How AI Is Transforming The Communications Industry: Lessons From FleishmanHillard

The communications industry is experiencing what Ephraim Cohen calls “change indigestion.” As global head of data and digital at FleishmanHillard, one of the world’s leading PR consultancies, Cohen has a front-row seat to perhaps the most dramatic transformation the industry has ever faced.

“In the past, we usually had decades to prepare” for technological shifts, Cohen tells me. “This is at such a rapid pace that it’s causing chaos. I do this full time, and I have a hard time keeping up with all the changes.”

The media ecosystem is fragmenting at breakneck speed, with audiences scattered across traditional outlets, influencers, Reddit threads, and AI platforms. Simultaneously, communications professionals must rapidly evolve their skills to serve clients navigating this landscape.

Democratizing AI Expertise Across The Enterprise

FleishmanHillard's response offers a blueprint for integrating AI without creating a two-tier system. The key, according to Cohen, is "hands-on keyboard learning."

Cohen’s point is that expertise comes from time spent doing the real work, learning by practice rather than theory. Because generative AI is still so new, he argues that everyone, including seasoned professionals, needs to re-engage with the basics and build hands-on capability. That means developing strong prompting, learning how to build AI agents, creating several of them, and then connecting them into a broader agentic solution that can deliver meaningful outcomes.

This builds on a foundation established five years ago when the firm began democratizing data fluency. Now, through Omnicom's platform, every professional has access to major frontier models. "Most of the major models in text like ChatGPT, Gemini, Nova, Llama," Cohen notes. "In video and image, models like Nano Banana Pro, Veo3, and so on."

Cohen describes their approach as a four-ingredient mixture: people with subject-matter expertise who know how to develop AI solutions, data fluency, AI technology, and knowledge bases digitized for AI agent access.

The Power Of Curated Knowledge Libraries

Knowledge bases are carefully curated libraries built around a specific topic or client context. Cohen compares this to the difference between working with a brilliant student who learned from an untidy, mixed-quality library, versus someone who has been given a well-organized reading list of trusted sources. With a curated knowledge base feeding the model, outputs tend to be more relevant, consistent and accurate because they are grounded in approved material rather than whatever the open internet happens to contain.

FleishmanHillard is building digitized libraries of proven case studies. "If we're doing a crisis simulation, we are building it off of a library of proven case studies and best practices, not whatever happens to live in ChatGPT. You get a much better, more accurate output."

This enables what Cohen calls a "bottom-up innovation wave." Some of the firm's most powerful solutions were developed by senior subject matter experts rather than technical specialists. "When you have the subject matter expert acting as the engineer, you get a better solution, both in terms of its usability and in terms of the output quality, and in terms of value to clients," Cohen observes.

How AI Changes Daily Workflows

Cohen walks me through a typical campaign workflow. It starts with synthetic audiences, which FleishmanHillard has been using since last year. "You always want to start with your audience insight," Cohen says. What once required expensive research now happens at a fraction of the cost and time.

Creative teams work with agents to develop concepts, then optimize content across channels. "We want people writing copy. You don't want to create AI slop," Cohen emphasizes. "You also want to take that copy and say, check it to make sure it's optimized for LinkedIn. Check it to make sure it's optimized for Instagram."

On the corporate side, teams use agents to forecast how different company actions will elicit stakeholder reactions. "We have about six months of real-world experience," Cohen notes. "We can say this is working, this is taking out a lot of guesswork, and it's adding a lot of precision to our counselors' direction to clients."

The platform makes this possible through elegant design. "You're literally going down and flipping switches on your boxes," Cohen describes. A single professional can move from getting insights to drafting an article to creating a graphic to producing a video, all within one interface.

Keeping Humans In The Driver's Seat

Despite powerful AI tools, Cohen is adamant that humans must lead the creative process. AI has a fundamental limitation, he explains: it's trained on everything that's already been created, which means its output can feel generic and derivative. "It is people that come with something original," Cohen emphasizes. The real value comes from human creativity using AI as a tool to test and refine ideas, not as a replacement for original thinking.

He compares AI to a talented assistant. "The best assistants are learning with you. They want to learn with you, and then they want to help test your thinking," Cohen explains. "If you want average work, outsourcing your job to AI is probably the fastest path to producing consistently average work. If you want great work, you will use it to test and refine."

Cohen sees a pattern emerging similar to when smartphone cameras arrived. Rather than eliminating professionals, the market exploded into two tiers: user-generated content and highly polished professional work. "There could be a market where AI-created copy, such as for hyper-personalized copy, is fine. Whereas high-stakes, high-impact hero content is going to be overseen by a person who's going to make sure every single word, every letter, every flow, every beat of what they're trying to do is as it should be."

Security Comes First

When ChatGPT launched, Omnicom, of which FleishmanHillard is a part, took an immediate pause. Employees rushed to use the new tool, but the firm quickly restricted access to ensure proper security and governance frameworks were in place first. Once Omnicom addressed the security concerns, the firm gave everyone the green light to move forward rapidly.

The platform operates in a highly secure environment where client work isn't used to train AI models. "The models do not learn off our work, period and end of story," Cohen states firmly.

Practical Lessons For Other Organizations

For companies beginning their own AI transformation, Cohen offers straightforward guidance. First, decide on your strategy. Do you want a few hero solutions developed by a central team, or do you want to lift your entire enterprise by making everyone capable? These approaches require fundamentally different resource allocation.

FleishmanHillard chose the enterprise-wide path, which Cohen believes yields better results over time. "Having that come out of a bottom-up innovation approach, we're seeing now a quick catch-up, and we're going to have more solutions that are more powerful because they were really built by people that know our clients, know their challenges and so solutions are much more bespoke to that."

Second, focus on fundamentals before getting distracted by shiny new objects. "Learn the basics," Cohen advises. "Once you've learned the basics, that will help you start coming up with higher-quality ideas of what your future may hold."

This hands-on learning creates what Cohen calls a "creativity tipping point." When people hear about AI in theory, they're excited but nervous about whether their skills are becoming outdated. "Once you get your hands on that keyboard and start learning how to create, you start thinking to yourself, oh, I can do this for my clients. I can do this for my work," he explains. "Then they get excited about the possibilities. And now they're organically creating new skills, they're creating their job in the future."

Finally, remember that the pace of change demands constant adaptation. "Everything I say to you is accurate as of today at 9:22 New York time," Cohen jokes. "By 9:23, it may change. That's an exaggeration, but not as much of an exaggeration as it was 10 years ago."

The Path Forward

Cohen's insights reveal a fundamental truth: the technology itself matters far less than how you deploy it. FleishmanHillard's success comes from ensuring every employee can use the tools effectively.

This democratization of capability creates an environment where innovation emerges from anywhere rather than being bottlenecked through a central team. As the industry evolves, this approach may prove the template for success: hands-on learning, bottom-up innovation and AI as an assistant to human creativity.

"We'll look back on this in a couple of years and say, well, of course this is just how we work," Cohen predicts. "But today it's a real push to get people inspired, excited and then to make it a habit."

That habit formation may be the most important factor in determining which organizations thrive in the age of AI.