Where can we retain the human touch, impactfully, in the age of AI? Thomas Scherer, cloud architect & computer scientist working for Google joins Lisa. One Saturday night, Thomas sat down with Gemini and asked, "What will make me the happiest person in the world?" Over the course of the next few hours, he got some fascinating results. All of this is part of the story of AI in our lives today, but there is so much more. This conversation is a small reflection of where we are with AI and why we should embrace its benefits, learning as much as we can with careful curiosity. From Horses to Cars “What do I do with my horse-riding skills now that the car has been invented?” With this statement, Thomas reminds us that mega shifts in our human experience is historically normal, and a reflection of the human mind’s brilliance. The AI Shift is just another technological step change. AI is replacing ‘commodity tasks’ - those which are repetitive, standardised processes, providing us with more time to lean into creativity. We become the navigator whilst the more mundane jobs could be taken over by AI. A new way to Search Traditional search engines try to match words whereas modern AI systems match meaning. When you search for trousers for instance, AI systems can use images and semantic understanding to infer style, intent, and context rather than just scanning for the keyword ‘pants or trousers.’ Large language models (LLMs) such as Gemini, ChatGPT, Claude, Perplexity, and so on, predict the most likely next word, turning colossal amounts of data into fluent conversation, explanation, and even advice based solely on statistical probability of word patterns. We don’t even need to invent the perfect query as they can also predict this. AI as Your Collaborative Partner Used well, AI is more like a creative collaborator: a brainstorming partner that proposes alternative angles, structures, and prompts. For small businesses, it can become an extra “virtual team,” generating draft podcasts, social posts, or marketing visuals that can then be curated and refined. But all the while, it remains the human who sets the objectives and the required tone. This also lends itself to the possibility of many people becoming autonomous, single-person businesses. Agents: When AIs Start Working Together When you give an AI tools and sub-tasks, it can orchestrate them toward a goal. One agent might create images; another might check whether those images match the brief (e.g. 'sunny landscape, not rain’); together, they negotiate improvements until the output fits what you asked for. Even non-technical people can use early agent-like products. NotebookLM, for instance, lets you upload documents, then: - Ask questions about them in natural language. - Generate personalised podcasts from your own material that you can listen to during a commute. - Work across multiple languages, both in sources and in the audio you generate. A recurring complaint in companies is: “Our data is too messy to do AI.” That is partly true for training bespoke models: bad data in, bad model out, but paradoxically, AI is also very good at cleaning data in the first place. You can literally give such a tool a messy folder of information and ask to make sense of it. Because it understands patterns in addresses, email formats, names, and categories, AI can, for example: - Standardise your contact lists so mailings no longer bounce. - Extract fields from scanned paperwork and fill out forms for you. - Help you perform a “data spring clean” on everything from CRM records to home admin. For an individual drowning in paperwork, this is transformative: scan, upload, and ask the AI to pre-fill or summarise, then you simply review and sign. Everyday Simplifications with AI You do not need to be a computer scientist to get real value from AI. A good starting sequence for a normal day could include: - Identify what you hate doing: repetitive emails, calendar logistics, summarising long documents, or form-filling. - Ask the AI directly: “Show me how to use you to spend less time on this task,” then iterate based on its suggestions. - Start with non-sensitive data and low‑risk tasks, and only move to personal or client material once you understand the provider’s terms and privacy guarantees. People in Luxembourg working across languages can also benefit from live translation and dubbing: tools already exist that let you speak in German and be heard in French or English in your own voice, with a slight delay, in meetings or recorded content. Jobs, Risk, and the Human Edge AI is reshaping the job market. In the UK, one study found that companies using AI had eliminated 11% of previous roles and left another 12% unfilled, while creating 19% new roles, which is a net loss of 4% overall, with the UK faring worse than the US on the balance between jobs lost and created. That reality naturally fuels both excitement and anxiety. What AI targets first are commodity tasks: copy-pasting, routine classification, basic template writing, or standardised analysis. The more your work relies on unique human context, judgment, empathy, and rapport, from live concerts to therapy and even parenting, the harder it is to replace. The opportunity, and pressure, is to climb the value chain: stop being the engine that moves the data and become the navigator who decides where to go. Trust, Safety, and Owning Your Self Image and Voice As AI systems get better at imitating voices and faces, distinguishing fake from real becomes a societal survival skill. Voice scams already exploit cloned speech to convince parents their child is in danger, and manipulated images can travel faster than fact‑checks. Two layers of protection are emerging: - Technical safeguards such as watermarking in generated images or audio, which allow downstream tools to flag AI‑created content. - Legal and ethical frameworks like GDPR in Europe, which treat your appearance and voice as personal data requiring your consent for alteration and reuse. - Providers also increasingly commit to indemnifying users when material generated within the rules is later challenged on copyright grounds, shifting some of the risk back to the platforms that trained the models. Prompting: Talking to AI so It Really Helps You do not need to be a prompt engineer, but a few habits make a big difference. First, describe what you do want rather than only what you do not want: “Keep the face unchanged and brighten the background” works better than “Don’t change the face.” Second, you can use AI to improve your own prompts: - Tell it your goal (“I want a video that shows X for Y audience”). - Ask: “Write a detailed prompt I can paste into a video/image generator.” - Edit the suggested prompt so it fits your tone, context, and constraints. Over time, this becomes a self-teaching loop: the AI drafts the prompt, you tweak and observe the output, and your intuitive sense of what to ask for gets sharper. AI, Emotions, and the Limits of the Machine Some people now confide in chatbots as if they were friends or therapists. In one late-night experiment, Thomas asked Gemini to interview him and figure out what would make him “the happiest person in the world”; the system eventually pointed out contradictions in his answers and nudged him toward deeper reflection. That shows how AI can mirror back patterns in your own thinking and ask probing questions. But it still lacks the embodied empathy, nuanced perception, and ethical responsibility of a trained human therapist, who reads not just words but tone, pauses, posture, and history. AI can supplement support; it should not replace serious care. Why You Should Start Now Paradoxically, Thomas’s biggest fear is not that AI will take over, but that people will be left behind because they are too afraid to try it. Like refusing to learn to drive when everyone else has moved to cars, opting out of AI entirely risks shrinking your options just as the toolset explodes. The most practical stance is curious, critical use: test it, set boundaries, keep the human touch at the centre, and let the machines handle the drudgery.
