The Sovereign Hub Podcast

Human Led, AI Forward.

The Sovereign Hub offers conversations and resources to help you decide where AI belongs in your life and work, where it is risky, and where it is not needed at all. Hosted by Carolyn Shepherd, champion of human judgement in the age of AI, creator of Jobscaping and one of HR Magazine's Most Influential Thinkers 2025. carolynshepherd.substack.com

  1. 4d ago

    The Truth about Applying for HR Jobs in 2026

    What happens when someone who has spent their career putting people at the centre of change ends up on the receiving end of it? That’s the question I put to Janet O’Hehir this week. Janet is an organisational development consultant and social scientist whose career runs from emergency services to magic circle law firms, in the UK and her native Australia. Janet spoke very openly about job hunting through applicant tracking systems. Applications vanish without a word, ghosting has become normal, and the one employer that promised a human would read her CV stood out simply for doing so. I’ll admit her story made me uncomfortable. I’ve been on that side of the process myself, and I’ve also been the one relying on those systems to sift candidates. So we looked at what good would look like. That means starting with the real problem, listening to unsuccessful candidates and recognising that your employee value proposition begins at the very first touch point. Thanks for reading The Sovereign Hub Podcast! Subscribe for free to receive new posts and support my work. From there we turned to judgement. We talked about why a human in the loop needs the training and authority to push back, and what the ICO found when 30 organisations volunteered to have their recruitment pipelines reviewed. Janet describes her work as living in “the white space between the lines on an org chart”, where people and systems meet. We also got into the pressure on the messy middle of management, the lack of investment in leaders and why every organisation should set a minimum standard of AI skills so that everyone has the same footing. It’s an honest, human conversation, and I think you’ll recognise a lot of it. I hope you enjoy it as much as I did. Thanks for listening. Stay Sovereign, This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

  2. Sep 23

    What If I'm Missing Something Important?

    Shownotes What If I'm Missing Something Important? (The Sovereign Hub, Season 2, Episode 2) You've spent years being the person people trust to get it right. Then AI landed on your desk, and one question keeps stopping you from signing anything off: what if I'm missing something important? What if it goes wrong, and everyone wonders why you didn't see it coming? In this episode, Carolyn Shepherd gives the honest answer (yes, you probably are missing something, and so is everyone else) and explains why the question is really about your professional reputation, not your knowledge. She then shares four plain tests for knowing when you know enough to make the call, none of which need any technical expertise. You'll hear: Why "what if I'm missing something?" feels so personal for HR leadersWhy reading one more report doesn't make the worry go awayFour simple tests: can you explain the options, is more reading changing your mind, can you undo it, and would you only find out by tryingWhat changes once you stop waiting to feel ready, and how you get back to the people work A question to think about: what are you still researching that, if you're honest, you already know enough to decide? Join Carolyn live. On the last Wednesday of every month at midday, she runs a free session for HR leaders who've been handed AI to own: how to get your leadership team deciding with you, how to get to a plan your CEO will sign, and how to get back to the people work. Next session: Wednesday 30 September 2026. Register here: https://us06web.zoom.us/meeting/register/yH74tlpYQn-CI_E8fp0jWA The Sovereign Hub is hosted by Carolyn Shepherd, barrister, author and creator of Jobscaping®. Human led, AI forward. AI can be delegated, but leadership judgement cannot. Stay sovereign. About Carolyn Shepherd Carolyn is a barrister, author and keynote speaker who works with people leaders and leadership teams in SMEs who've been handed responsibility for AI. She's the founder of Emmeline AI, creator of Jobscaping®, author of The ChatGPT Advantage and a LinkedIn Learning instructor to over 40,000 learners. She was named on the Digital Leaders AI 100 UK list and as one of HR Magazine's Most Influential Thinkers. Connect with Carolyn on LinkedIn: https://www.linkedin.com/in/carolynjshepherd/ AI can be delegated, but leadership judgement cannot. Stay Sovereign. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

    What If I'm Missing Something Important?
  3. Sep 16

    Could AI really kill us all by the end of the decade?

    Hi, I had a different episode ready for you this week. Then a researcher walked out of Anthropic saying the people building AI believe it could kill us all by the end of the decade, four chief executives said the industry should slow down, the President of the United States said it shouldn’t, and Sheryl Crow told Instagram to pay attention. So that episode has moved to next week, because this one needed answering first. The question I’ve been asked most since, in the corridor, the car park and by email: * How could AI actually kill us all? * Should I be worried? * And what am I supposed to do about it from my desk? In this episode This episode is my answer, and it’s the middle way. Not doomer, not accelerationist. Using our judgement before handing things to AI. * What actually happened, in three minutes, without the hype. * The three facts that make the worried, worried: AI learns as a crowd while we learn in silos, it improves itself, and it doesn’t follow rules slavishly, so tell it not to use the door and it may use the window. * Why the labs can’t just add another rule, and why it looks a lot like a prescription drug cascade. * The evidence on the other side, which is stronger than the headlines let on. * Who should be in the room before you say anything to your staff, and whether gut feel is worth anything when the experts admit their numbers are a wild guess. * Six things you can do this month, none of which needs you to be an AI expert and none of which you do alone. Thanks for reading this far, The Sovereign Hub Podcast! Subscribe for free to receive new posts and support my work. Download the companion 30 minutes, no jargon. In the show notes there’s a free companion download, called “Quite a lot, actually”: the six actions you can take - with action points and getting started tips for each one. Download the guide to this episode here… Are you going in circles with your AI plans? And if you’re the HR leader who’s been handed AI, I’m running a free 45-minute webinar on Wednesday 30 September at midday, “How HR leaders get to an AI plan the CEO will sign” and I’m delighted to say that it’s only been announced today and it’s already filling up. Link below. Here’s a question for you to think about:When the AI headline reaches your organisation, who gets asked first about it, and is it the right person or just the nearest one? Register for the webinar here… Registration link [How HR leaders get to an AI plan the CEO will sign, Wednesday 30 September, 12:00] Stay Sovereign, This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

  4. Sep 9

    I don't lie awake worrying about AI governance.

    Hi, Six months into hosting the show, I noticed something changing. People weren’t bringing me career questions any more. They were bringing me decision questions. Have we thought this through? Are we moving too fast? Who carries the can if it goes wrong? So the show has evolved with them. The Sovereign Career Hub is now The Sovereign Hub. Guest interviews stay, and you’ll hear more solo episodes from me on the conversations that never quite make it into the SLT meeting. The first one is out today, and it’s about the noise in your head when a decision is yours to own. * Why “nobody ever got fired for buying IBM” is a familiar trap. * Why case studies are reassurance dressed up as evidence and * Why you don’t need to become an AI expert to make a good AI decision, but you do need to get comfortable exercising judgement when the experts disagree and; * why the sceptic on your leadership team might be the most helpful person in the room (sometimes). Twenty minutes, no jargon. And in the show notes there’s a free companion guide: the techniques from the episode on a digital download, plus The Decision Noise Collaboration, a twenty-five minute structure for working through a difficult AI decision with a colleague or two. No AI expertise required, just the willingness to talk about the decision noise. Here’s a question for you to think about: What AI decision are you carrying right now, and who else needs to be in the room before you make the call? Download the guide to this episode here… Stay Sovereign! Carolyn This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

    I don't lie awake worrying about AI governance.
  5. Sep 2

    Who Cares if the Boss is Watching?

    Ep 13: Who Cares if the Boss is Watching? First, Some Important Context In 2006 Gaelle Watson was given the task of running virtual classroom sessions. The internet was different, the infrastructure was different, and so was the mindset: she had to fight her corner to show that the same learning happens behind a screen, and that people can be just as active there. What that work came down to was trust and engagement. Twenty years on she sees a similar situation with AI. People are not always engaging with it because they do not trust it. Not just the output. They are unsure how they will be seen using it, and what it means for their work if they are seen as replaceable. Thanks for reading! Subscribe for free to receive new posts and support my work. The Questions We Get Into How do you build trust with a team you rarely sit next to, and does AI make that harder or easier? If transparency means being told the truth about your future, can you take it? Could your organisation say it? Can you be your true self at work when so much of how you come across now passes through an AI filter first? And if AI can do the manager’s job, do we still need managers? Or do humans essentially need other humans? Growth or Surveillance? The tools most of us use every day can already track how long you spend in an application and the words you use with colleagues and clients. For a learning specialist that is gold: you can revisit a conversation, get personalised feedback and do it better next time. For an employee it looks like a permanent performance review. Growth and surveillance, Gaelle says, are the same coin. An employee is very unlikely to trust that data which can be accessed will not be. So ring-fence it. A separate, anonymised space for practice that cannot be accessed for performance purposes. If your practice was shocking, fine. You will learn from it, and it will not cost you your career. Thanks for reading! This podcast is public so if you like please spread the word. Unmuting Emotions Gaelle is Director of SyncSkills Performance, a facilitator working on communication, collaboration and leadership in hybrid and remote teams, and host of the Learning Horizons podcast. She invited me onto her AI in the Workplace series last year, and I recommend it, (and not only because I was on it). She has a law degree and a business degree, no technical background, and describes herself as a Mediterranean woman who waves her hands when she cares. She has been working remotely for over twelve years. When I asked her where she sits on the AI spectrum she gave herself a six, with a caveat that being French, a twenty out of twenty is unheard of. She is comfortable with the strategy and the tools, curious about what is under the bonnet, and clear-eyed about the risks, ethical and environmental. Powerful, she says, but we have to be intentional about how we use it. The Sandwiched Manager I put to her a figure from Bertie Tonks’ LinkedIn post on the quiet collapse of the middle manager: around 6% fewer managers than four years ago. When people go, the work does not, and managers are already sandwiched between leadership and their teams, expected to be coach, systems specialist, ethics lead and health and safety officer at once. Equipping them with AI is essential. But learning a tool takes headspace, permission to make mistakes and time to start with the easy stuff, and that space is very seldom available to managers. Who Carries the Can? We also get into accountability, and whether you can hold a person answerable for AI’s mistakes when they are not in control of the output. Two lawyers on a call, so it went somewhere I did not expect: we are accountable because we care, and AI is not driven by that. And we talk about the double dash. If the first thing your reader sees is one, they are not thinking “keen attention to detail”. They are thinking you did not care enough to write it yourself. The irony being that they have probably used it too. Her line of the episode: ‘If your practice was shocking, it’s fine. You will learn from it, and it will not be at the detriment of your career.’ Where in your organisation could someone practise badly, safely? And if that place does not exist, who decides what the data gets used for? Lots to think about in this one. Listen in and stay sovereign. Show Notes About Gaelle Gaelle Watson is Director of SyncSkills Performance. She works with individuals, teams and organisations on trust, psychological safety and behaviour change, with a particular focus on hybrid, remote and international teams. Her approach is practical and human-centred: she diagnoses behaviours, builds them through practice and feedback, and measures change over time so the impact is visible. She offers leadership facilitation, executive and leadership coaching, and learning and behaviour change consultancy, and works bilingually in English and French with organisations and learning consultancies across the UK and Europe. Connect with Gaelle on LinkedIn: Gaelle Watson Learning Horizons podcast (Gaelle's bimonthly show; Carolyn guested on the AI in the Workplace series): [link to add] Bertie Tonks, THE QUIET COLLAPSE OF THE MIDDLE MANAGER (and why most boards haven't noticed yet), LinkedIn. Source of the 6% figure quoted in the episode. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

  6. Aug 26

    Leaning into her Personality

    Somewhere inside a council, a woman whose job was to assess support applications for children with additional needs was given an instruction. ‘Turn down 98 out of every 100.’Jess Rad heard that first-hand, from an advocate she paid to help her fight for her own daughter. It took Jess three attempts to secure an EHCP, the legal document that guarantees a child support at school. She did it alone, papers covering the floor, a process she compares to writing a dissertation on a topic nobody would choose. She calls the system a sieve: it filters for whoever has the tenacity, the time, the energy, the money, English as a first language. The privilege, in other words, to keep going.I went into this conversation with a theory. I have watched AI arrive on both sides of that fight, systems processing applications on one side, exhausted parents using AI to write better appeals on the other, and I asked Jess whether AI is getting between families and fairness.She stopped me in my tracks. It is not AI turning them down, she said. It is people. These statistics existed long before AI. The systems are underfunded, and they were built by people with no lived experience of what they administer. Her point, borrowing from Ruth Bader Ginsburg, is that you need diverse experience in all the places decisions are being made. If parent carers had built these systems, they would not work the way they do, she says.Jess is the founder of The WomenHood, a gender equity and neuroinclusion business she has run for seven years, helping organisations close gaps and boost allyship. She is a national award-winning keynote speaker with a 100% NPS score, and her strapline says it all: Outspoken about the unspoken. Before that she spent 12 years in film and TV with Disney, the BBC and 20th Century Fox, blowing up caravans with Richard Hammond along the way, until motherhood met a workplace with no flexibility and she built something better. When I asked her where she sits on the AI spectrum, she surprised me: an eight out of ten. Her reasoning is one of the most compelling cases for AI I have heard on this show. A neurodivergent brain has what she calls a spiky profile, effortless brilliance at some things, real difficulty with others. Anything that lifts the weight of the difficult parts creates room for the brilliance. In her words: I’m game for it. I’m here for it. We also get into the one percent change that shifted how men became allies at Depop. Why these conversations must stop being opt-in, and what it took to be diagnosed autistic and ADHD as an adult, in a world where the diagnostic criteria were built around white males. Her line of the episode: ‘I don’t want my daughter to grow up feeling any shame for who she is. I just refuse that.’ Where in your organisation are decisions being made by people, or systems, with no lived experience of what they decide? And who is carrying the load in the meantime? The companion article will be published in the next few days. Listen in and stay sovereign. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

  7. Aug 20

    From 5,000 LinkedIn Views to 300,000 overnight

    First, Some Important Context It was a Friday lunchtime. Not a great time to post on LinkedIn, as anyone who watches the algorithm will tell you.Caroline Daniel posted anyway.By the evening she was watching the numbers climb past anything she had seen before. Her usual 5,000 views became 300,000. Three thousand people downloaded the free tool she was offering. And the shape of her business changed dramatically.The tool was an ERA 2025 matrix, a practical compliance tracker for the new Employment Rights Act, mapping every significant legislative change by date, category and priority so employers know exactly what to do and when. She built it with AI. Without AI, she says, she could never have created it at all. Clear Leads Caroline is founder of Clear HR Consulting and co-founder of Clear People, with over 25 years in HR and 14 of them running her own consultancy. Her clients have always been SMEs, startups and scale-ups. That is changing. Off the back of one post, the enquiries now coming in are from businesses of 250 to 300 employees looking for a fully outsourced HR function. She has partnered with Emily Perry to launch Clear People, and they are hiring. Who was the ‘target customer’? Here is the part I find most intriguing. Caroline posted the matrix as lead generation aimed at business owners. It spoke far louder to HR professionals. Her explanation is one every HR person will recognise: so much of the work is reactive, and the Employment Rights Act demands the opposite. When the removal of the two-year qualifying period for unfair dismissal lands, recruitment and onboarding stop being routine and become risk management. Seeing the whole act laid out in red, amber and green made people realise just how much proactive work is waiting for them. Checking the AI’s work But this is not a story about AI doing the work for you. Caroline built a handbook agent that updated all her client handbooks for the ERA in ten minutes. It also confidently got the day one unfair dismissal position wrong. Had she sent those handbooks out unchecked, in her words, it would have been ‘a bloodbath’. The human check is not optional.Her line of the episode: Claude can still write the policy, but Claude can’t have the difficult conversation.We also get into gut instinct in hiring, accountability when a chatbot goes off grid, and why a paddle club does not need AI to thrive, but does need its manager freed up to engage with people rather than monitor holiday requests.**What could you build this week that solves a real problem for the people you serve? And would you know how to check it before you shipped it?**This week’s resource is a copy of Caroline’s ERA matrix, free to download here: I’m figuring out how the hub can be most useful to you so please let me know if you want more or less of anything. In the meantime, I’ve got a few super interviews already recorded so listen in and enjoy. Take care for now. This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit carolynshepherd.substack.com

About

The Sovereign Hub offers conversations and resources to help you decide where AI belongs in your life and work, where it is risky, and where it is not needed at all. Hosted by Carolyn Shepherd, champion of human judgement in the age of AI, creator of Jobscaping and one of HR Magazine's Most Influential Thinkers 2025. carolynshepherd.substack.com