Niche down or stay a generalist in the AI era?

Niche down, with a margin. In 2026, broad and routine work is the work AI does cheapest, so breadth no longer protects you the way it once did. The safer shape is deep in one domain and competent in the skills next to it. A niche that is too narrow has its own risks, covered below.

Why staying broad used to make sense

The generalist case was not wrong for its time. When the main danger was an uneven calendar, breadth protected you. A designer who also wrote copy, ran a little social media and built simple sites had four doors instead of one. A slow month in one was offset by another.

Breadth also matched how small clients buy. A small business does not want four contractors. It wants one capable person who can handle the website, the newsletter and the brand. For about twenty years, "be the versatile one" was reasonable advice.

It assumed that the floor under routine work was stable. That is the part that moved.

What AI changed, according to the data

AI did not lower demand evenly. It hit routine, easily substituted tasks first.

The Oxford Internet Institute analysed more than three million job postings from a global freelancing platform. Demand for substitutable skills such as writing and translation dropped by 20 to 50 percent. Chatbot and natural language jobs almost tripled over the same period, and net demand for freelance work rose, according to the Oxford Internet Institute. So the market grew and shifted at once.

A Brookings study using Upwork data found that freelancers in AI-exposed occupations saw about a 2% fall in contracts and a 5% fall in earnings. The declines kept growing over the months studied. Its surprising finding was that the higher-skilled, more experienced freelancers were hit hardest, as Brookings reports. The authors suggest AI levels the field: when a tool can produce passable routine work, the premium for doing that routine work well shrinks.

That is the inversion. Being a better generalist at commoditised tasks stopped being worth extra. The broad bundle of routine skills that once hedged your income is now the bundle AI replaces fastest.

The squeeze is not limited to independent work. SignalFire found that new-graduate hiring at big tech fell 25% from 2023 and by more than 50% from 2019, in its State of Talent report. A Stanford Digital Economy Lab paper found employment of workers aged 22 to 25 in AI-exposed jobs now stands 19% below where it would have been, and calls it a descriptive early indicator rather than proof of cause, in Canaries in the Coal Mine.

What the demand side says

Upwork's In-Demand Skills 2026 report found demand for the top skills tied to applying AI inside existing roles up 109% year over year, with AI video generation and editing up 329% and AI integration up 178%. It also found established skills such as graphic design, full stack development and data analytics staying steady, per Upwork's release.

Read that carefully. It does not say generalists are finished or that every established skill is safe. It says the growth sits in applied, specific work, and that clients still pay people who bring judgment on top of a tool.

Four ways to niche

"Niche down" has become hollow advice, so be precise about which axis you mean.

1. A skill niche goes deep on one capability. Not "design" but conversion-focused landing page design. Not "writing" but documentation for developer tools.
2. An industry niche picks a sector. A marketer who works only with dental practices already knows the language, the rules and the customers.
3. A client-type niche picks a kind of buyer: early-stage startups, solo founders, companies past a certain size. The work varies. The buyer's situation does not.
4. A problem niche is built around an outcome: reducing churn, fixing a slow site, getting a clinic's bookings under control.

Industry, client-type and problem niches tend to be the strongest, because they run on context, and a general model does not have context about your client's specific world. A pure skill niche is more exposed, since a skill is closer to the kind of thing tools can approximate. The best positioning usually joins a skill with a context niche: one capability applied to one world.

Why depth works as a moat

AI is good at tasks that are common in its training data and loosely specified: a generic blog post, a passable logo, a standard email. It is weaker where work depends on deep, specific context, such as a particular industry's regulations, a type of client's hidden constraints, or the judgment that comes from solving the same narrow problem a hundred times.

A deep niche is a store of that context and judgment. That is why a client will pay more for someone who has seen their exact problem before. It also changes how you sell. A specialist can quote a fixed package for a known problem. A generalist ends up quoting hours against a vague request, which is the easiest kind of work to compare on price. The guide on productized services shows how that works in practice.

The case against niching too hard

This is where the argument has to push back on itself. A niche that is too narrow is fragile in a different way. If you are the expert in one platform and it fades, your moat drains with it. If you serve one micro-industry and it has a bad year, so do you. Hyper-specialising trades an uneven calendar for concentration risk.

The shape that holds up is deep in one thing and competent in the things adjacent to it. Be clearly the specialist in a domain. Keep enough range to follow a client problem one or two steps sideways, and enough adaptability that you are not married to one tool. The depth is your moat and your pricing power. The range around it is insurance. For a list of areas where human context still wins, see these AI-resistant niches.

How to test a niche before you commit

Many people stay generalists because they treat a niche as irreversible. It is not. Treat it as an experiment.

1. Look backward. Which past projects did you enjoy most, do best and get paid fairly for? A niche you already have evidence for beats one that only sounds lucrative.
2. Change the first sentence of how you describe yourself, on your site, your profile and your discovery calls.
3. Write two or three pieces aimed squarely at that niche.
4. Keep serving your current clients. Niching changes your positioning and your pipeline, not your obligations.
5. Judge it after a few months, not a few weeks. The signal is whether you get better-fitting inquiries, not more of them.

If the niche pulls in better conversations, lean in. If not, you lost little and learned something. The real risk is choosing none and staying discoverable only as "available for anything", which reads as "the work AI does for free."

How Delivvo supports a specialist practice

A specialist needs the process around the work to look as focused as the pitch. Delivvo runs that process for service businesses, so your niche shows up in how a client experiences you from the first message.

Here is what that looks like once you have picked a niche:

1. Build an inquiry form with questions only your niche asks. The form builder has 26 field kinds and takes up to 40 fields, so the form itself shows the client you know their problem.
2. Inquiries land on your lead board as New, then move through Contacted, Quote sent, Won or Lost, so you can see which niche message actually produces work.
3. Turn your offer into fixed packages and add-ons in Proposals. A known problem with a fixed scope sells better than an hourly estimate.
4. The client opens one link, picks a package, signs the contract and pays the deposit.
5. Delivery and approvals happen in a client portal with no password, branded with your logo and colour.

One detail a competitor page would not offer: the client portal, forms, quotes, emails and PDFs all carry your branding. Clients pay through your own gateway and Delivvo takes 0%. If you later change niche, you change the form, the packages and the portal colours, and your old clients and projects stay where they are. Nothing is rebuilt from scratch, which keeps the test from step one of the list above cheap. Look at the inquiry form builder and Proposals for the details.

Delivvo is a CRM and client portal for service businesses. See Delivvo for teams.

FAQ