evaluated.

That includes:

• Align: Defining values and guardrails early.
• Invest: Supporting learning without demanding immediate results.
• Amplify: Encouraging experimentation without penalizing failure.

There is widespread fear that AI will take our jobs. A human-centered AI adoption plan protects the humans. AI should be a tool for us, not a replacement.

AI is not going away. Independent publishers are well-positioned for AI adoption. You’re closer to your authors. You understand your audiences, and you can move faster than larger organizations. Thoughtful AI adoption may feel slower at the start, but it works much better over time.

---