Every publisher wants to make workflows more efficient. But how can organizations successfully integrate AI into existing processes when, in practice, many are still dealing with fragmented systems, disconnected teams, and employees experimenting with AI without clear organizational guidelines?

For Larrauri, publishers should approach AI adoption in three phases:

1. Executive alignment: The first phase begins with the CEO, senior executives, and the executive committee, who must first develop a solid understanding of AI—both its technical capabilities and its human impact.

   “That's very clear to me because sometimes you talk to senior managers at this level and they don't know exactly what they are talking about,” Larrauri said.

2. Managerial involvement: Only after leadership understands the right questions to ask should middle managers and operational teams become fully involved. Otherwise, she argued, publishers will struggle to maintain control over how AI is introduced across the organization.

   “There are a lot of studies showing that middle managers are struggling—not knowing what to do, having to face expectations transferred from CEOs, and dealing with the pressure of team members using plenty of tools without any deep knowledge,” she said.

3. Workflow redesign: Once the organization reaches a higher level of maturity, publishers can move into the third phase, in which teams begin redesigning workflows around AI rather than simply adding new tools to existing processes.

Publishers also need to recognize a series of risks that are already emerging as AI adoption accelerates. Among them are limited C-level understanding, uncontrolled experimentation by employees, and the absence of governance frameworks capable of guiding responsible AI use.

“In marketing and sales, employees start using these models on a private basis, accessing personal accounts while using company content—which is a total mess because a lot of private content is going into private accounts,” Larrauri warned. “This is a risk that CEOs and senior managers are not identifying: that people are using these tools with the company's private content.”

Another challenge, she noted, is the different paces at which AI adoption takes place inside organizations. While many employees are already incorporating AI tools into their daily work, senior leadership often remains at an earlier stage of understanding, leaving middle managers caught between executive expectations and rapidly evolving team practices.

This disconnect creates a new management dilemma. Without clear governance, employees who use AI to complete tasks more quickly may be perceived as more productive than colleagues who spend additional time reviewing, verifying, and challenging AI-generated outputs.

For Larrauri, another overlooked issue is that AI adoption does not stop at the publisher's walls. Libraries, booksellers, and other partners are also developing their own AI strategies, often without sufficient dialogue across the industry.

As a result, publishers are failing to communicate how they expect their content to be used throughout the value chain. She cited the example of AI-generated audiobooks. While many publishers are turning to AI-assisted production to make audiobook publishing economically sustainable, some public libraries have already begun rejecting AI-produced audiobooks in public tenders.

“If publishers don't explain why they are using AI, libraries don't see the value,” she said. “Unless we discuss this together—unless retailers, booksellers, and publishers understand each other and have forums together—there will be friction. And this is happening today on a daily basis.”

Asked what publishers should prioritize in the short term, Larrauri returned to the message that underpinned her entire presentation: start with leadership.

She recommended that publishers dedicate the first 90 days to:

• Training senior executives

• Assessing available AI tools

• Evaluating existing workflows

• Drafting an initial governance framework capable of guiding future implementation

At the same time, she stressed that successful AI adoption cannot be imposed strictly top-down. Middle managers and operational teams must be involved throughout the process, contributing ideas and sharing practical knowledge about how work is actually done.

Ultimately, Larrauri argued, the success of AI adoption depends less on technology than on communication.

“Only if you communicate well within the company and care about transferring what you are doing to the entire organization—not only with AI or technology, but with everything—only then can you be successful,” she concluded. “Otherwise, there will be a lot of noise in the middle and the models probably won't work as well as they could."