Build, operate, or partner: choosing the right engagement.
Three structures, three different distributions of risk. Picking the wrong one is the most common reason good work fails.
Most failed engagements are not execution failures. They're structural ones: a fixed-scope project bought by a company that actually needed an operator, or a retained relationship sold to a company that just wanted a thing built. The work was fine. The container was wrong.
There are really only three structures, and they differ in one variable — how much of the outcome the vendor is holding.
Build — you hold the outcome.
Fixed scope, defined deliverable, clean handover. Right when you know what you need, you have a team that can run it afterwards, and the value is in the artefact rather than in its ongoing operation.
It goes wrong when the buyer doesn't actually have the team to operate what gets built. A beautifully engineered system with nobody to run it decays faster than a mediocre one with an owner.
Operate — the vendor holds execution.
Retained. The vendor runs what they built and reports against outcomes rather than activity. Right when the system needs continuous judgement — media buying, content operations, anything where last week's decision changes this week's.
It goes wrong when the reporting is activity-shaped. If the monthly deck is a list of tasks completed rather than a number that moved, you're paying for motion.
Partner — the risk is shared.
Equity, revenue share, or build-operate-transfer. The vendor's upside depends on the operating result. Right when you need capability you can't hire quickly and you're willing to trade economics for alignment.
It goes wrong when the vendor doesn't have real control over the outcome they're being paid on. Shared risk without shared authority is just a discount with extra steps and a worse ending.
How to choose in one question.
Ask: if this fails, whose problem is it? If the honest answer is yours, buy Build and make sure you can operate it. If it's genuinely shared, structure it as a partnership and give the partner enough authority to earn it. If you want it to be someone else's problem entirely, that's Operate — and you should be paying accordingly and reading a number, not a task list.
Any firm worth engaging will tell you when you've picked the wrong one.
- What an AI development company should actually deliverMost AI engagements die between the demo and the deploy. Here's the checklist that separates a vendor who ships from one who presents.
- Your attribution is lying to youPlatform-reported conversions are a marketing document, not a measurement. What to build instead, and roughly what it costs.
- Publishing AI content without wrecking your rankingsThe difference between scale that works and scale that gets you deindexed is not the model. It's the gates.