In-house, agency, or AI: who should design your product
Build, buy, or generate is not a budget question. It is a question about where judgment has to live.
- Guide
- June 2026
- 6 min read
Choosing between an in-house team, an agency, and AI tools is a decision about where design judgment needs to live, not about which is cheapest. Generate the commodity parts, hire in-house for the problems you face daily and must own, and bring in a partner for the high-stakes decisions you make rarely.
Build, buy, or generate. Every team that makes something now faces the same three-way choice for design, and most of them frame it as a budget question: which is cheapest. That framing quietly produces the worst decisions, because cost is the one axis that tells you nothing about where the work will actually be good.
Cost is the wrong axis. Judgment is the right one
The useful question is not what each option costs, but where the judgment needs to live. Some decisions are low-stakes and reversible, and you just need volume. Some are the texture of your product, faced every single day. And some are rare, expensive, and hard to undo. Those three call for three different answers, and price is incidental to all of them.
Generate the commodity. Staff the recurring. Partner for the irreversible.
A decision rule that holds up
- Generate it with AI when the work is high-volume, low-stakes, and easy to reverse: variations, drafts, the hundredth banner. Speed matters more than authorship here.
- Build it in-house when the problem recurs weekly and you have to own it end to end: the product surface your users live in, the system that has to stay consistent as you ship.
- Partner with an agency when the decision is rare, high-leverage, and hard to reverse: positioning, a core identity, a flagship launch. You are buying judgment that has made this call many times, which you cannot build in-house for a decision you make once.
Read the stakes, then the frequency
Two questions settle most cases. How often will you face this, and how expensive is it to get wrong. High frequency pulls work in-house, because you need it close and you need to own it. Low frequency plus high stakes pulls it to a partner, because experience you will not accumulate yourself is exactly what you are renting. Low stakes plus high frequency is where AI earns its place. Map the need on those two axes and the answer usually picks itself.
The hybrid that usually wins
In practice the right answer is rarely one of the three. It is a stack. AI handles the volume underneath an in-house team that owns the daily craft, with an outside partner brought in for the handful of decisions that set the ceiling. The mistake is treating it as a single permanent choice instead of a portfolio you allocate, problem by problem, against where the judgment has to live.
Asked & answered
Decide by where judgment has to live, not by price. Generate high-volume, low-stakes work, staff an in-house team for the problems you face weekly and must own, and bring in a partner for rare, high-stakes decisions you cannot easily reverse.
No. In-house is best for recurring problems you own day to day. For rare, high-leverage decisions like positioning or a core identity, an outside partner who has made that decision many times is usually the better bet.