Home / Articles / The jobs that never scaled: on-demand expertise, the work before and after the meeting, and why a historian, a librarian or a consultant can now work for every project that needs one

The jobs that never scaled: on-demand expertise, the work before and after the meeting, and why a historian, a librarian or a consultant can now work for every project that needs one

By Dinis Cruz · 2026-10-21 · article v1.0.0 · newsroom v0.13.18 · gen-z-bookworkagentsexpertiseconsultingstartupsarticle

Abstract: In the book I say that the opportunity is not only in new jobs, but in jobs we have done for a very long time that never really scaled, and now can. The editor asked for more detail, and fairly: without examples it is a claim without a case. So here it is. What never scaled is on-demand expertise: getting the value of someone with skills the team does not have, for the few days it is needed. An engagement like that has three phases. Before, the expert has to gather context, and the questions they ask are their experience. Then the meeting, short, where the value is. After, the write-up, for every audience. The first and third took weeks, most experts dislike them, and they are not what the expert is hired for, so only big companies could pay for it, and only when the company, the talent and the travel all lined up. The clinic has long worked differently: others take your history, so the specialist decides fast. Agents can now be that support for any expert, anywhere, even twelve hours a week. My own agent team has a historian, a librarian, a cartographer, an accountant and a sherpa: roles I could never have had. That is the opportunity, for the people who have spent twenty years in one of those jobs, and for the startups that never scaled before.

In short

• What never scaled is on-demand expertise. Skills the team does not have, for the days they are needed.
• Every engagement has three phases. Before, gathering context; the meeting; after, the write-up.
• The before and the after were the cost. Weeks of work most experts dislike and are not hired for.
• So it only happened when everything lined up. The company, the talent, the travel and the budget.
• The clinic already works the other way. Others take your history, so the specialist decides fast.
• Agents can now be that support for any expert. Anywhere, even twelve hours a week.
• That is the opportunity. For people who know an old job well, and for startups that never scaled.

Where this comes from

In the book, Generation ZX Developers (GenAI Startup Edition), the chapter Where do you land? ends with a section called The jobs that never scaled. It says that the opportunity "is not only in new jobs. It is in jobs we have been doing for a very long time, sometimes for millennia, that never really scaled and never had good technology, done one person at a time." The idea comes from You are lucky to be here, where I first wrote it down.

The editor's note on that section is direct: it needs "three examples of jobs that never scaled, and now can", because "without them this section is a claim without a case." That is fair. This article is the case.

What never scaled: on-demand expertise

Take the simplest example: a consultant. What is a consultant? Somebody who has skills the team does not have. You bring them in because they know something you do not, and you need it for a while, not for ever.

So I think the core of this is on-demand expertise: getting the value of someone's skills for the days you need them. That is what never scaled. Not because the skills were rare, but because of everything around them.

Three phases: before, the meeting, after

Any engagement with an outside expert has three phases, and each one matters.

Before: gathering context. The expert needs a lot of information, and they have a very specific set of questions and data points they want. That is part of their magic: what questions to ask. The more experienced they are, the more interesting those questions are, because they go for the non-obvious things. When I look at application security, the first thing I want to know is what their CI pipeline looks like, because if they do not have a good one, I can already tell you they are going to have a lot of problems. Then: how is the team set up, how are they funded, how are they rewarded, what are their non-functional requirements, what is the scope. Every professional has a mental model like that, built from experience.

The meeting, or meetings. These are short, and they are where you really get a sense of it. You meet the stakeholders, you present what you have found, you gather more, you maybe brainstorm with a few teams. By now you have pieces of the puzzle, and you are connecting them. You arrive prepared, because you know the system. You are not arriving cold.

After: the write-up. Then the expert has to put it all together in a way that each audience can consume: the board, the engineers, the managers, the people who will do the work. That is a lot of work if you want to do it properly.

Traditionally, to do a decent job, that was about three weeks: a week or more to meet people, get the gist, read and ask the right questions; a day or two of meetings; and three days or more to write it all up. It was hard to have an impact any faster.

The parts experts do not like

Here is the thing. Most consultants, most creative people, most people who like solving problems, do not like the first phase, and hate the third. They are usually not that good at them, and that is not what they are measured on. You do not bring that person in because they are a good writer. You bring them in because they are a very good architect, or librarian, or strategist, or specialist in a domain, or a doctor.

And there is a quieter problem with the third phase. Often the expert understands the problem much better than the client does, but can only write down so much, because there is only so much the client can take in. The rest of what they understood is lost.

What changes with agents

Now, because of agents, because of the ability to collect data and to distribute it, the first and third phases can be done far more effectively.

Before the meeting, there can be a whole set of questionnaires, agents and structures that collect the information ahead of time. That is already how I work: I send a note saying, here is what I need to know. What comes back can be packaged, filtered and sanitised, with sensitive information removed, and then given to the outside expert, who can analyse it, form a view and come to the meeting with a diagnosis to test.

After the meeting, agents can turn the decisions into the actions, the roadmap and a version for each audience, and hand the actions out. And because each audience gets its own version, more of what the expert understood can reach the people who need it, instead of the one document a client has time to read.

The clinic already works this way

This is not new as a shape. Think of what happens when you go to a clinic or a hospital. The first people you meet are the ones collecting information from you: the forms, the reception, the nurses. Eventually that is presented to the doctor or the specialist, who gets the key pieces of information and can make a decision very fast. And then a team carries out what was decided.

The specialist is not effective because they work less. They are effective because everything around them is built so that their time goes only on what needs them. What is new is that this support no longer needs a building full of people. Agents can do it, for any expert.

The intersection that made expertise rare

Most consulting in the past happened at a strange intersection. You needed a company, in a location, with a problem; somebody who had hired the talent that fitted that problem; someone who could physically go to the office and meet them; and the money to pay for all of it. That is very inefficient and very expensive, so it was only used for specific things, and mostly by big companies.

And the talent is not, most of the time, going to live close to the client. It is spread around the city, the country, the world. When the context travels to the expert, and the delivery travels back, that intersection stops mattering.

Jobs that never scaled, and now can

You can go around all the houses and find professions where you could never get the value of the person without them having a large support network around them:

• The historian. One of the most important roles in a project, because they give it the context of its own history. We complain that we do not learn from history; it is because we do not capture history in a way that can be consumed. Imagine every project with a historian. It does not need one full time: it needs one at specific points, to set things up, to connect and map what happened.
• The librarian. Everything findable, connected and current. My favourite agent.
• The cartographer. The map of what connects to what, and where the boundaries are.
• The architect, and the software architect. Someone who has seen a hundred systems, for the week a decision is made.
• The on-demand consultant. Skills the team does not have, for days, not months.
• The lawyer reviewing a contract. The clauses that matter, found before the meeting.
• The accountant. What things cost, what they return, and where the money should flow.
• The doctor and the specialist. The history taken first, so the decision is fast.
• The project manager. Who is doing what, what is blocked, and what is next, kept for everyone.
• The security lead. The questions that find the real problems first.
• The sherpa. A guide for each user, watching where they struggle.
• The translator. The same information, in the right language and form, for each audience.

Can agents do all of this on their own? Some parts of it. But agents can go in all directions, and you need somebody who provides the direction. That is the role of the expert: the experience that knows which questions matter, and the judgement that decides.

The team I could never have had

I can see this in my own work. The SG/Send project's agent team has nineteen roles, and many of them are ones I could never have had before. The historian's mission: "Track every decision, spec change, and architectural evolution so the team always knows what was decided, why it was decided, and what it superseded." The librarian's: "Maintain knowledge connectivity across all project artifacts, ensuring every document is discoverable, cross-referenced, and current." The cartographer maps "the system topology, data flows, security boundaries, and dependency relationships". There is an accountant, whose mission begins "Make the financial reality of the project visible", a sherpa for users, a journalist, a translator, and the usual developers, architects and operations roles around them.

None of those roles was affordable for a project of that size before. Each one is a job that never scaled, now working for one project, and directed by a person.

Every field has its geeks

As a CISO, I work with almost everybody in an organisation, and something has always fascinated me: most roles that have any agency, which is most roles in a modern company, attract a particular kind of person. We call the technical people geeks, but every industry has them. Talk to someone deep in an industry and they have their own words, their own three-letter acronyms, their own definitions, their own ontology and taxonomy. They geek out on things you would never think of, and they look at what you love doing in the same way.

Their culture and their experience shaped what they focus on, and so did their limitations: what they are good at and what they are not. In a way, those limitations are what make them good. That is exactly the kind of person whose value was locked behind the before and the after.

Twelve good hours

These are roles you could never hire part-time and get real value from. Think of a parent working from home: a remarkable resource, who often needs the mental stimulation, but may only want to work twelve hours a week for this period of their life. Nothing wrong with that. You just need to make sure those twelve hours are very productive, so they get the best of both worlds.

I know many people who moved, or changed jobs, and who miss what they did: the challenge, the ideas, the moments when something clicks. They are highly qualified, and they cannot connect to that work again, because until now there was no environment that prepared things for them and made them productive.

Which is one of the things I have always found strange about companies. They spend money on all sorts of things, and very little on making their people productive. There should be a department whose whole job is to make people productive: to manage their data, and to focus on what they need, not on what it happens to be able to do. Agents can now be that department.

Startups that never scaled, and now will

This connects to startups, and it is worth being precise. Paul Graham's essay of July 2013 starts: "One of the most common types of advice we give at Y Combinator is to do things that don't scale", such as "to recruit users manually". That is about doing unscalable things on purpose, at the start, to learn.

What I mean is different. There are whole kinds of company that never scaled, because they were services built on experts' time: the consultancy, the agency, the specialist practice, the one-person service. Their growth was bound by the hours of the people in them, most of which went on the before and the after. Foundation Capital, a venture firm, calls this shift "Service-as-Software", and puts it at "a $4.6 trillion opportunity", which is an investor's estimate, not a measurement. Whatever the number, the shape is the one in this article: when the before and the after are carried by agents, a services business can grow with the expert's judgement, not with their hours.

And that is the point for the reader of the book. If you have spent twenty years inside one of those jobs, you can build the first version of the support it never had. And if you are starting out, the jobs that never scaled are a good place to look.

Three examples, for the book

For the chapter, these are the three I would use, because I have lived each one:

• The on-demand security consultant. Three weeks for a two-day engagement, most of it gathering and writing. With agents asking my questions ahead of time, and writing up for each audience afterwards, it becomes two prepared days, for companies that could never afford the three weeks.
• The historian. Every project needs one at a few moments, and almost none had one. Now a project can have a historian agent recording every decision and why, with a person directing it at the moments that matter.
• The librarian. Making everything findable and connected was always valuable and never staffed. My librarian agent does it every day, for one project.

Related writing

The same shape runs through other things I have written: why graphs never scaled for everyone else; removing code, which needed a map and the tests to prove nothing broke; the minimum code below; due diligence, which never scaled because it was a form; and reading with a pen, where capturing what I learn from books never scaled. And the expert founders is about the people best placed to build what comes next: the ones who know the problem.

What this does not claim

• "Never scaled" means for most companies. Big firms did buy this work; most could not.
• The three weeks are from my experience. Not a survey.
• The twelve hours are illustrative. The shape is the point, not the numbers.
• Agents do not replace the expert. They carry the before and the after; the direction is the person's.
• The $4.6 trillion is Foundation Capital's estimate. An investor's figure, quoted as such.

Where this comes from

A voice memo of 21 October 2026, prompted by the editor's note in the book's chapter Where do you land? on gen-z-developers.sgit.ai, and its source, You are lucky to be here. The role definitions of the SG/Send agent team, read the same day. Paul Graham, Do Things that Don't Scale, July 2013. Foundation Capital, AI leads a service as software paradigm shift. Related: the graph companies, the code you do not ship, the expert founders, start a company. Written up and checked by this site's agent, newsroom.writer; the editorial responsibility is mine.

Want the next issue by email. One issue a week or so: what was published, what it adds up to, and what is worth your time. Subscribe to the SGit Newsroom →

← All articles