How AI Is Transforming Communications: Lessons From FleishmanHillard
The industry has a new problem: change indigestion. Channels multiply, audiences scatter, and the pace outstrips old planning cycles. As global head of data and digital at FleishmanHillard, Ephraim Cohen calls it out plainly-keeping up is hard, even for people who live in this work every day.
The answer isn't a bigger "innovation team." It's practical, hands-on capability for everyone. Cohen's core belief: expertise comes from time on the keyboard-prompting, building agents, connecting them into workflows, and shipping outcomes.
The Operating Model: Four Ingredients That Actually Work
FleishmanHillard's approach blends people, process, and tech without creating a haves-and-have-nots divide. The focus is democratized skill-building across the firm, supported by integrated tools.
- Subject-matter experts who build - The people closest to client problems design the AI solutions.
- Data fluency for all - A culture set years ago now pays off as AI scales.
- Access to top models - Through Omnicom's platform, teams can use ChatGPT, Gemini, Nova, Llama for text and models like Nano Banana Pro and Veo3 for video and image.
- Digitized knowledge bases - Curated libraries tuned for agents to reference.
Curated Knowledge Beats "Whatever's On The Internet"
Give an agent a messy feed, get messy output. Give it a vetted library, get practical, consistent results. FleishmanHillard builds topic- and client-specific knowledge bases-approved case studies, best practices, and reference material.
Example: crisis simulations driven by a library of proven cases produce guidance you can trust. The bonus effect is a bottom-up innovation wave, where senior counselors create stronger tools because they know the nuance that matters.
What Daily Work Looks Like With AI
Start with your audience. Teams use synthetic audiences to generate insight at a fraction of the old cost and time. Creative groups then partner with agents to develop concepts, iterate, and pressure-test across channels.
The bar for content stays human. Write the copy yourself; use AI to optimize it for LinkedIn, Instagram, and other formats so distribution hits the mark. On the corporate side, agents forecast stakeholder reactions to different moves-reducing guesswork and adding precision to counsel.
All of this happens in one interface. As Cohen puts it, you're flipping switches: insight, article draft, graphic, video-end to end without tool-hopping.
AI Is The Assistant, Not The Author
AI is trained on what already exists. That's why it can drift generic. People bring the original spark; agents help test, refine, and stress-test ideas fast.
Think smartphone cameras: we didn't lose pros, we got clearer tiers. At scale, AI-generated, hyper-personalized copy will be fine. For high-stakes "hero" content, a human will tune every beat.
Security Before Speed
When ChatGPT launched, Omnicom paused first, then moved fast with guardrails. Now teams operate in a secure environment where client work is never used to train models. As Cohen says, the models do not learn from their work-period.
Practical Steps For PR And Communications Leaders
- Pick a path - Central "hero" builds or enterprise-wide uplift. If you want compounding returns, make everyone capable and let solutions emerge from real client problems.
- Make hands-on learning mandatory - Teach prompting, agent building, evaluation loops, and channel optimization. Ship small wins weekly to build confidence and momentum.
- Stand up curated knowledge bases - Start with crisis, corporate reputation, and key clients. Version them. Assign owners. Keep them clean.
- Instrument your workflow - Track time saved, quality lifts, and performance by channel. Promote what works; retire what doesn't.
- Codify security and governance - Clear policies on data handling, disclosures, and review. Treat it like media compliance: baked-in, not optional.
- Expect constant change - Toolsets shift weekly. Build habits and principles that outlast model versions.
A Simple AI Workflow You Can Roll Out This Month
- Audience - Generate synthetic personas and pain points; validate with a quick pulse check or SME review.
- Concept - Co-create angles with an agent; force-rank by potential impact and risk.
- Draft - Human writes. Agent suggests variants for tone, channel, and length.
- Optimize - Channel-specific checks (LinkedIn, Instagram, newsroom). Add alt text, CTAs, and compliance notes.
- Predict - Run stakeholder-reaction scenarios for corporate updates or issues responses.
- Package - Auto-generate social snippets, a graphic, and a short video cut where relevant.
- Review - Human sign-off. Publish. Measure. Feed results back into your library.
Why This Works
The tech matters less than how your people use it. FleishmanHillard's edge is giving every professional access and training, then grounding the system in curated knowledge. Innovation shows up everywhere, not just in a lab.
Make hands-on learning a habit. Encourage bottom-up builds. Treat AI as the assistant that sharpens your best ideas. Do that, and in two years, this won't feel novel-it'll just be how your team works.
Want structured upskilling?
If you're building capability across a PR or comms team, consider a focused path like the AI Certification for Marketing Specialists for practical, job-ready skills.
Your membership also unlocks:
