You’ve got a great idea for a blog or LinkedIn article, if only your internal subject matter expert can find time for an interview.
Human experience and insight are key ingredients in quality content in the age of AI. Bots can churn out loads of content at a rapid pace, but without human perspective, it feels generic.
The bottleneck is finding the time to interview your CEO, CMO, Chief Content Officer or other internal SME who can provide the viewpoint you need to elevate your content.
We face this challenge at SmarterX, which is what led the team to start building an interview agent. It’s an experiment in progress but worth sharing for you to consider for your content and marketing team.
Original Insight Is the Scarce Resource
As AI-generated content becomes easier and cheaper to produce at scale, the things AI can’t replicate become more valuable: original quotes, firsthand experience, strong opinions, insight earned through actual work and experiments.
At SmarterX, the content team values an “expert-first” approach to content, where the original seed material comes from a human, and AI assists in drafting and distributing.
That means repurposing human content being created throughout our organization:
- Quotes and material from our weekly podcast, The Artificial Intelligence Show
- Information from AI Academy courses
- Findings from original research
- Q&As with people across the team to surface knowledge and perspective that AI can’t replicate
For that last one, Q&As, the problem is time and access. Subject matter experts aren’t always available when the content team needs them or can’t allocate the time for an interview, even a short, informal one.
The Experiment: An AI Interviewer
To solve this, the SmarterX team has built a MVP (minimal viable product) of an AI agent that can interview experts within ChatGPT.
It works like this:
Let’s say the content team has five posts planned for the week. For three of them, they want to incorporate original perspectives from an internal expert. Rather than scheduling a meeting or sending an email, the agent takes over the coordination and interview process.
It follows these steps:
Research first. Before the interview begins, the agent researches the assignment, the topic, the audience, and relevant things the expert has said in recent podcast episodes. It runs several research tasks in parallel, so it’s prepared with real context.
Adaptive interviewing. The agent then interviews the expert one question at a time. Each answer helps shape follow-up questions. The goal is to draw out useful, specific perspectives not just generic or surface-level takes.
Creates a usable brief at the end. When the interview is done, the agent produces a structured brief for the content team: main takeaways, pull quotes, a full transcript, and relevant source context. It also fact-checks claims the expert made during the interview. Fact-checking matters because even subject matter experts make mistakes, and building verification into the process saves the content team a step.
As discussed on Episode 233 of The Artificial Intelligence Show, the interview and output brief are already happening.
But someone still has to manually kick off the process, telling the agent what the posts are about and initiating the interview. The team is working toward automating this. For example, having someone tag a post in the project management system as needing an expert interview, which would then prompt the agent to reach out and schedule one autonomously.
What It’s Actually Solving
The agent is not a content generation tool but instead is making it easier for a busy expert to more easily contribute their thinking to content. For example, an SME could conduct their AI agent interview during a walk or a morning commute, instead of taking time out of their work day.
Whether this scales meaningfully, and whether the briefs it produces translate into noticeably better content, is something the team is still measuring. It’s an early test of a narrowly scoped agentic workflow, one that seems achievable without a long build cycle.
The takeaway is this: As AI changes the speed and scale of production, content teams should be redirecting their energy to ensuring what’s produced is insightful, original and based on what only humans can provide.
This post draws on the AI Use Case Spotlight segment of Episode 233 of The Artificial Intelligence Show, hosted by Paul Roetzer and Mike Kaput. To listen to the full Episode 228 of The Artificial Intelligence podcast, visit: https://podcast.smarterx.ai/shownotes/233
For more on building AI-ready marketing teams, explore AI Academy at academy.smarterx.ai.
