In-house teams know the business but are hard to scale. Agencies scale but sit outside. The model that works now: an embedded core with overflow built in.
For years the debate was framed as a choice. Bring marketing in-house and get people who know the business inside out, or use an agency and get specialists you could never hire yourself. Pick one.
Most of the teams we work with have stopped picking. They have worked out that the question is not in-house or agency. It is how to get the closeness of one and the flexibility of the other.
What in-housing gets right, and where it breaks
In-house teams sit in the meetings. They know the product, the politics and the priorities, and they do not need a brief for every task. That is real value, and no agency fully replaces it.
But an in-house team is sized for an average month, and no marketing calendar has average months. There is the launch, the rebrand, the peak season, the new market. Either you staff for the peak and pay for idle capacity the rest of the year, or you staff for the average and drown every time something big lands.
The specialist skills are harder still. Engineering, data, CRM and motion design are hard to justify as full-time roles for most teams, yet every team needs them some of the time.
Then there is headcount itself. In a lot of businesses, getting approval for a new role is harder than getting budget for the work. The money exists. The role does not.
Overflow, not overlap
The model I believe in is simple: an embedded core, with overflow built in.
- The core is a small team of people who work inside your business. They use your tools, join your stand-ups and know your brand. To your colleagues they feel like part of the team, because they are.
- The overflow is a wider bench of specialists behind them, on pre-agreed terms, for the peaks and for the skills you need some of the time rather than every day.
- The orchestration is someone owning how the two fit together, so work flows to whoever should do it, inside or out, without a new brief and a new quote each time.
The embedded people do not compete with your in-house team. They extend it. That is the difference between overflow and overlap.
Headcount without the headache
This model also solves the hiring problem. The capacity is bought as a service and sized month to month, so it grows when the work does and shrinks when it doesn't. No recruitment cycle, no notice periods, no restructure when priorities change. You get the team you need, rather than the team you could get approved.
We have worked this way inside fintech marketing teams since 2020, and the pattern holds: the embedded team runs the day to day, and the overflow absorbs the launches that would otherwise have broken it.
Making it work
- Embed for real. Your systems, your rituals, your Slack. A partner who only appears at the monthly report is not embedded.
- Agree the overflow before you need it. Pricing, turnaround and who can call on it, set in advance, so a busy week does not start with a negotiation.
- Keep the roles clear. Strategy, enablement and execution are different jobs. Know which one you are buying, and do not let one partner quietly blur all three.
- Measure utilisation, not vibes. Track where the hours go. Grow the core when the data says the work is steady, not because a quarter felt busy.
- Automate the grunt work. Every hour spent on manual reporting, QA checklists or moving content between systems is an hour not spent on judgement. That is where AI earns its keep.
Summing up
The best marketing teams I see now are small, embedded and technically literate, with elastic capacity behind them. Not a giant permanent department, and not an agency that sits outside the building. Keep the core close to the decisions, buy the peaks and the specialists as you need them, and let the machines take the busywork.
If that sounds like the team you need, here is how we do it.