Glossary
The terms behind the thinking.
48 short definitions of the ideas I write, advise and teach on, from physical AI to GCC 4.0.
Physical AI and robotics
- Actuator
- The component that turns energy into motion, such as a motor, gearbox or hydraulic unit. In humanoids, actuators are among the largest cost items and set how strong, precise and efficient the robot can be.
- Cobot
- A collaborative robot designed to work safely alongside people, usually smaller and slower than traditional industrial robots, and easier to redeploy.
- Dexterous manipulation
- A robot's ability to grasp and handle varied objects with hand-like precision. It is one of the hardest problems in robotics and a key bottleneck for useful humanoids.
- Embodied AI
- AI that learns through a body interacting with its surroundings, rather than from text and images alone. A robot that grasps, walks or drives is embodied; a chatbot is not.
- Humanoid robot
- A general-purpose robot with a roughly human form, designed to work in spaces built for people: stairs, doors, tools and workstations. Its promise is flexibility; its challenge is cost, safety and reliability.
- Humanoid tech stack
- The layers that make a humanoid robot work: actuators and hands, sensors, batteries and power, onboard compute, low-level motion control, a foundation model for perception and action, and the data and fleet software that improve it over time.
- Moravec's paradox
- The observation that what is easy for humans, such as seeing, walking and grasping, is hard for machines, while what feels hard for humans, such as calculation, is easy. It explains why AI automates desk work faster than physical trades.The plumber, the developer and the dignity of labour →
- Physical AI
- AI that senses, decides and acts in the physical world through machines such as cars, robots and factory equipment. Its output is motion rather than text, so safety, latency, energy and hardware cost shape every design and business decision.
- Robotics foundation model
- A large model trained on diverse robot and video data so it can generalise across tasks and robot bodies, much as language models generalise across topics.
- Sim-to-real
- Training robots in simulation and transferring the skills to physical machines. Simulation is cheap and safe; the hard part is closing the gap between the simulated and the real world.
- Teleoperation
- A person controlling a robot remotely. Beyond operating in hazardous places, it is now a key way to collect demonstration data that teaches robots new skills.
- Vision-language-action (VLA) model
- An AI model that takes in camera images and instructions in plain language, and outputs robot actions. VLAs let one model handle many tasks instead of programming each one by hand.
Mobility and software-defined vehicles
- ADAS
- Advanced driver-assistance systems, such as adaptive cruise control, lane keeping and automatic emergency braking. They are the stepping stones between manual driving and automated driving.
- Central and zonal architecture
- An electronic layout in which a few high-performance computers run the vehicle's functions while zone controllers manage everything in one area of the car. It cuts wiring, weight and cost, and makes over-the-air updates practical.From spaltmaß to software →
- Connected tyre
- A tyre fitted with sensors that measure pressure, temperature, load, wear and road contact. The tyre is the only part of a vehicle that touches the ground, which makes it a valuable source of data.When the rubber hits the road →
- Coopetition
- Competitors cooperating where it suits both, such as on shared platforms or standards, while still competing elsewhere. The software-defined vehicle is likely to be won this way, not by one player taking all.
- Electronic control unit (ECU)
- A small computer that controls one function in a vehicle, such as the engine, brakes or windows. Premium cars have carried around a hundred of them, each with its own software and supplier.
- Over-the-air (OTA) update
- A software update delivered wirelessly to a vehicle or device, fixing problems or adding features without a workshop visit. It turns a car into a product that can improve after sale.
- Software-defined vehicle (SDV)
- A vehicle whose features are defined and updated mainly through software running on a few central computers, rather than fixed at the factory in dozens of separate control units.The car is becoming a supercomputer on wheels →
- Spaltmaß
- German for the panel gap between parts of a car body, long a symbol of build quality. It stands for the hardware-first mindset that carmakers now have to balance with software excellence.From spaltmaß to software →
AI and business models
- AI agent
- AI that pursues a goal by planning and taking actions across tools and websites, not only answering questions. As agents spread, a growing share of internet traffic may come from machines rather than people.The internet's next users won't be human →
- AI diffusion
- How widely and deeply AI is put to use across an economy, by citizens, companies and governments. For national advantage, diffusion matters more than who builds the best model.Diffusion matters more than models →
- AI wrapper
- A product built on someone else's AI model, adding an interface, workflow or data. Wrappers can be valuable, but their margins and defensibility depend on what they own beyond the model.Compute is the new railway →
- AIoT
- The combination of AI with the Internet of Things: connected devices whose data is analysed by AI to predict, automate or create new services.
- Default layer
- The interface people use without actively choosing it, such as the preinstalled assistant, browser or app. In AI, owning the default may decide winners more than benchmark scores.In AI, defaults win →
- Ecosystem moat
- An advantage built from owning many connected layers, such as chips, models, products, distribution and data, so each strengthens the others. It is harder to copy than a better model.In AI, defaults win →
- Ice-cream and studying products
- Ice-cream products reward users in the same moment they ask for effort, so they spread on their own. Studying products ask for effort now and pay off later, so they must fight for every user.Studying isn't like eating an ice cream →
- Sovereign AI
- A nation's capacity to build and run AI on its own data, compute and infrastructure, rather than depending entirely on foreign providers.Compute is the new railway →
- Voice-first AI
- AI designed to be spoken to rather than typed to, working across languages, accents and noisy settings. For the next billion users, voice may be the main way to reach AI.The next billion AI users will speak →
Strategy and organisation
- Accountability arbitrage
- The idea that a capability centre's next advantage is not lower cost or scarce skills, but its willingness to own results. Cost arbitrage got centres in the room; accountability decides who leads.GCC 4.0: from cost centre to command centre →
- Bricoleur
- Someone who builds with what is at hand rather than waiting for perfect tools, an idea from the anthropologist Claude Lévi-Strauss.Ajith, the bricoleur →
- Deep tech
- Ventures built on scientific or engineering breakthroughs, such as robotics, semiconductors or new materials. They carry technical risk as well as market risk, and take longer to reach revenue.Institutions produce conviction →
- GCC 4.0
- The next stage of global capability centres, in which a centre moves beyond cost and capability to owning outcomes: end-to-end platforms, products and decisions measured on the parent company's results.GCC 4.0: from cost centre to command centre →
- Global capability centre (GCC)
- An offshore centre owned by a multinational to deliver work for the parent company, from engineering and IT to finance and research. India hosts more than any other country.
- Industry 5.0
- A European vision of industry that builds on the automation of Industry 4.0 by putting people, sustainability and resilience at the centre.
- McKinsey 7S model
- A framework that looks at seven parts of an organisation that must fit together for strategy to work: strategy, structure, systems, shared values, style, staff and skills.Used in my strategy execution programme →
- Ownership ladder
- A way to judge a capability centre by what it owns rather than how many people it employs: tasks, then modules, platforms and finally global mandates. Each rung is earned through trust built on the one below.A case study on Consulting →
- Patient capital
- Investment that accepts long timelines and delayed returns in exchange for backing hard problems. Deep tech depends on it.Institutions produce conviction →
- Three-box strategy
- A framework by Vijay Govindarajan for balancing time horizons: manage the present, selectively forget the past, and create the future. It helps leaders fund innovation without starving the core business.Used in my strategy execution programme →
- Where to play and how to win
- The two choices at the heart of any strategy: which markets, customers and parts of the value chain to compete in, and the advantage that lets the company win there.How I apply it in consulting →
Decisions, skills and leadership
- AI fluency
- The ability to use AI well at work: knowing what it can and cannot do, framing tasks clearly, checking its output and using it responsibly. It is becoming a core skill for every role.Built in my AI @ Work programme →
- Cynefin framework
- A sense-making framework by Dave Snowden that sorts situations into clear, complicated, complex and chaotic, each needing a different way of deciding.Used in my decision-making programme →
- Issue tree
- A diagram that breaks a big question into smaller, MECE sub-questions until each can be answered with analysis. It keeps teams focused on what actually decides the answer.Taught in my problem-solving programme →
- Learning in the flow of work
- Learning that happens inside daily work, in small doses and at the moment of need, rather than only in classrooms. AI makes it practical at scale.AI for L&D on Training →
- MECE
- Mutually exclusive, collectively exhaustive: a rule for breaking a problem into parts that do not overlap and together cover everything. It is the backbone of structured problem-solving.Taught in my problem-solving programme →
- Pyramid principle
- A way of structuring a message, developed by Barbara Minto, that starts with the answer and then groups the supporting arguments beneath it. Busy readers get the point first.Taught in my problem-solving programme →
- Skill Twin
- Skillstr's digital twin of a professional's skills: a mirror of how a person leads, used to personalise practice and feedback over time.
- VUCA and BANI
- Two ways to describe a turbulent world. VUCA stands for volatile, uncertain, complex and ambiguous; BANI for brittle, anxious, non-linear and incomprehensible.