AI for Time Travel

I Called My Grandparents in the 1970s

My brother is into genealogy. He's the kind of person who actually reads through old family letters—boxes of them, handwritten, folded, posted. Most people don't. It feels like homework.

One day he showed me a letter from our aunt Carol, written to her parents back in the 1970s. Ordinary family correspondence: news about the kids, plans for a visit, everyday life. The kind of artefact that's technically rich—full of context and personality—but emotionally distant. You have to want to spend time with it. You have to be into genealogy.

I looked at that letter and saw something else. Not a document. A conversation trapped in the wrong medium.

Output 1: The Radio Play

First, I created what you might call a "radio play"—a produced audio piece where you hear Carol calling her father. It's the 1970s on the call. He doesn't know about the future. He's just... Dad. The conversation contains everything from the letter: family news, the kids, plans for visits, period-appropriate concerns.

But it's no longer text you read. It's a conversation you listen to.

Output 2: The Interactive Phone Call

This is where it gets different.

You can actually ring my grandparents. Or ring Carol herself. And have a real conversation—not listening to a recording, but in the 1970s, talking to them. The AI knows what was in the letter. Knows the family context. Responds as them, in their voice, in their time.

You're not consuming a narrative. You're participating in a conversation.

"Same source material. Two completely different experiences. One is a produced piece you consume. The other is a conversation you participate in."

The radio play is passive—cinema-style, crafted narrative. The interactive version is active—you ask questions, you explore, you have agency.

Both are built from the same fragmentary evidence: one letter. Both feel emotionally real despite being synthetic.

My Brother's Reaction

"Gobsmacked" is the word he used.

But not because it sounded realistic. That's the obvious reaction, and it's the least interesting one. He was gobsmacked because he suddenly saw his genealogy research differently.

All those letters and documents he'd been collecting for years—carefully preserving, cataloguing, transcribing—they didn't have to stay as archives. They could become conversations.

What Actually Changed?

The information was always there. The letter existed. The facts were known.

What changed was the medium.

We're not built for documents. We're built for conversations. Families remember through stories told out loud, not through filing systems.

AI didn't create new information. It translated existing information into a form humans actually want to engage with.

Stumbling Into Something Bigger

As I watched my brother's reaction, something clicked. This wasn't just a genealogy hack. This wasn't just a "cool AI project."

The experience felt... larger. Like I'd accidentally walked into a room that connected to many other rooms.

I started thinking about other things this reminded me of:

These all felt... related somehow. But they'd never connected in my mind before.

Then I realised: they're all variations of the same thing.

"AI as a way to let emotionally meaningful conversations happen across time—with people who exist outside their normal time slot."

It turns out, those questions have answers. And the answers are stranger—and more useful—than I expected.

But first, let me show you five wildly different ideas that turned out to be the same idea wearing different clothes.

Five Ideas That Seem Unrelated

After the Carol experiment, I started noticing other ideas rattling around in my head. They'd never connected before. But now they seemed to rhyme. Let me walk you through five of them.

These will seem unrelated at first. Your job: spot the common thread. By the end of this chapter, you should see it. And once you see it, you can't unsee it.

Business Applications Already Deployed

Chapters 1 through 4 established the pattern, the framework, and the timing thesis. But pattern recognition isn't enough for business readers. You need proof that this works and delivers ROI.

This chapter presents five domains where Time-Shifted Proxies are already deployed — not as speculation, but with measurable business outcomes validated in production environments.

Domain 1: Sales & Customer Success

AI-simulated customers for training represent one of the most mature applications of the Time-Shifted Proxy pattern. Sales representatives practice discovery calls, product demonstrations, objection handling, and compliance scenarios with AI avatars that simulate realistic customer behaviour.

The traditional alternative — roleplay with colleagues — suffers from predictable failure modes. Colleagues don't know how to object convincingly, iterations are limited before participants lose interest, and the social awkwardness undermines psychological safety. AI roleplay eliminates these constraints: unlimited practice, realistic resistance, immediate feedback, no social cost.

Oracle Case Study

Oracle adopted Second Nature's AI training platform with striking results. While the volume of introductory calls remained constant, the quality of outcomes transformed: opportunities per sales representative per month rose from 2.78 to 6.02, and new logos per rep per month grew from 0.49 to 1.04 — more than doubling pipeline creation efficiency.¹

Time savings compound these gains. In the Outreach Prospecting 2025 report, 100% of AI-powered SDR users reported measurable time savings, with nearly 40% reclaiming 4 to 7 hours per week.²

"AI roleplays provide a psychologically safe environment that can help learners feel more prepared and confident in their ability to handle difficult situations."

Deployment Variations

Domain 2: Medical Education

Medical communication training faces unique constraints: conversations are high-stakes (patient outcomes directly affected), low-opportunity (cannot ethically practise on real patients), and emotionally demanding (breaking bad news, end-of-life discussions, complex diagnoses). AI simulation provides the missing infrastructure for safe, repeatable practice.

Oxford Medical Simulation deploys voice-controlled virtual patients using large language models to enable natural conversation. The system creates what researchers describe as "a psychologically safe environment" where learners can rehearse difficult conversations without risk to actual patients.³

Mayo Clinic Study

Mayo Clinic deployed an AI-based communication platform with Arizona-based PGY-2 neurology residents to improve communication with stroke patients in emergency settings — scenarios where communication errors carry severe consequences. The results validated both efficacy and acceptance: 80% of trainees reported they would use the AI tool again for future communication practice.⁴

Cornell's MedSimAI

Researchers at Cornell University developed MedSimAI, a large language model system that creates realistic interactive patient conversations. In a study involving 104 medical students, 78% identified focused history-taking as the platform's most valuable function, while 62% highlighted question phrasing practice.⁵

Domain 3: Marketing & Video Testimonials

Video testimonials represent a direct application of the "future customer talking to present prospect" pattern. The Time-Shifted Proxy manifests as AI-generated customer avatars delivering composite experiences built from multiple real customer stories.

The Retention Advantage

Customers remember 95% of a video testimonial compared to only 10% of written reviews — a near 10x differential in retention. This isn't surprising: video engages multiple sensory channels and emotional processing pathways that text cannot access.⁶

Production Economics

AI tools reduce video testimonial production costs by up to 70% compared to traditional filming, eliminating the need for actors, studios, and production crews. Professional-quality testimonials can now be generated in minutes rather than days.⁷

"85% of consumers now expect personalised interactions from brands. For businesses still clinging to one-size-fits-all strategies, the future is bleak."

Emotional AI Market Growth

The global emotion AI market reached $2.9 billion in 2024 and is projected to grow at 21.7% CAGR through 2034. Sales teams deploy emotional AI to read facial expressions during product demonstrations and adjust responses when confusion is detected, driving measurably higher conversion rates.⁸

A Realeyes study analysing 130 automotive advertisements found a direct correlation between emotional engagement and social media performance. Volkswagen's narrative-driven "The Force" advertisement generated 86% higher brand exposure and 329% more social media interactions than Ford Fiesta's product-focused approach.⁹

Domain 4: Decision Support

Temporal discounting — the human tendency to heavily discount future pleasure and pain — undermines long-term decision-making across financial planning, health behaviours, and career choices. MIT researchers developed "Future You" to address this cognitive bias through conversational AI.

MIT's Future You Tool

The system enables users to have text-based conversations with an AI-generated simulation of their potential future self at age 60. Users provide information about current circumstances and aspirations; the large language model constructs a plausible future persona that can answer questions, offer advice, and describe life trajectories based on present decisions.¹⁰

The initial study included 344 participants aged 18 to 30. Results showed that users experienced decreased anxiety and lower levels of negative emotions after conversing with their future selves. More significantly, participants reported a stronger connection to their future — a psychological shift linked to improved decision-making regarding health, education, and financial planning.¹¹

The mechanism works by transforming "future me" from an abstract concept into a conversational partner. When future consequences feel personally real rather than intellectually distant, temporal discounting weakens and decision quality improves.

Domain 5: Memorial Services

Memorial AI services represent the most emotionally charged application of Time-Shifted Proxies: conversational interfaces with deceased loved ones. Multiple commercial platforms now offer these services at accessible price points.

Commercial Deployments

Market Size

China's virtual digital human industry reached a core market size of approximately $6.7 billion USD by 2025, with surrounding markets driven by the technology approaching $90 billion. This represents explosive growth from 2022 figures and validates substantial commercial demand for digital resurrection services.¹⁵

Consumer Acceptance with Consent

Research by Masaki Iwasaki found that 58% of U.S. survey respondents support digital resurrection when the deceased had explicitly consented. Acceptance plummets to 3% when consent is absent — demonstrating that demand exists but is contingent on ethical frameworks that respect autonomy and consent.¹⁶

"The question isn't whether this works. It's: which conversation across time would unlock the most value in your domain?"

The Design Space Matrix

You understand the pattern. You've seen proof it works. The question now: how do you apply this to your domain?

Most innovation frameworks give you principles. This chapter gives you something better: a systematic idea-generation tool. A design space matrix where every combination of axes produces a potential product, feature, or use case.

Not random brainstorming. Deliberate exploration.

The solution: four axes that combine to form hundreds of possible products. Let's walk through each one.

Axis 1: Time Direction

The first decision: where in time does this conversation exist?

Each direction unlocks different value. Past gives you memory and learning. Future gives you decision support and persuasion. Parallel gives you scalable training and pattern recognition.

Axis 2: Role to Simulate

The second decision: who is speaking?

Axis 3: Medium Level

The third decision: what form does the interaction take?

This isn't just about production cost. The medium fundamentally shapes the emotional experience and, therefore, the product's effectiveness.

"Phone call to grandma hits differently than reading her letter. The medium isn't decoration—it's the product."

Same underlying AI. Same data source. Radically different user experience. Choose deliberately based on the emotional outcome you need.

Axis 4: Purpose

The fourth decision: what outcome are you trying to achieve?

Purpose determines success metrics. A training application succeeds when performance improves. A marketing application succeeds when conversion increases. Choose the wrong purpose and you'll measure the wrong thing.

The Matrix in Action

Now combine the axes. Every combination is a potential product.

Domain-Specific Idea Generators

Here's the matrix applied to common business domains. Use these as starting points for your own exploration.

"Which impossible conversation would unlock the most value in your domain?"

Not "what AI tool should we build?"

But "who do we wish our team or customers could talk to—but can't?"

Answer that question. Walk through the four axes. Write the one-line product spec. You've just systematically generated a Time-Shifted Proxy application.

What This Is Fighting Against

The incumbent mental model keeps you from seeing the pattern. Time to name what's in the way.

If you've read this far and the examples still feel disconnected—talking to grandma in 1975, dating an AI, rehearsing conversations with your boss, watching future-you rave about a holiday—you're likely caught in a frame that makes these ideas invisible.

That frame is powerful. It's institutional. And it's wrong about what AI is for.

The Incumbent Mental Model

Here's what most people think AI is for:

• Productivity. Making tasks faster.
• Automation. Reducing headcount.
• Efficiency. Time saved per employee, cost reduction, workflow optimisation.

This isn't wrong. These are valid use cases. But they share the same frame: AI as a tool that makes existing work faster or cheaper.

How does this show up in practice? Listen to the questions:

All sensible. All measurable. All the same underlying assumption: AI replaces human effort on existing tasks.

Why does this frame dominate? Three reasons:

What Productivity Thinking Misses

When AI equals productivity, you only see automation opportunities. The design space shrinks to three questions:

• What's slow? Speed it up.
• What's expensive? Automate it.
• What's repetitive? Remove humans.

That's a narrow aperture. Here's what doesn't fit:

Yet every single one of these has proven ROI. Oracle's AI sales training: opportunities per rep doubled (2.78 → 6.02 per month). Video testimonials boost conversions by 80%. MIT's "Future You" tool reduced anxiety and improved long-term decision-making in 344 participants.

The productivity frame can't see these because it asks the wrong question.

The New Frame: AI for Relationships Across Time

The Time-Shifted Proxy pattern starts from a different assumption:

"AI isn't just for making tasks faster. AI is for relationships across time."

Not automation. Simulation.
Not efficiency. Connection.
Not tasks. Conversations.

Here's the category distinction:

Why does this matter? Because different frames generate different questions:

• Productivity: "How do we automate X?"
• Relationships: "Who do we wish we could talk to?"

The second question opens doors the first one can't see.

The Core Mechanism: Media Translation

Here's the deeper insight: the constraint was never technical. It was media format.

The archive already exists. It always has:

• Letters from grandma in the 1970s
• CRM notes from churned customers
• Research on what makes holidays memorable
• Recordings from top salespeople
• Family photos, genealogy records, old voice notes

The raw material was always there. What AI enables is translation.

Why does this work on the brain? Because conversation is our native interface for memory.

"Humans are lazy mammals with big feelings. We want the meaning, not the microfiche."

We suck at reading archives. We're great at talking to people. Forty pages of letters? That's homework. A six-minute phone call that compresses those letters into a voice that sounds like grandma? Instantly consumable.

AI doesn't create new information. It makes existing information consumable in the form our brains actually want. That's the unlock. Not AI intelligence, but AI accessibility.

What We're NOT Saying

Before we go further, three important caveats:

If This Frame Wins

What changes if Time-Shifted Proxy thinking becomes mainstream?

For Individuals

• AI becomes a tool for emotional connection, not just productivity
• Heritage becomes accessible—talk to ancestors, not just research them
• Decision-making improves through future-self connection
• Difficult conversations get rehearsed before stakes are real

For Teams

• Training, onboarding, and change management get a new design pattern
• "Talk to" becomes a training modality (simulated customers, stakeholders, executives)
• Knowledge transfer shifts—capture not just information but voice and perspective
• Empathy scales—experience what it's like to be an angry customer, not just read about it

For the Industry

• A new product category emerges: "time-shifted proxy" applications
• Memorial AI moves from fringe to mainstream
• Simulation becomes standard for high-stakes preparation
• Marketing shifts from "brand tells you" to "future-you tells you"

"The companies that see this pattern early will define new categories. The ones that don't will keep building marginally better chatbots."

The Recognition Test

How do you know if you've shifted frames? Here's the test:

Signs you're still in the productivity frame:

• Every AI idea is about "faster" or "cheaper"
• You dismiss emotional applications as gimmicky
• You can't see the connection between genealogy and sales training
• "That's interesting but not practical for business"

Signs you've adopted the relationship frame:

• You see impossible conversations everywhere
• You ask "what data traces exist?" about people
• You think about medium (text/voice/video) as a design choice
• The examples in this book feel connected, not random

If you've made the shift, the next question isn't "is this real?" It's "where do I apply it first?"

That's Chapter 8.

Applying the Pattern: Your Next Steps

You've seen the pattern. You understand the framework. You know why late 2025 is the inflection point. Now the question is: what conversation across time will you enable first?

This chapter gives you five practical steps to move from concept to implementation, domain-specific applications to inspire your thinking, and a closing that brings us back to where we started.

References & Sources

This ebook synthesizes research from industry analysts, academic institutions, technology publications, and practitioner experience. Sources are organized by category for easy reference. All statistics and quoted material are attributed to their original sources.