This is a real decision with a defensible answer either way. Here is how we would think about it if we were on your side of the table.
Side by side
| Agency | In-house | |
|---|---|---|
| Time to competence | Immediate | 3 to 6 months of learning |
| Cost shape | Retainer, scales with spend | Salary, fixed regardless of spend |
| Cross-account learning | Yes, patterns from many accounts | Only your own data |
| Product knowledge | Has to be transferred | Already there |
| Risk | Dependency on a vendor | Key-person risk |
| Best when | Channel is new, volume is uncertain | Spend is large and steady |
When in-house wins
When your spend is high enough that a specialist salary is cheaper than a percentage-based retainer, when your product is complex enough that outside context is a real handicap, and when you have someone who can spend months learning a channel without other duties slipping.
When an agency wins
Early in a channel's life, which is exactly where ChatGPT advertising is now. Nobody has years of experience in it, so the differentiator is how many accounts a team has already tested it across. That pattern library is not something a single in-house hire can build alone.
The hybrid most brands land on
An agency runs the channel for the first two to four quarters while an internal owner sits alongside, learning the account. When volume justifies it, the work transfers in-house with documentation and the agency stays on for strategy. We build our engagements to make that transition possible rather than to prevent it.
Frequently asked questions
What does an agency actually cost?
Typically a monthly retainer, sometimes with a spend-linked component. Compare it against a loaded specialist salary plus the cost of the months spent learning, not against zero.
What should we keep in-house regardless?
Your data, your accounts and your tracking. Whoever runs the channel, ownership of the assets should never leave the business.