Stop asking AI to write. Here’s what the NowThis audience strategist does instead.

Sirui Hua of NowThis explains why communicators should stop thinking of AI as a writer and start thinking of it like a team of specialists.

Sirui Hua, head of audience and analytics at NowThis,  is the infrastructure behind the storytellers, the person who makes sure great work finds the audience it deserves. At NowThis, Hua leads audience strategy and AI innovation for the global video brand, building tools that support creators and turn data into faster, smarter creative decisions.

Hua will speak at Ragan’s upcoming virtual AI Communications Conference, where he’ll break down agentic workflows and how they’re starting to show up in comms, from multi-step research and drafting to automated content flows. 

 What first sparked your interest in audience strategy?

My first job was at a PR agency in China, helping global brands build their social presence on Weibo and WeChat. That’s where I first saw the power of social media and realized how much I enjoyed connecting brands directly with their audiences. The real spark came later, from production. I joined NowThis 10 years ago as an associate producer, learning how a story actually gets built, from the opening hook to the emotional payoff. So when I eventually moved into analytics, I wasn’t just looking at numbers on a chart. I could see the narrative beats inside them. I’ve believed ever since that even the most brilliant storytelling can get lost without the right strategy and support, and I decided my job was to become the infrastructure that supports creators: decoding audience interests, algorithms and trends so storytellers can focus on making magic, knowing their work will find the audience it deserves.

 If you’re only using AI to draft content, what opportunities are you missing?

Honestly, almost all of them. Drafting is the smallest part of what the latest AI tools can do now. If AI only writes your first drafts, you’re still the operator: you’re doing the research, moving the files, formatting the report, and hitting send. The shift I keep pushing on is going from operator to manager or editor. You set the goal and the standard, the AI runs the steps in the middle, and you stay where your judgment matters most, which is deciding what’s worth doing and whether the result is good enough to ship. Using a frontier model just to write drafts is like buying the latest smartphone and only making phone calls with it. The blank page used to be the bottleneck. Now the bottleneck is the workflow around the draft, and the context you’ve built up that the AI can reuse the next time.

 Can you give us a sneak peek at one real-world workflow you’ll walk through during the session?

The one I’m excited to show is our paid social agent, a tool my team uses to run paid campaigns on Meta and TikTok. Under the hood, it’s a multi-agent system, a small team of specialist agents that each own one step of the campaign lifecycle. One sets up the boosts, one pulls performance into client reports, and one watches pacing and handles optimization. Before any of them execute or recommend anything, they gather context by reading the places where campaign reality actually lives: the Slack threads, the contract PDFs, the emails, and our data warehouse. Then the pacing agent checks delivery against what we promised the client and drafts the changes it thinks we should make, like reallocating budget, pausing an underperforming ad, or spinning up a new one. Nothing executes without a human review and approval.

What’s something communicators are still doing manually that AI can already help with?

Status reporting. Communicators spend an astonishing amount of time assembling information that already exists: pulling numbers from different platforms, checking whether the client replied, and turning it all into a recap. Almost none of that is judgment work, and agents are already good at it. Audience comments are another one. Nobody can read 20,000 comments, so most teams sample a few and go by vibes, when AI can actually read all of them and tell you what’s resonating and with whom. If you’re wondering where to start, my rule is simple. Find the recurring task someone on your team genuinely hates, automate that one end-to-end, and keep a human at the front and the back, because you’re still responsible for whatever AI ships under your name. One shipped workflow will teach you more than dozens of AI strategy decks.

What book, podcast or creator has influenced your thinking lately?

Two podcast episodes. Anthropic co-founder Jack Clark and its head of economics, Peter McCrory, went on Odd Lots and said, “You shouldn’t buy AI like software, you should think of it as hiring thousands of chiefs of staff.” That reframed how I think about rollout, because a chief of staff is only as good as the access and context you give them. The other is Derek Thompson’s Plain English episode with Georgetown neuroscientist Adam Green, about a study that found AI made student writing sound more creative while making the ideas measurably more alike. It resonates with me a lot: AI raises the floor, but your taste sets the ceiling.

What’s a hobby or passion that has nothing to do with work?

Long-distance running. I live in Jersey City and run along the Hudson River waterfront, with the Statue of Liberty and the Manhattan skyline as my backdrop. There’s something about the rhythmic exertion that clears out the mental clutter. My best ideas rarely show up at my desk; they usually show up around mile two to mile three.

Don’t miss Sirui’s insights on agentic AI frameworks at Ragan’s virtual AI Communications Conference on July 22. Register now.

Isis Simpson-Mersha is a conference producer/ reporter for Ragan. Follow her on LinkedIn.

The post Stop asking AI to write. Here’s what the NowThis audience strategist does instead. appeared first on PR Daily.

Source_link