How to start an agency after a career in tech

To start an agency after a career in tech, pick one type of client and one outcome you've already delivered. Sell it as a fixed, productized offer before you take custom work, and set up delivery so a small senior team and written processes do the work. Win your first clients from people who know your work, and price for a margin that still holds once you're paying other people. That last part is the one I'd check twice.

What kind of agency should you start?

Four agency models work with a small senior team: the niche specialist, the productized agency, the AI automation agency and the studio. They differ in what you sell, how you price it and where the margin comes from.

A boutique agency is a services firm with a small senior team, often under ten people. It sells a narrow set of services to one type of client and runs the work on documented processes and contractors.

Pick one model and stick with it, because each needs its own offer page, sales pitch and delivery process. An agency suits people who enjoy designing how work gets done and are happy to hand the doing to others. In my Builder Archetypes, that's the Orchestrator. If you'd rather hold one leadership role inside a company for part of the week, read how to go fractional instead.

How do you pick a niche for an agency?

Pick a niche by pairing a client type with an outcome that client already pays for, like paid acquisition for Series A B2B SaaS companies in DACH. "Marketing agency" on its own is too broad to count.

My rule: choose the client type before the channel or the tool. Clients in the same space talk to each other, so referrals travel and one case study sells the next.

Narrow also pays better. Promethean Research's Digital Agency Industry Report (2026, based on surveys of 1,452 agency leaders) found that agencies that cut their services in 2025 grew almost twice as fast as the average and averaged 30% net margins. Agencies that added services averaged 10%.

I'd test a niche when most of these are true:

• You've delivered the outcome more than once, in a job or for clients.
• You can name 30 companies that fit and reach at least 10 people inside them.
• The buyer already spends money on the problem: another agency, a tool or a hire.
• Something makes them buy now, like a funding round, a new head of growth or a migration deadline.
• Your tenth client will look enough like your second that the process carries over.

If you're still choosing between angles from your career, start with what to sell after a tech job.

What should a new agency sell first?

A new agency should sell one productized offer first: a defined result, for a defined client, in a fixed time, at a fixed price.

A productized service is packaged like a product, with the same scope, steps and price for every client. You can write it down, train a contractor on it and know your margin before you start. That's why I'd never open with custom work.

The shape I like is an entry offer that leads into ongoing work. Say you run tracking and attribution for e-commerce brands. The entry offer is a three-week tracking audit at a fixed price that ends with a prioritised list of fixes. The core offer is a monthly retainer to implement and run those fixes.

Before you sell it, write down the outcome in the buyer's words, what's included and what isn't, what the client must provide, the timeline and the price. Alex Hormozi's advice in $100M Offers helps here: list what could stop the client getting the result, then build fixes for the likely ones into the offer. The Offer agent in The Stack runs this for a services business and writes it up as one paragraph you can put on a page.

Take custom work once the core offer has sold several times and someone other than you can deliver it.

How should you price agency work, and what margin should you aim for?

Charge a monthly retainer for ongoing work with defined outputs and a fixed project fee for work with a clear end. Add a performance fee only when the result is measurable and your work clearly causes it.

• Retainer. A fixed monthly fee for a defined scope of ongoing work. Predictable revenue lets you commit to contractors. Define it by outputs per month, because clients start questioning a retainer the moment they can't see what it produces.
• Project fee. One price for a defined result, paid in milestones. Keep hours out of the proposal, so every time your process gets faster, your margin goes up. Expect uneven revenue between projects.
• Performance fee. Part of your pay tied to a result such as revenue or qualified leads. Put it on top of a base fee. On its own it invites arguments about attribution, and Promethean notes it tends to make an agency look like a vendor.

Most agencies mix these: Promethean found that 8% or fewer rely on a single pricing model. The report also warns that agencies billing time and materials are exposed to AI, because when the same output takes fewer hours, the invoice shrinks. I wouldn't start an agency on hourly billing in 2026.

Blair Enns, in Pricing Creativity, argues for showing several options at once, at different scopes and prices. The Pricing agent in The Stack uses his approach for services. For day rates and value pricing in euros, see how to price consulting services.

Margins: a worked example

Gross margin on a client is the fee minus the direct cost of delivering it. Net margin is what's left after every cost, overhead and tax included. In Promethean's report, the average digital agency earned a 13% net margin in 2025, and studios with fewer than 10 full-time staff earned more than twice the average margin of agencies with 50 or more.

Here's an illustrative client. Say you charge 5,000€ a month for a lifecycle email retainer:

• Senior contractor, 3 days at 600€: 1,800€
• Production contractor, 4 days at 250€: 1,000€
• Tools and AI usage: 150€
• Your own review and client time, 1 day valued at 800€: 800€

Delivery costs 3,750€, which leaves 1,250€ of gross margin, or 25%. Sales, admin, insurance and software still come out of that, so it's thin.

Now say templates and a review checklist cut the senior work to 2 days and your time to half a day. You raise the fee to 6,500€ because the offer now includes a monthly test plan. Delivery costs 2,750€ and gross margin is 3,750€, about 58%. One rule I'd never bend: count your own delivery time at a market rate. Leave it out and the numbers only work while you work for free.

How do you deliver without doing all the work yourself?

Write down how the work gets done before you hire anyone. Then hand steps to senior contractors one at a time, and keep the kickoff, the review and the client relationship yourself.

Build the delivery system

Deliver the first two or three clients yourself and document as you go. You want an intake form and onboarding checklist, a written process for each step with a short screen recording, templates for everything you hand over, a quality checklist someone else can use, and a fixed format for client updates. Keep every client's work on one project board. I led 160+ people at Too Good To Go, and I'm convinced of one thing: nobody can deliver well from instructions that live in your head.

First hires and contractors

Your first person should be a senior contractor who can deliver the core offer from your documents with light review. A cheap junior who needs you at every step adds cost and saves you no time, so I'd pay more here.

Your second person may not work on delivery at all. Promethean describes the founder of a studio-sized agency as its only seller, so decide early who owns the pipeline.

Hire employees once the workload has been steady for months. Until then, contractors keep your costs flexible. In many European countries, a contractor who works set hours only for you, under your direction, can be reclassified as an employee, so check with a local accountant or lawyer first.

How do you get your first agency clients?

Your first agency clients will most likely come from people who have already seen your work: former colleagues, managers, founders and vendors. Promethean reports that most agencies still get the bulk of their leads through referrals, and that the average agency spends 7% of revenue on sales and marketing.

1. List 50 to 100 people who know your work. Mark the ones at companies in your niche, and the ones whose company is in a buying moment right now.
2. Message them about the specific problem your offer solves and ask for a conversation.
3. Sell the entry offer first. A fixed-price audit is a much easier yes than a year-long retainer.
4. After each project, write a short case with the problem, what you did and the result in the client's numbers, with their permission.
5. Publish the thinking you'd sell. David C. Baker argues that experts should publish their point of view, and a buyer who has read yours already knows how you'd work. This is the step I'd start earliest.
6. Find partners who meet your clients first: software vendors whose tools you implement, investors' portfolio teams, agencies that don't offer your service.

What does AI change about agency delivery?

AI shortens hands-on work like first drafts, analysis, code and design options. So clients expect to pay less for the doing and more for strategy and for changing how their work runs.

Promethean's report says a third of the agency industry had fully implemented AI by March 2026, 70% of agencies changed their service mix in 2025, and some recent margin pressure comes from clients expecting cheaper services because of AI. It also sees new demand for workflow redesign and automation. That's where I'd point a new agency.

Clients are adopting it unevenly. Eurostat reported that 20.0% of EU enterprises with 10 or more employees used AI in 2025, up from 13.5% in 2024. Its detailed statistics put the figure at 17% for small enterprises and 55% for large ones, with marketing or sales the most common use.

An AI automation agency designs, builds and maintains automated workflows that combine AI models with a client's existing tools, for jobs like lead routing, support triage or document processing. If you start one, plan for three things:

• Maintenance. Tools, APIs and models change. Sell a monthly maintenance fee with every build, or you'll be fixing things for free.
• Client data. If your workflows process personal data for an EU client, you're usually a processor under GDPR and need a data processing agreement with them.
• Pricing. A workflow you build in two days can save a team hours every week. Set a fixed fee based on that saving, and automate the same process for the same kind of company so you can reuse components.

What mistakes turn a new agency into a freelancer with overhead?

A new agency becomes a freelancer with overhead when the founder still does most of the delivery and now also pays for a company, contractors and software. My test is simple: take two weeks off and see whether client work still ships.

• Scope creep. This is work added after the price is agreed, with no change to price or timeline. Prevent it with an out-of-scope list in every proposal and a priced change request for anything new.
• Saying yes to every service. Each new service needs its own process, templates and people. The 30% versus 10% margin gap above is what that costs you.
• Building delivery before demand. Processes designed for ten clients don't help while the pipeline holds one. Spend the first months selling, and document as you deliver.
• Taking any client who pays. Clients outside your niche break your process and give you case studies your niche ignores.
• Selling hours. Hourly billing caps revenue and hands every efficiency gain, AI included, to the client.

Questions people ask