Think about the last time you gave up.
Not gave up on something dramatic. Not a relationship or a career or a dream. Something smaller. Something so routine in its frustration that you barely registered the moment you stopped fighting it.
The insurance claim you filed three times and then abandoned. The bill you suspected was wrong but could not face disputing. The gym membership you kept paying for months after you stopped going because the cancellation process required a letter, sent by post, to an address that took twenty minutes to find. The energy tariff you knew was not the best one but switching felt like a project you never had time to start. The medical referral that needed a form that needed an appointment that had a six-week wait. The invoice that sat unpaid for three months while the chasing emails went unanswered and the relationship felt more important than the money.
The expense claim from the work trip that had seven steps and a receipt format the portal would not accept so you wrote off the £40 and moved on. The tax rebate you were probably entitled to but the form was four pages long and you were not entirely sure you were reading it correctly. The cashback offer you registered for, met every condition of, and never received because one step in the redemption process was buried in the terms. The energy refund that sat as a credit on a closed account for months until you gave up asking. The parking charge you successfully appealed but whose refund never arrived and by the time you noticed it felt too late to start again.
Small money. Every single time. But real money. And the same rational calculation made every single time: the cost of pursuing this is greater than the amount I am owed. Multiplied across millions of people, that is billions surrendered every year to processes designed, consciously or not, to outlast the people they failed.
These are not failures of character. They are rational responses to irrational systems.
So you absorbed it. Filed it under the category every adult carries: things I have quietly accepted.
That category is enormous. It has been growing your entire life. It costs you money you should not be losing, time you will not get back, and a low-level background hum of things that are unresolved and probably always will be.
Until very recently, that was simply the deal. There was genuinely nothing to be done about it.
I have spent the past year building AI-powered solutions against exactly this kind of pain, working across industries and problem types that share the same underlying frustration. What I have learned is that the opportunity to change this is far larger, and far more accessible to ordinary people, than anything being written about AI right now would suggest. This piece is about why. And more importantly, it is about you.
It was never an accident
Here is the thing nobody says directly.
A significant portion of the complexity you navigate every day was not the result of incompetence. It was the product, deliberate or semi-deliberate, of systems designed to require expert intermediaries.
Tax codes written by accountants, lobbied by accountants, administered by accountants. Insurance policies drafted in language that makes the average claimant dependent on someone else to interpret. Planning regulations layered over decades until a person who wants to build a modest extension on their own home needs to hire a consultant just to understand what is being asked of them. Employment contracts so complex that both the employer and the employee need professional advice before either can act with any confidence. Healthcare referral pathways so fragmented that navigating them is a full-time job that nobody is paid to do on your behalf. Subscription services with cancellation processes deliberately designed to outlast your patience.
The complexity is the product. The friction is the business model.
This is not a conspiracy. It is something more mundane and more pervasive than that: institutional self-preservation. Every layer of process, every form, every requirement for a certified professional to sign off on something, every forty-eight page terms and conditions document: these exist partly because they serve a genuine function and partly because they create a dependency that justifies a role, a department, a profession, a company.
The moat was always artificial. It was built from language, from process, and from the assumption that ordinary people would not have the time, the knowledge, or the access to route around it.
For most of human history, that assumption was correct.
The parent managing a school appeal alone because she cannot afford an education consultant. The small contractor whose client has not paid in sixty days and who cannot afford a lawyer to pursue it. The couple who were missold a financial product and have no idea how to complain effectively. The freelancer underpaid by a platform whose only recourse is a process deliberately designed to exhaust them into giving up. The elderly man who received a letter from a government agency that he does not understand and has nobody to ask.
These are not edge cases. These are the textures of ordinary life. Hundreds of millions of people absorbing losses they should not be absorbing, because the tools to fight back were priced somewhere they could not reach.
That is what is changing now. And the change is more significant than almost anything written about AI has acknowledged.
Why nothing could fix this before
For this to change, three things had to be true at the same time. Until very recently, not one of them was.
The first was genuine intelligence. Not the ability to retrieve information or follow a script. The ability to read a specific situation, understand what actually matters in it, and work out what to do. A simple automated system cannot dispute your water bill effectively. A search engine cannot negotiate your lease renewal. A standard form cannot chase an unpaid invoice with the kind of escalating precision that actually gets results. What was needed was something that could think about your particular circumstances rather than return a generic answer that applies to everyone and therefore fully helps no one.
The second was the ability to act. Understanding a problem is not enough. You need someone, or something, to actually do the work. Draft the letter. File the form. Send the email at the right moment. Check the regulation. Monitor what comes back. Escalate when nothing comes back. Follow the thing through over days and weeks, across multiple parties and systems, without losing the thread or giving up. That is not a single request answered. That is a sustained effort conducted on your behalf, without you having to be present at every step.
The third was cost. Every previous attempt to solve these problems involved people: advisers, consultants, caseworkers, specialists. Those people were expensive. That cost meant the solution was only ever available to people who could afford it, or to businesses with enough margin to absorb it. The single parent disputing an unfair charge cannot afford a solicitor. The café owner chasing a corporate client’s invoice cannot afford a credit management firm. The family trying to understand what benefits they are entitled to cannot afford a financial adviser. The cost of expertise has always been the final wall between people and the outcomes they deserved.
All three of those conditions now exist. Not perfectly. Not without problems. But sufficiently, and for the first time all at once.
That is the shift. Not AI in the abstract. Not any single product or platform. The specific combination of genuine judgment, real-world action, and a cost that ordinary people can actually afford. The wall is not gone. But there is a door in it now, and the door is getting wider.
What it actually costs now
Let us be specific about the economics, because this is where the reality of the change becomes undeniable.
Previously, navigating complexity cost either your time or your money. Usually both. That had been the equation for as long as professional services existed. You paid someone who spoke the language, or you found the time to learn it yourself, or you gave up. Those were the options.
The equation has changed.
A lawyer’s letter costs between £150 and £500. An AI agent that drafts a formal demand with full legal precision, references the correct legislation, and personalises it to the specific situation costs almost nothing to produce. The legal weight is the same. The result is frequently the same. The access is now universal.
A bookkeeper costs between £300 and £800 a month for a small business. An AI agent that categorises every transaction, reconciles accounts, flags anything unusual, and prepares everything needed for a tax return costs almost nothing to run. Not a replacement for a senior accountant making complex decisions. A genuine replacement for the grinding, time-consuming routine work that most small business owners are currently doing themselves at ten o’clock at night.
A reputation monitoring service costs £500 to £2,000 a month. An AI agent that watches review platforms, social channels, and local news, drafts responses to negative reviews within the hour, and flags emerging problems before they become crises costs almost nothing at the scale a small independent business actually needs.
An immigration specialist charges £200 to £500 an hour. An AI agent that understands the current rules for a specific visa, assesses your situation against those requirements, builds your documentation checklist, and drafts your application letter operates at a cost that is not measured in hours.
A benefits adviser, where you can find one and get an appointment, costs time, travel, and waiting. An AI agent that knows the current entitlements in your situation, identifies what you are eligible for but not currently receiving, and guides you through each application step is available any time, on any device, in your own home.
And here is something you can do right now, today, at no cost. Upload your last three months of bank statements to Claude and ask it to find every subscription, recurring charge, and direct debit leaving your accounts. Most people are genuinely surprised by what comes back. The trial that converted to a paid plan. The app deleted but never cancelled. The duplicate charge across two cards for the same service. The small monthly amount that has been leaving quietly for two years for something you cannot even name. This is not a future capability. It is available to anyone with a bank statement and five minutes. It is also the simplest illustration of what is now within reach: a tool that does the patient, methodical work that humans rationally avoid because the effort outweighs the individual reward, but that adds up to something significant when done properly.
The comparison is not always perfect. An AI agent is not always the equivalent of a seasoned professional. There are situations where human judgment, formal accountability, and professional indemnity genuinely matter and should not be substituted. But for the problems that ordinary people have been quietly losing for decades? The gap between what an AI agent can now do and what a professional was being paid to do is closing faster than most of those professionals would like to acknowledge.
The old equation: solving this problem costs your time or your money. Usually both.
The new equation: solving this problem costs the price of an AI subscription. And that price is falling every month.
The scale of what is sitting unsolved
Think about the parent who received a letter saying their child’s school placement appeal has been rejected and who has no idea whether they have further rights, what the timeline is, or how to write a response that will actually be taken seriously. The process was designed by people who navigate it professionally. The parent is entering it once, under enormous stress, with everything that matters to them at stake.
Think about the person who bought a used car that turned out to have finance outstanding against it and now needs to understand their consumer rights, draft a formal complaint, and escalate to a regulator if the dealer does not respond. Each of those steps requires knowledge they have never needed before and time they may not have.
Think about the small landlord who has a tenant who has stopped paying rent, does not know the correct legal process, is terrified of getting a single step wrong, and cannot afford the lawyer who would charge £1,000 to handle it properly. The knowledge exists. The forms exist. The process is navigable. But only if you know the language, and most people do not.
Think about the freelancer whose long-term client went quiet after the last piece of work was delivered, whose agreement was informal, and who does not know whether to send a formal demand, involve a mediator, take the matter to small claims court, or write off the debt and move on. Each path has a right sequence. None of them are explained anywhere that a normal person would find.
Think about the family who inherited a small property and have no idea what the tax position is, what needs to be filed, by when, or what happens if they get it wrong. The rules are publicly available. Understanding them well enough to act on them correctly is a different matter entirely.
Think about the person on a zero-hours contract who believes they have been treated unfairly, knows there are employment rights that probably apply to their situation, but has no idea where to start and even less idea whether starting would be worth the risk.
Think about the carer managing prescriptions, GP referrals, specialist appointments, social care assessments, and benefit claims for a family member, while also working, while also running a household. Each of those systems has its own language, its own timeline, its own process. Navigating them simultaneously is a second full-time job that nobody applied for and that nobody is being paid to do.
Every single one of those situations contains a solvable problem. Every single one represents thousands, sometimes hundreds of thousands, of people experiencing the same friction in the same way at the same moment. And in every single one of them, the person experiencing it has quietly accepted it as the cost of being in their situation, because they had no other realistic option.
That is the map. It is vast. And most of it is untouched.
This is for you
Here is what I want you to do now.
Think about your week. Not the interesting parts. The grinding parts. The thing you did that felt like a waste of two hours but had to be done. The form. The call that put you on hold for twenty minutes. The email chain that went back and forth six times to resolve something that should have taken one exchange. The process you completed by hand because one system does not talk to another. The document you chased for the third time. The thing you eventually wrote off because fighting it any longer was simply not worth it.
Now ask: who else has that problem?
Not just you. How many people in your line of work, your neighbourhood, your community, your country deal with the exact same thing every week? Hundreds? Thousands? Hundreds of thousands?
And ask: do you understand it more deeply than most? Do you know why the process works the way it does, who created it, where it always breaks down, what the person stuck inside it actually needs to get out?
If the answer to those questions is yes, you are holding something that two years ago would have required a team of software engineers and a significant amount of investment capital to act on.
You are holding the beginning of a business.
Picture Sarah. She spent twelve years working in HR for a mid-sized company. She processed redundancies, managed grievances, sat in the room when difficult things were said. She knows exactly where people get the process wrong, exactly what their rights are, and exactly what information they would need at each step to protect themselves. She is not a lawyer. She has never built a product. She would not have described herself as someone who could start a technology business.
But she knows the problem at a level of depth that no one who has not lived inside it could replicate quickly. And the tools that would allow her to build something that guides a person through a redundancy situation, explains their rights in plain language, drafts the letters they need, and flags when something is not right: those tools exist today, are accessible without specialist technical training, and cost a fraction of what they would have cost eighteen months ago.
That is one person. One problem. One business that does not yet exist.
Multiply it by every person who has accumulated that kind of knowledge in their field and the map becomes extraordinary.
The accounts manager who knows exactly why the reconciliation process breaks every quarter and what it costs the people trying to fix it. The nurse who knows exactly why discharge coordination fails and what it takes to get someone through the system without falling through a gap. The letting agent who knows exactly why deposit disputes take six weeks and what both sides actually need to resolve them in six days. The restaurant owner who has run a kitchen for eight years and knows every compliance requirement, every supplier pressure point, and every moment where independent operators lose money they should not be losing. The parent who spent two years navigating the special educational needs assessment process and came out knowing more about it than half the professionals she dealt with. The retired HR director. The experienced mortgage broker. The former insurance claims handler.
Every single one of these people has something that cannot be downloaded, cannot be googled, and cannot be replicated quickly by someone who simply read about the problem. They have lived knowledge. Real depth. The understanding of not just how the process works but why it fails, who it fails, and what would actually fix it.
That knowledge is the asset. The technology has finally caught up with it.
The question is not whether you are technical enough. The barrier is not code. Platforms and tools available today are designed to be used by people who are not software engineers, and they are becoming more accessible every month. The barrier is commitment: the willingness to sit with a problem long enough to understand it at genuine depth, to find your first ten customers before thinking about your first hundred, and to build something that genuinely solves one specific problem before expanding to the next one.
The economics are more accessible than most people assume. A solution that helps five hundred people or small businesses save three to four hours a month, and charges £39 or $49 a month, generates more than £230,000 in annual revenue. One person. A problem you already understand. The cost of serving your five hundredth customer is barely more than the cost of serving your fifth. That has never been true before for a product that requires real intelligence to deliver.
This is the democratisation of building. Not for engineers. Not for founders with contacts in San Francisco or connections in venture capital. For ordinary people with extraordinary knowledge of a specific broken thing.
The tools have arrived. The knowledge was always yours.
What is genuinely hard
This piece would not be worth reading if it stopped at the opportunity.
AI agents fail. This needs to be said clearly and without softening. The final stretch of a complex task is genuinely difficult. Systems that work well during testing behave unexpectedly when real people use them in ways you did not anticipate. A person who has been confidently misled by an automated system before is not easily persuaded to trust one again, especially with something that matters to them.
The legal and regulatory environment is genuinely uncertain. Who is responsible when an agent files an incorrect form? Who is liable when an automated letter goes to the wrong person with the wrong information? These questions do not yet have settled answers, and anyone building in this space needs to hold that uncertainty honestly rather than quietly hoping it resolves itself.
Trust in systems that act on your behalf builds slowly, and it should. People will allow a tool to draft something before they allow it to send it. They will allow it to monitor before they allow it to act. They will allow it to recommend before they allow it to decide. Building for where people actually are, rather than where you wish they were, is the difference between something people genuinely use and something they try once and put down.
And the knowledge requirement is not flexible. The person who builds something lasting here is not someone who read about the opportunity and went looking for a problem to solve. It is someone who has been frustrated by the same broken process for years, who understands it at a level of depth that is not easily replicated, and who finally has tools that match what they already know. The technology is accessible to everyone. The real understanding of the problem is what cannot be faked.
These are not reasons not to build. They are reasons to build carefully: to start narrow, to work with real people early, to treat your first version as something you are learning from rather than something you are launching, and to resist the pull to grow before the core problem is genuinely solved.
The field is tilting
The field is not level. It is worth being honest about that.
The established firm still has regulatory standing, brand recognition, professional insurance, and relationships built over decades. The large platform still has distribution that a new entrant cannot replicate overnight. The incumbent still has the advantage of being the default, the known quantity, the thing people reach for without thinking.
But the field is tilting. And it is tilting in a direction that has never been true before in the history of professional services.
For most of recorded history, access to expertise was rationed by wealth, geography, and connection. You got a good lawyer because you could afford the hourly rate, or because you knew someone, or because you happened to live in the right city. You got sound financial advice because you had enough money to be worth advising. You got real help navigating a complex system because you already understood how systems worked, which almost always meant you were already educated, already employed, already advantaged in ways that compounded.
The people who could not access expertise did not get worse outcomes because they were less capable of understanding their situation. They got worse outcomes because the tools to act on their understanding were priced somewhere they could not reach.
AI agents are not a perfect solution. They will not eliminate every advantage that wealth and connection confer. But they represent the most significant shift in who can access capability since the internet made information free. The internet made knowing things free. Agents are making doing things affordable.
That shift is not reversible. The cost of running these systems will continue to fall. The capability of what they can do will continue to rise. The number of people who understand how to use and build them will continue to grow. The problems that are solvable at real scale will continue to multiply.
The greatest concentration of opportunity right now is not in building a new AI model. It is not in competing with the large platforms. It is in the unglamorous, specific, deeply human work of finding a problem that real people have been quietly losing, understanding it more deeply than anyone else has bothered to, and building something that handles it for the price of a monthly subscription that costs less than what the problem currently costs people to absorb.
The complexity was always the product. The moat was always artificial. The expertise that felt out of reach was never truly beyond the people who needed it. It was placed somewhere they could not reach, by systems that depended on that distance to justify their existence.
That distance is closing. Faster than those systems would like. And the window in which genuine problem knowledge, without a large team or significant capital behind it, is sufficient to build something that matters: that window is open right now.
Not indefinitely. Right now.
The question is not whether this is real. It is. The question is not whether the moment is right. It is. The question is what problem you have been quietly carrying, absorbing, filing under things I have accepted, that you now have the tools to finally do something about.
That problem is the business. You already understand it better than anyone who might try to compete with you. The tools have arrived. The people who need what you could build are waiting, even if they do not yet know that help is possible.
The only question left is whether you are ready to stop accepting it.
If this resonated, pass it to someone who needs to read it. Subscribe for what comes next: a practical guide to assessing whether a problem is genuinely buildable, how to find your first customers before you have finished building, and what the path from idea to real revenue actually looks like. The opportunity is real. The path is learnable. And it starts with a problem you already know.
Craig Hepburn is an AI strategist and Perplexity Fellow who builds advanced agentic systems. He spent years leading digital transformation at Art Basel and UEFA. Now he works on the harder question: not whether organisations adopt AI, but how they govern it when they do.