The competency map

to write down, describe & document

Turning what happened into a written record is close to fully automatable. This is core LLM territory.

It is tasks that get automated, not jobs →

to tell or give information to others

Delivering information on request is what assistants do best. The act of informing is largely automatable.

It is tasks that get automated, not jobs →

through text, absorb knowledge, facts & thoughts

Ingesting and absorbing text at scale is exactly what models do. Reading-to-extract is highly automatable.

It is tasks that get automated, not jobs →

to create order & order

Ordering, scheduling and arranging is classic automation territory, now easier than ever.

It is tasks that get automated, not jobs →

to calculate & create

Calculation is the oldest thing we automated. Fully machine territory.

It is tasks that get automated, not jobs →

to collect data, things or knowledge

Gathering data and information is something machines do tirelessly and at scale. Highly automatable.

It is tasks that get automated, not jobs →

to boil things down without losing the important thing

Summarization is the textbook strength of large language models. As automatable as it gets.

It is tasks that get automated, not jobs →

to recognize patterns

Recognizing patterns in data is the literal definition of machine learning. This is machine ground.

It is tasks that get automated, not jobs →

to sort, cure or select from

Classifying and selecting from a pile is core automation. Machine territory.

It is tasks that get automated, not jobs →

to create & tell stories

Generative models produce competent stories on demand. Distinct voice, lived detail and knowing what is worth telling is where the value concentrates.

The generative frontier is where the pressure is highest now →

to create, layout & design

Layout, variation and first drafts are increasingly machine work. Framing the problem and judging what is good still need a designer.

The generative frontier is where the pressure is highest now →

to map & mindmap

Structuring a domain into a map is something models do quickly and well. The framing choices behind the map matter more than ever.

The generative frontier is where the pressure is highest now →

to make things clear & simple

Making the abstract clear and simple is a signature LLM strength. The judgment of what to keep is the human layer.

The generative frontier is where the pressure is highest now →

to dive deep or to do research

AI research agents are advancing fast at gathering and synthesizing sources. Verifying and judging the sources is the human guard.

The generative frontier is where the pressure is highest now →

to draw or illustrate

Image generation turned illustration into a prompt. Direction, taste and a recognizable hand are where artists keep value.

The generative frontier is where the pressure is highest now →

to write down thoughts & writing texts

Writing is the most exposed knowledge task of all. Volume is automated; voice, accuracy and judgment are the differentiators.

The generative frontier is where the pressure is highest now →

to create structure in everything from texts to projects

Imposing structure on text and projects is something AI does well and quickly. Deciding the right structure is the human call.

The generative frontier is where the pressure is highest now →

to draw, visualize & create images

Generating images and visuals is now a prompt away. Art direction and meaning are where the human value sits.

The generative frontier is where the pressure is highest now →

to investigate, understand & see connections

AI does the legwork of finding connections in data; deciding which connections matter, and standing behind the read, stays with you.

Most real AI use augments, it does not replace →

to present arguments for or against something

A model can draft both sides of a case in seconds. Reading the room and choosing which argument to actually make is the human part.

Most real AI use augments, it does not replace →

to protect things, people or ideas

AI strengthens detection and monitoring, but someone accountable has to decide what counts as a threat and what to do about it.

Most real AI use augments, it does not replace →

to be philosophical

A model can rehearse every position ever written. Sitting with a question that has no answer, and meaning it, is human.

Most real AI use augments, it does not replace →

to solve things that do not work so well, or are broken

AI is strong at diagnosis. The hands-on, in-the-world repair, and the call on what to fix first, stay yours.

Most real AI use augments, it does not replace →

to make things better

Models suggest endless refinements. Knowing which improvement actually matters to the people involved is judgment.

Most real AI use augments, it does not replace →

to learn how things work & are there

AI is the fastest explainer ever built, which makes learning quicker, not optional. You still have to do the understanding.

Most real AI use augments, it does not replace →

to understand problems, risks & challenges

A model maps a problem space fast. Sensing which problem is the real one, in a messy context, is human framing.

Most real AI use augments, it does not replace →

to create hype around something, from a party to a product

AI floods every channel with content; cutting through the noise with a genuine cultural spark is the scarce part.

Most real AI use augments, it does not replace →

to invent & develop ideas

Models generate options endlessly. Taste, the judgment of which idea is worth pursuing, is what is scarce.

Most real AI use augments, it does not replace →

to talk, tell & discuss

AI drafts and translates the message; reading the person in front of you and adjusting is still human.

Most real AI use augments, it does not replace →

to like, find fault or give grades

Models give instant feedback and grades. Whose standard applies, and the responsibility for the verdict, stays human.

Most real AI use augments, it does not replace →

to find solutions to problems

AI proposes solutions quickly; defining the real problem and judging the fix in context is the human half.

Most real AI use augments, it does not replace →

to see, study or observe something that is happening

AI monitoring and sensing is powerful, but deciding what is worth noticing, and acting on it, needs a person.

Most real AI use augments, it does not replace →

to see the whole & the context

Models synthesize a lot into a summary view; holding the whole picture and its stakes is human judgment.

Most real AI use augments, it does not replace →

to be strategic & win over someone or something

AI can lay out the strategic options; choosing the play, and being answerable for it, is yours.

Most real AI use augments, it does not replace →

to get people to agree on something

Persuasive copy is cheap now. Persuading actual people, who can tell when you mean it, is still relational.

Social skills are getting more valuable, not less →

to plan & structure

AI produces plausible plans instantly. Adapting the plan as reality breaks it is the part that matters.

Most real AI use augments, it does not replace →

to show & present

Models build the slides; standing up and reading the room while you deliver is human.

Most real AI use augments, it does not replace →

to sell things, from apartments to ideas

Transactional selling is being automated; complex, trust-based selling stays human. AI handles the volume, you handle the relationship.

Most real AI use augments, it does not replace →

to be funny (or at least try)

Models can be funny on paper. Timing, read and the in-person laugh are still a human gift.

Social skills are getting more valuable, not less →

to be creative, do & create

Raw generation is cheap; creative direction, the why and the for-whom, is what holds value.

Most real AI use augments, it does not replace →

to finish & complete

AI gets you to a draft fast; the last mile of finishing, with ownership, is still human follow-through.

New work appears where old tasks disappear →

to think about which paths things could take

Models sketch scenarios on demand. Choosing which future to bet on is a human judgment with consequences.

Most real AI use augments, it does not replace →

to think about things based on books & knowledge

Models recombine existing theory fluently. Forming a genuinely new account of how things work is human.

Most real AI use augments, it does not replace →

to create solutions or come up with new things

AI accelerates the search for solutions; the leap to something genuinely new, and the courage to back it, is human.

New work appears where old tasks disappear →

to understand what worked well or badly

Models assess against criteria fast. Setting the criteria, and owning the judgment, is the human part.

Most real AI use augments, it does not replace →

to try, try & take risks

AI designs and runs tests at scale; choosing what to risk, and reading the result that matters, is human.

Most real AI use augments, it does not replace →

to weld people & groups together

Trust between people is not a deliverable a model can ship. Belonging is built in presence, over time.

Social skills are getting more valuable, not less →

to build companies, projects or associations

Founding and running things turns on judgment under uncertainty and people who will follow you. Tools help; accountability does not transfer.

Social skills are getting more valuable, not less →

to charm & make people like you

Charm lives in real-time, embodied connection. It is among the least exposed of all the verbs.

Social skills are getting more valuable, not less →

to help others succeed or find their way

AI tutors are useful, but being coached by someone who knows you and is invested in you is a relationship, not an output.

Social skills are getting more valuable, not less →

to drive tasks, projects & ideas forward

Initiative, persistence and owning the outcome are not things you can delegate to a model.

New work appears where old tasks disappear →

to get others to be motivated & engaged

Getting people to genuinely care depends on credibility and relationship. A generated pep talk reads as exactly that.

Social skills are getting more valuable, not less →

to find a common solution & agreeing

Real negotiation reads pauses, leverage and trust in the room. It is deeply relational and barely exposed.

Social skills are getting more valuable, not less →

to make things practical

Physical, in-the-world doing is where robotics still lags badly. Hands-on work is among the safest ground.

It is tasks that get automated, not jobs →

to find a creative or intuitive solution in the moment

Finding the move in the moment, with a body in a room, is the opposite of a batch process.

Social skills are getting more valuable, not less →

to understand, build & maintain relationships

Building and holding a real relationship is presence over time. There is no shortcut a model provides.

Social skills are getting more valuable, not less →

to help, lead & drive others forward

People follow people. Leadership rests on trust and accountability that cannot be generated.

Social skills are getting more valuable, not less →

to listen & take in things you hear

Being truly heard by another person is irreplaceable. Transcription is automatable; listening is not.

Social skills are getting more valuable, not less →

to resolve conflicts between people

Resolving conflict between people demands presence, neutrality and trust. It sits at the human core.

Social skills are getting more valuable, not less →

to talk & tell

Live conversation, with presence and timing, is barely exposed. Voice tools assist; talking is human.

Social skills are getting more valuable, not less →

to work & collaborate with others

Working well with other people is a social act. It is augmented by tools but rooted in humans.

Social skills are getting more valuable, not less →

to put things in motion

Putting things in motion takes initiative and nerve. A model will draft; it will not start for you.

New work appears where old tasks disappear →

to take care & care for others

Caring for another person is presence, touch and attention. It is the safest ground on the board.

Social skills are getting more valuable, not less →

to present & speak in front of people

Holding a room in person is embodied and relational. Slides are automatable; presence is not.

Social skills are getting more valuable, not less →

to think & think

Judgment, the weighing of what matters and what to do, is the irreducible core. AI extends thinking; it does not own it.

Most real AI use augments, it does not replace →

to teach other people things

Teaching people is a relationship with attention and adaptation. AI tutoring helps; the teacher still matters.

Social skills are getting more valuable, not less →

to think big & wide

Vision, deciding where to go and why anyone should care, is direction-setting. Models do not have a stake.

New work appears where old tasks disappear →

to let the thoughts flow away

Letting the mind wander into genuinely your own imagination is not a retrieval task. It stays human.

New work appears where old tasks disappear →