This morning, I asked Claude to create a content library for OriginalVoices. Ten blog posts, each covering a different industry and use case. Each one grounded in real audience research, with real quotes from real people.
I didn’t brief a research agency. I didn’t run a survey. I didn’t wait two weeks for a report and another two for the content.
I opened Claude Code, described the audiences I wanted to understand, and let the agent take it from there. It designed the research questions, queried thousands of real people through OriginalVoices, analysed the responses, identified the patterns, pulled the most compelling quotes, and wrote ten complete blog posts — each one directly shaped by what real consumers actually said.
I edited and approved. The agent did everything else.
The whole thing — research, analysis, and writing — took 15 minutes.
This is what happens when AI agents can actually understand people.
And to be clear: these aren’t AI-generated personas or synthetic audiences. There’s no model guessing what a “typical millennial” might think. Every response came from a real person’s Digital Twin — trained, updated, and validated by the individual it represents. Real people, real perspectives, at the speed AI agents need.
The content isn’t special because it was fast. It’s special because it’s relevant. Every quote reflects how real consumers actually think and feel. The headlines aren’t clickbait — they’re findings. The arguments aren’t hypothetical — they’re evidence-based.
That’s the difference between AI content generated from training data and AI content informed by real human context. One sounds plausible. The other sounds true — because it is.
The workflow is simple. And it changes everything.
Here’s what actually happened, step by step:
- I described the audiences in plain language — “Gen Z adults aged 18-25 who buy skincare online” or “UK adults who use traditional high street banks”
- The agent designed research questions for each audience and ran all ten studies simultaneously through OriginalVoices
- Real people’s Digital Twins responded within seconds — not summaries, not averages, but individual perspectives with the texture and nuance of real human thought
- The agent analysed the responses, identified the non-obvious patterns, pulled the most compelling quotes, and wrote ten complete blog posts directly from the research
- I reviewed, edited, and approved
No research brief. No waiting. No separate tools. The research, the analysis, and the creation happened in the same workflow, in the same session, in the same conversation. My role was editorial — direction and judgement. The agent handled everything else.
Speed matters. But depth matters more.
The speed is impressive — seconds instead of weeks. That’s what Agents need. But what makes this genuinely different is the depth. Each response came from a real person who has trained, updated, and validated their own Digital Twin. These aren’t synthetic personas generated from demographic data. They’re representations of real individuals who have actively taught their Digital Twin how they think, what they value, and how they make decisions.
That means the responses have texture. When a Gen Z consumer says “I prefer emails that feel like they care about my skin, not just my wallet,” that’s not a language model guessing what Gen Z might say. It’s a real person’s perspective, expressed through their Digital Twin.
And if I wanted to go deeper — ask follow-up questions, probe a specific finding, test a hypothesis that emerged from the first round — I could. The same real people are available to query again. The research doesn’t end when the first responses come back. It can evolve in real time, iterating as fast as the questions form.
Our methodology also enables us to go back to the real people behind the Digital Twins for further validation. If a finding is surprising or high-stakes, we can verify it directly with the humans who generated the insight. The Digital Twins are the speed layer. The real people are the truth layer. Both work together.
This isn’t a content trick. It’s infrastructure.
What happened this morning with blog posts is just one application. An AI agent with access to OriginalVoices could do the same with anything:
- A product team querying real users before building a feature, understanding not just whether people want it but why, and what would make them actually use it
- A creative agency testing ad concepts across five audience segments simultaneously, getting honest reactions before a single pound of media spend
- A brand strategist validating positioning in real time, finding out whether the message lands or falls flat with the people who matter
- An e-commerce platform generating product descriptions informed by what actual buyers care about — not keyword-stuffed copy, but language that reflects real purchase drivers
- A startup founder testing their pitch with their target market before walking into a funding meeting
The pattern is always the same: understand people first, then act. The problem has always been that understanding people takes too long, costs too much, and arrives too late to shape the work.
OriginalVoices removes that bottleneck entirely.
Why this matters now.
AI agents are getting more capable every month. They can write, code, analyse, plan, and execute across almost any domain. But there’s one thing they can’t do on their own: understand what real people think and feel right now.
Training data is historical. It tells you what people (or the internet) said months or years ago, aggregated and anonymised beyond recognition. Web search is useful, but everyone has access to this data and it’s rigid - you’re constrained by what exists. Neither tells what a 32-year-old mother in Manchester thinks about your sustainability messaging, or whether a gym-goer in Austin would actually pay for your AI coach feature, or what makes a Gen Z consumer in London open an email instead of deleting it.
OriginalVoices gives AI agents that missing layer — real-time access to authentic human perspectives. Not synthetic. Not simulated. Real people who have chosen to train AI representations of themselves, who actively update and validate those representations, and who are available to query on any topic, for any purpose, in seconds.
That’s the future we’re building. Not AI that replaces human understanding. AI that finally has access to it.
Ten blog posts. Ten audiences. Ten different industries. 20 minutes.
Every piece is grounded in what real people actually said. Every insight can be traced back to a real human perspective. Every piece of content is more relevant, more credible, and more useful because it started with the people it was written for.
That’s not a productivity hack. That’s a fundamentally different way to create.