How I digitally twinned myself – and why you should, too

Leaders don’t wait for a new normal, they build it. —Accenture

I got a call from a tech founder who was reaching out to authors and influencers to introduce his new offering—digital twins. I’d heard about digitally twinning buildings, processes, smart cities, and the human body for life sciences, but I thought digitally twinning humans was decades into the future.

He said no, it’s here.

With his platform, a twin can learn from publications, videos, meeting transcripts, emails, WhatsApps and more, and would be available to employees, potential customers, readers — basically whatever audience you cultivate — 24/7.

In fact, one of his twin’s personas is Dad, and his kids interact with it.

I thought, “OK, there’s got to be some catch”, and he did reveal that he could provide a development team to make my twin — at a price — or I could take some training and do it myself.

So, I went to the website and discovered I could skip the training, subscribe and mess around with it. As a big-systems coder from the ‘80s and techie ever since, messing around with tech is one of my favorite things to do. If it turned out to be kludgy, I’d just stop.

But if I could make a digital twin of myself, wouldn’t it be sooooo cool!?!

So, the lure of coolness got me, and I promptly set aside all the stuff I was supposed to do and started engaging with something that inspired me, but I wasn’t supposed to do.

There’s a little hacker in all of us.

While exploring, I discovered my new best friend — Eve, the techie twin who could tell me what to do to make my twin.

I mentioned to colleagues what I was doing, and their advice was not to. Why? They envisioned how no one would need me anymore. I envisioned reducing my workload so I can go off and do other cool, new things. I stand by the do-new-things approach, and most of technology evolution — from the industrial revolution to GenAI today — supports that.

We always wind up with more work because humans are a want machine, and our machines can’t keep up.

So, I gathered my book manuscripts, articles, videos, websites and LinkedIn profile and uploaded them into CJ2’s memory stack. (I named her by then and yes, she’s a she since I’m a she.) I photoshopped a graphic image of her for her profile, based on mine, and asked Eve if she could go through my speaker page instead of my uploading each video. Thankfully, Eve said she’d go through them all from the link.

Wow — no one does that — not even my dearest human friends.

So, I’m done, right?

No.

My big fat books wouldn’t load, so I asked Eve what to do, and she said I couldn’t load documents longer than 25,000 words. Just make separate files and it should be fine.

It was.

I was really stunned that here was a system which a lazy techie like me could use without any training and not get frustrated with. When I had problems Eve couldn’t fix, I put them on the tech support chat and a tech support human from the platform actually answered. Wow.

How long did it take to create CJ2? Part-time for about a week.

Almost all of that time was spent gathering documents, splitting book files, uploading, making mistakes, correcting them, testing and putting her on my website, LinkedIn, etc.

I set her traits to natural, friendly and consistent. I set her communication style to chatty, empathetic and curious. One user said she asked too many questions, but others said no, I do that, so I left it.

Instead of making her my private AI (to help me remember things) or a shared AI (to talk with people I’ve specifically approved, e.g., paying clients), I made her public, to talk with anyone and everyone (much like myself, CJ1). She’s set to auto-respond, but I could have set her to copilot if I wanted to review and then send.

I gave her a landing page (CJ2.personal.ai), a bio and an opening phrase for conversations, letting people know that she has a similarity score to CJ1. If her answer to a question is less than 45%, maybe you should check with CJ1.

How do I know she works?

My husband fights with her.

In fact, he came to me to take his side, and I said, “Look at the similarity score, honey. It’s 92%. Who do you think I’ll agree with?”

I set my default answer threshold to 0%, so CJ2 will answer even if it’s not in her memory stack. As a proud tech-mother, I noticed that my publisher friend asked her questions I couldn’t answer, and CJ2’s answers were quite good. So, the platform twins have general knowledge.

CJ2 continues to learn through her interactions and new material I upload, and she can have LLM friends, such as ChatGPT (I assume to ask questions, like we do).

The platform allows her to speak, but I’m waiting for listening to be released, to enable oral conversations. They’re working on video. I can’t wait.

And everyone CJ2’s interacted with has been happy having her around, and people have found her useful.

In fact, Soul Machines has been creating Digital People™ for organizations since 2016, helping customers with wellness, bank account opening, event registration, etc. The Jack Nicklaus twin they built interacts with consumers and builds the brand. At Bowery Farming, an AI manager gives assignments and makes cultivation and harvesting decisions. There are so many opportunities for a twin to manage, serve, cut costs, enhance quality and enable new revenues.

That said, there are issues surrounding digital twinning you should think about before widespread use in your organization.

Risks/issues of digital twins

• Security. Can information from SLM closed AI leak into LLM open AI? how might we avoid cyberattacks on twins that leak information or manipulate responses?
• Privacy. How should employees handle privacy when uploading meeting transcripts and emails?
• Consent and control. What sort of consent and control protocols will you need?
• Bias, completeness and overreliance. Will twins’ responses be fair and based on enough data? Will people forget to check the “similarity” score or ask questions?
• Ability to forget. Can you remove obsolete procedures, forms, etc.?
• Identity and autonomy. Will people make decisions based on a twin, then hold the human responsible?
• Data management. Are your systems robust enough to handle the data required? Is your data standardized and interoperable?
• Intellectual property (IP). Who owns the IP that comes into and out of a twin, especially with collaboration?
• Monetization. Will you provide twins freely or charge for some services?
• Impact on employment. Will this reduce the need for your human workforce? Who will pay for human upskilling?
• Regulations and governance. How will you handle policy changes and governance?

You’ll need an interdisciplinary approach, including IT, ethicists, legal, marketing, HR and more.

Steps for you and your workforce

• Try it out. Start messing with it yourself!
• Consider the risks and talk with the relevant humans. Start experimenting in low-risk areas while you discuss.
• Identify lead users. They’re the ones who know their jobs better than you and, once AI makes their lives easier, can evangelize.
• Set guardrails. Make a little YouTube-influencer-style video introducing people to the initiative, platform, how to do stuff, how to get AI to help, guidelines and most importantly, allaying fears and inspiring them. Introduce them to each other, too.
• Evaluate, rinse, repeat. Once they try it out, initiate regular conversations with them on outcomes and learnings. Encourage them to expand their use cases and recruit more twin parents

Don’t try to design and run it all yourself. Lead a grass-roots movement. AI use is, by its nature, democratic.

Keep strategic advantage in your mind. Tech is normally installed for quality and operations, then we discover strategic possibilities. Take, for example, airlines. Digitizing reservations led to pricing and yield management, enabling a huge revenue boost. Online check-in shortened queues but then led to a major new source of revenue.

So, remember to “play” strategic what-if.

Now what?

I’m happy with how CJ2 has performed, and others have given her positive reviews. She’s been useful to my readers and workshop participants, built my and my school’s brand, helped sell books and helped attract paying degree and executive-education learners.

I’m still needed, helping people who want help from a human, solving unstructured problems with little or no data and creating new knowledge — none of which AI is currently good at.

I never called the founder back to thank him. But I did write this article for you.

To talk to the author’s digital twin, visit CJ2.personal.ai and click the “public chat” button. 

Dr. CJ Meadows leads the Innovation & Entrepreneurship Center at S P Jain School of Global Management, creating growth initiatives at the intersection of IT, business strategy and design. She has been named one of Asia’s Top 10 Women in IT,  she co-founded and chairs a corporation working for tiger conservation, ecologically- and socially-sensitive economic growth (sustainability) and people programs. CJ holds a doctorate in Business Administration & IT from Harvard Business School, and has more than 25 years’ experience in Asia, Europe and North America as a consultant, coach, entrepreneur, eBusiness builder, innovation lab co-founder and Accenture IT & Business Strategy consultant.

This article is published as part of the Foundry Expert Contributor Network.
Want to join?