One person, a team of agents, and clients who never notice the difference
by Mat Siems
PARTS I–X · CHAPTERS 1–100 · MATSIEMS.COM
Part I
The Agency of One
What a solo agency is, and why now.
Chapter 1 · Part I
One Person, Several Desks
Picture a small office with four desks in it. At one sits the owner, who decides what the firm sells and to whom. At the next sits the account manager, who answers the client's emails and notices when they sound worried. At the third sits the production team, turning briefs into finished work. At the fourth sits the bookkeeper, who sends invoices and frowns at the bank balance. Now walk round the room and notice that every chair has the same person in it. That is a solo agency. It always was. What has changed is the third desk.
This book is about running that office in 2026, when the production desk can be staffed by AI agents that draft, research, build, format, test and revise at a pace no single human could match. It is a book about the business and the operating model: what you sell, how clients find you, how the work gets made and checked, how money moves, and how you stay sane while it does. It is not a productivity manual. Your morning routine is your own affair.
An agency, for our purposes, is a firm that sells a repeatable outcome to a series of clients and runs a system to deliver it. The word that matters is repeatable. A one-off favour for a friend of a friend is not an agency. Neither is a single brilliant engagement you will never do again. The agency is the thing that can do the same kind of good work for the tenth client as well as the first, without the owner reinventing it each time.
An agency is a promise you can keep more than once.
For years, the solo version of this was a contradiction that people managed by working very long hours. One person could make the promise, but could only keep it a few times at once before the production desk buckled. You either stayed tiny, raised prices until few could afford you, or hired, and hiring turned you into a manager of people rather than a maker of work. Plenty chose hiring and discovered they did not enjoy it.
The agents do not abolish that trade-off, but they move it a long way. Production that once needed three juniors can now be done by one person who knows how to brief, check and assemble the output of several agents. The other desks are still yours, and they are still demanding. The chapters that follow are about doing all four jobs well, in the right proportions, without pretending any of them is optional.
So this week, do something small. Draw the four desks on a piece of paper and write, beside each one, roughly how many hours of last week went there. Most people find the production desk is fat and the owner's desk is starved. That ratio is what this book is trying to reverse. The chairs stay. The weight moves.
Fig 1 · One Person, Several Desks. One person fills all four desks; AI agents now staff the production desk.
Chapter 2 · Part I
Not a Freelancer, Not a Consultant
People will call you a freelancer, and some days you will let them because it is easier than explaining. But the distinction is worth getting straight in your own head, because it changes what you sell, how you price it and how you spend your week.
A freelancer sells capacity. The client has work and not enough hands, and the freelancer supplies hands for a period. The unit of sale is time, whether it is dressed up as a day rate or a project fee calculated from an estimate of days. The client manages the work, sets the direction and often decides what good looks like. This is honourable, useful and, in the age of agents, increasingly squeezed, because capacity is precisely the thing that has become cheap.
A consultant, in the small-engagement sense, sells a judgement. A client has a question, the consultant studies it and gives an answer: a diagnosis, a recommendation, a review. The engagement is usually short and often unique. The value is in the thinking, and the deliverable is frequently a document or a conversation. There is a whole other book to be written about that model, and it is not this one.
An agency sits between and slightly above them. It sells an outcome, delivered by a system, repeatedly. The client does not manage the work; they buy the result and trust the firm to produce it. The agency owns the method, the quality standard and the schedule. It brings both the judgement of the consultant and the production of the freelancer, but packages them as a service with a shape the client can understand before buying.
A freelancer is hired. A consultant is asked. An agency is engaged.
Why does this matter now? Because the agents have made the freelancer's core asset, hours of skilled production, far more abundant. If you sell hours, you are selling the thing whose value is falling fastest. If you sell only advice, you are bounded by how many conversations you can have. If you sell an outcome and own the system that produces it, the cheaper production becomes, the better your margins, provided you price the result rather than the effort.
There is a practical test you can run on your current work. Look at your last three invoices. If the line items describe time, you are freelancing. If they describe a recommendation, you are consulting. If they describe a result the client now has, under a name you would recognise on a menu, you are running an agency. Most people find a mixture, and that is fine as a starting point. The job of the next few parts is to move more of the mix towards the third kind, on purpose, one offer at a time. The labels are not the point. The economics behind them very much are.
Fig 2 · Not a Freelancer, Not a Consultant. Freelancer, consultant and agency compared by what they sell and who directs the work.
Chapter 3 · Part I
What Changed When the Agents Arrived
It is tempting to describe the arrival of capable AI agents as a revolution, and then to spend the rest of the conversation in a state of either excitement or dread. Neither is especially useful when you have a client deliverable due on Thursday. A calmer description is that one input to your business became very much cheaper, and the others did not.
The cheap input is production. Drafting copy, assembling research, writing code, building a slide deck, cleaning a spreadsheet, transcribing and summarising a call, generating twenty variants of a headline, checking a document against a style guide. Work that once took a person an afternoon can now be done in minutes by an agent that has been briefed well. It is not free, and it is not always right, but its cost has fallen far enough that it no longer sets the limit on what a small firm can deliver.
The inputs that did not get cheaper are the ones that were always scarce. Knowing what a particular client actually needs, as opposed to what they asked for. Deciding which of five plausible drafts is right. Noticing that the brief has drifted. Earning enough trust that a client will let you make a decision on their behalf. Holding a difficult conversation about a missed deadline. These cost the same as they always did, which is to say they cost you, your time and your attention, and there is only one of you.
When one input gets cheap, the expensive ones decide the shape of the business.
That shift reorganises the solo agency from the inside. In the old model, the owner spent most of the week producing and squeezed judgement and relationships into the gaps. In the new model, production shrinks to a supervisory task, and the owner's week can tilt towards the scarce work. Can, not will. Plenty of people use agents to produce more of the same work at the same price, then wonder why they are just as tired and slightly worse paid.
There is also a change on the client side. Your clients have access to the same agents. They know, roughly, that a first draft of almost anything can be produced quickly. What they cannot easily produce is a first draft that is right for them, checked by someone who understands their situation, delivered on time, and followed by a second and third piece of work that builds on the first. That is what they will pay an agency for.
So the useful question is not whether agents will replace you. It is which part of your work they will replace, and what you will do with the time. Write down the three tasks that consumed most of your production hours last month, and ask honestly how much of each an agent could now do under your direction. The answer is rarely all of it. It is almost never none. Cheap production is not a threat to an agency. It is a gift to anyone who knows what to make.
Fig 3 · What Changed When the Agents Arrived. Production cost collapsed while judgement, trust and hard conversations cost the same.
Chapter 4 · Part I
The Clients Do Not Care How
The subtitle of this book promises clients who never notice the difference. That phrase needs handling with care, because it could be read as a promise of concealment, and concealment is a poor foundation for a business built on trust. What it actually means is simpler and more demanding: the work is so consistently good, and the service so steady, that the client has no reason to think about how it was made.
Clients buy outcomes. A restaurant customer does not ask whether the stock was made that morning or yesterday; they ask whether the soup is good, and they would certainly mind if it were not. In the same way, a client buying a monthly content programme, a website rebuild or a quarterly analytics report cares about whether it is right, on time and useful. The inside of your kitchen is your business, as long as what comes out of it is what you promised.
That said, there is a line, and you should know exactly where it is. If a client asks how the work is produced, tell them, plainly and without defensiveness. If a contract or a client's policy restricts the use of AI tools, or of particular kinds of data in them, honour it to the letter. If a deliverable's value depends on it being personally made by you, such as a signed opinion or a piece of writing under your name, do not quietly hand it to an agent and pretend otherwise. Seamless is fine. Secret is not.
The client should never notice the difference because there is no difference worth noticing, not because you hid it.
In practice the best approach is to state your way of working once, early and calmly, in your onboarding material: you use AI agents for production, every piece of work is directed and checked by you, and client data is handled according to the terms you have agreed. Most clients will nod and move on. A few will ask sensible questions. A very few will have constraints you need to know about before you start, which is exactly why you mention it.
What clients do notice is inconsistency. A brilliant first deliverable followed by a sloppy second one. A report with the previous client's name in the footer. A tone that wanders from week to week. Prose with the faint gloss of something nobody really read. These are the tells, and they are tells of carelessness rather than of tools. A hand-made deliverable full of errors is worse than an agent-made one that has been properly checked.
So treat the phrase in the subtitle as a quality standard, not a marketing trick. Look at the last thing you sent a client and ask whether anything in it would make them wonder how it was made. If so, the problem is not that they might find out. The problem is that it was not finished. Honesty about method is cheap insurance. Consistency of result is the actual product.
Fig 4 · The Clients Do Not Care How. A decision path for when agents may produce the work and when you must disclose or do it.
Chapter 5 · Part I
Judgement Up, Production Down
Every solo agency now has to decide, explicitly or by drift, how work is split between the owner and the agents. Drift is the default, and drift tends to produce a muddle in which you do a bit of everything and supervise nothing properly. It is better to draw the line on purpose.
The rule of thumb is short: production goes down to the agents, judgement stays up with you. Production is the making of things whose shape is already decided. Judgement is deciding the shape, choosing between options, and declaring something finished. A brief for a landing page is judgement. Drafting the copy is production. Choosing which of three drafts to send is judgement. Formatting it for the client's content system is production. Telling the client that the headline they love will confuse their customers is judgement of the most valuable kind.
It helps to picture the work as layers. At the top sits the relationship: understanding the client, keeping their trust, knowing what matters to them this quarter. Beneath it sits direction: briefs, standards, decisions about approach. Beneath that sits production: the drafting, building and assembling. At the bottom sits verification, which is shared, because agents can run checks and tests while you do the final read. The top two layers are yours. The third belongs to the agents. The fourth is a partnership.
Delegate the making. Keep the deciding. Share the checking.
The common mistake is to let the line creep upwards. An agent drafts a brief for itself, then executes it, then summarises the result for the client, and you approve the whole chain with a glance. Each step looks efficient. Together they mean that no human has actually decided anything. The work may well be competent. It will also be generic, because the one thing that made it specific to this client, your judgement about their situation, never entered the process.
The opposite mistake is to keep production for yourself out of habit or pride. Some owners rewrite every agent draft from scratch because they cannot bear to edit. That is understandable and expensive. If you find yourself doing this, the fix is usually upstream: a better brief, a better example, a clearer standard. Rewriting is a symptom of a brief that did not carry your judgement into the work.
Try this on your next deliverable. Before you start, write one sentence for each layer: what the relationship needs, what the direction is, what the agents will produce, and how it will be checked. Four sentences. If you cannot write the second one, you are not ready to delegate the third. Most delivery problems begin at the layer above the one where they show up.
Fig 5 · Judgement Up, Production Down. Relationship and direction stay with you, production goes to agents, checking is shared.
Chapter 6 · Part I
The Operating Model on a Napkin
Before going deep into any one part of the business, it is worth seeing the whole thing at once. A solo agency is not a collection of tasks. It is a loop, and the loop has four stations.
The first station is the offer: what you sell, to whom, in what shape. The second is the pipeline: how people who need the offer find out about it and come to talk to you. The third is delivery: how the work gets made, checked and handed over. The fourth is money: how you get paid, how cash moves, and how much is left. Then the loop closes, because good delivery creates referrals, case studies and repeat work, which feed the pipeline, and what you learn in delivery sharpens the offer.
It sounds obvious drawn on a napkin. It is less obvious in practice, because most owners spend their attention on whichever station is on fire. When the pipeline is empty, everything is marketing. When a big project lands, everything is delivery and marketing stops. When an invoice goes unpaid, everything is money for a week. The loop keeps turning, but jerkily, and the jerks are what make the business feel precarious.
A business is a loop. Treat it as a list and it will treat you as a firefighter.
The agents help most at the delivery station, which is why so much of this book lives there. But they can assist at every point. They can research prospects for the pipeline, draft proposals from call notes, prepare invoices and payment reminders, summarise delivery data into case studies, and track which offers are selling. What they cannot do is decide which station deserves your attention this week. That is the owner's desk, and it is the one most often left empty.
A useful habit is a weekly review of the loop, short enough to survive a busy week. Ask one question per station. Offer: is it still the right thing to sell, and has anything we learned changed its shape? Pipeline: how many good conversations are booked for the next fortnight? Delivery: what is late, at risk or quietly getting worse? Money: what is owed, what is due, and how many months of costs are in the bank? Four questions, perhaps twenty minutes, written down in the same place each week.
The value of the review is not the answers. It is the habit of looking at all four stations even when one is shouting. Over a few months you will notice patterns: the pipeline always thins two months after a big project starts, or cash is always tight in the same quarter. Those patterns are where the real improvements live.
Draw the napkin. Put it somewhere you will see it. Then book the twenty minutes for Friday afternoon, when the week is still fresh and the weekend is near enough to make you efficient. The loop does not need to be clever. It needs to be looked at.
Fig 6 · The Operating Model on a Napkin. Offer, pipeline, delivery and money form a loop, checked in one short weekly review.
Chapter 7 · Part I
The Myth of Infinite Capacity
There is a moment, usually a few weeks after an owner starts using agents seriously, when the arithmetic seems to change. A job that took three days now takes one. Surely, the thinking goes, I can take on three times as many clients. Some people act on this immediately. Most of them discover, a couple of months later, that they have tripled something, and it is not their profit.
What agents scale is output. They can draft, build and assemble far more than one person could. What they do not scale is the owner's attention, and attention is consumed by everything that sits around the production. Each client brings emails, calls, decisions, feedback, small crises and the steady background effort of keeping a relationship warm. Each deliverable needs a brief written and a result reviewed. Each new project needs onboarding. None of that shrinks just because the drafting got faster.
So capacity is not set by how fast the work can be made. It is set by how many clients, projects and decisions you can hold in your head at once without dropping any. That number varies with the person and the work, and it is much lower than the agent arithmetic suggests. Exceed it and the first thing to go is not the production. It is the judgement: reviews get shallower, briefs get thinner, and client messages get answered later and more briefly.
Agents multiply the work. Nothing multiplies you.
The practical response is to find your real limit before a client finds it for you. Look back over the past year for the periods when you felt in control and the periods when you felt behind, and count the active clients and projects in each. The number at which control turned into scrambling is your current ceiling. It is probably smaller than you would like, and that is useful information rather than a failing.
Then use the agents to raise the ceiling in the right places. Not by taking on more clients, at first, but by stripping the attention cost out of each one. An agent can prepare a briefing on every client before their weekly call. It can draft status updates for you to edit. It can keep a running log of decisions and open questions per project so that you never have to reconstruct context from memory. Each of these lowers the cost of holding a client in mind, and only then does it make sense to hold more of them.
There is a gentler point underneath all this. The temptation to fill all the capacity the agents create is a temptation to replace one kind of overwork with another. You could instead keep the same number of clients, serve them better and finish on time. That is not a failure of ambition. It may be the most profitable choice on the table.
Fig 7 · The Myth of Infinite Capacity. Capacity is set by attention per client; agents raise it by cutting overhead.
Chapter 8 · Part I
Small on Purpose
Most small firms describe themselves, at least privately, as not yet big. The single owner is a stage before the team, the team a stage before the department, and so on up the ladder of ambition. This book takes a different view. Staying small can be a strategy rather than a waiting room, and in the age of agents it is a more viable strategy than it has ever been.
Small has real advantages that large firms spend fortunes trying to imitate. Decisions are fast because there is one decision-maker. The client deals with the person who does the thinking, not a junior who reports to someone who once met them. The cost base is light, so the firm can survive a lean quarter without layoffs. Quality is consistent because the same standards are applied by the same eyes. And the owner gets to do the work they actually set up the business to do, rather than becoming a full-time manager of other people doing it.
The traditional disadvantages of small were capacity and range. One person could only produce so much, and could only be expert in so many things. Agents chip away at both. Capacity, as the previous chapter warned, is still bounded by attention, but the production ceiling has risen sharply. Range is extended too: an owner with strong judgement in their core field can now deliver competent adjacent work, research, formatting, data handling, light coding, through agents, without hiring a specialist for every corner.
Big is a size. Small is a choice.
Choosing small on purpose changes several decisions downstream. You price for margin rather than volume. You select clients for fit rather than size. You build systems for consistency rather than for onboarding staff. You turn down work that would require a team, or you partner for it rather than hiring. You measure success by profit, quality of life and the quality of the work, rather than by headcount or revenue alone.
Not everyone should stay small for ever, and the final part of this book takes the question of hiring seriously. But the decision to grow should be made from a position of choice, not drift. Many solo owners hire because the workload grew and hiring seemed the only answer, then find themselves running a business they enjoy less and earn proportionally little more from.
Try writing a single paragraph describing your agency three years from now if it stayed exactly this size. What would you sell? What kind of clients? How would the week feel? If the paragraph sounds pleasant, you may already have your strategy. If it sounds like a trap, you know what to fix, and it may not be the headcount. Ambition is not measured in desks.
Fig 8 · Small on Purpose. Size against intent: staying small on purpose beats growing, or staying, by drift.
Chapter 9 · Part I
Your First Honest Audit
Everything in the chapters ahead depends on knowing where your time actually goes, not where you think it goes. So before changing anything, it is worth spending a week watching yourself work, in the manner of a mildly suspicious time-and-motion inspector.
The method is unglamorous. For five working days, keep a running log in whatever tool you already use. Every time you switch task, write the time and a few words. At the end of each day, mark every entry with one of three letters: J for judgement, R for relationship, P for production. Judgement is deciding: scoping, choosing, briefing, approving, prioritising. Relationship is people: calls, client emails, sales conversations, the small kindnesses that keep work flowing. Production is making: drafting, building, formatting, researching, checking against a list.
An agent can help with the bookkeeping. Feed it the raw log at the end of the day and ask it to total the hours by category and flag entries that look misclassified. It is surprisingly good at noticing that "quick email to client" took fifty minutes and probably involved three decisions. The categorising, though, is yours. Only you know whether a given hour was spent deciding something or merely doing it.
You cannot delegate a week you have not looked at.
When the week is done, look at the totals. Most solo owners discover that production dominates, often by a wide margin, and that a lot of what felt like judgement was really production in disguise: reformatting a deck, rewriting a paragraph the agent had nearly got right, chasing a file. Relationship work is usually smaller than expected and squeezed into the edges. Pure judgement, the hours spent actually deciding what to do and why, is often the smallest slice of all.
Now go through the production entries and ask of each one whether an agent could do most of it under a clear brief. Mark the yes answers. Then go through the judgement entries and ask whether any of them were really decisions you could have made once and written down, so that you would never need to make them again. Mark those too. The first list is your delegation backlog. The second is your playbook backlog, which becomes important when we reach the delivery machine.
Keep the log. Repeat the audit in three months. The point is not a perfect ratio, because there isn't one, and the right mix depends on your offer and your clients. The point is to see the drift. Without a record, every week feels the same and the business changes without your noticing. With one, you can see the weight moving between the desks and push it where you want it. What gets measured gets managed, but what gets written down gets noticed first.
Fig 9 · Your First Honest Audit. A week of logged hours, tagged J, R or P, becomes a delegation and a playbook backlog.
Chapter 10 · Part I
The Edge Is Taste and Trust
If production is cheap and available to everyone, including your clients and your competitors, what exactly is a solo agency selling? The answer this book will keep returning to is two words: taste and trust. They are the edge. Headcount is not.
Taste is the ability to tell good from merely competent, and to know what good means for this client, this audience and this moment. An agent can produce ten versions of a homepage. Taste is knowing which one will make a cautious finance director feel safe enough to book a demo, and why the other nine will not. It is knowing when to cut, when a plain sentence beats a clever one, and when the brief is wrong. Taste is built from years of seeing work succeed and fail, and it is personal: two people with equal skill will have different taste, and clients choose between them partly on that basis.
Trust is the client's confidence that you will do what you said, tell them the truth, protect their interests and notice problems before they become crises. Trust is what lets a client stop checking your work, which is the moment an agency relationship becomes genuinely valuable to both sides. It is earned slowly, through consistency, and lost quickly, through a single careless email. Trust also travels: it is what makes clients refer you to people they respect.
Anyone can rent the production. Nobody can rent your taste or borrow your reputation.
Large agencies have always claimed taste and trust too, of course, but they deliver them through layers. The person with the taste is often not the person doing the work, and the person the client trusts is often not the person who will be there next year. A solo agency can offer both directly: the person who decides is the person who checks is the person who answers the phone. Agents make that offer scale further than it used to, without diluting it.
This has a practical consequence for how you invest your time. Every hour spent on production is an hour not spent on the two things that differentiate you. That does not mean production is beneath you, or that you should never make anything with your own hands. It means you should make things with your own hands deliberately, for the reasons of craft and taste-building, rather than by default because you never got round to building a system.
As you read on, notice how many of the practices that follow are really ways of protecting taste and trust: briefs that encode your judgement, reviews that catch the generic, communication that never surprises the client, money habits that let you say no to bad work. The machinery varies. The edge stays the same. Hold that thought for a hundred chapters. It is the spine of the book.
Fig 10 · The Edge Is Taste and Trust. Taste and trust overlap in one person who decides, checks and answers: that is the edge.
Free preview · Part I
Keep reading
You have read the first 10 of 100 chapters of The Solo Agency. The full book has every chapter and plate, as PDF and EPUB in English, French, Polish and Spanish.