20 January 20268 min read
Every agency that has considered a delivery partner has run the same thought to its end: we introduce them to our client, they do good work, the client realises they could buy directly, and in eighteen months we are a line item nobody can justify. It is not a paranoid fear. It happens.
It is also mostly preventable, and the prevention is two-thirds behavioural and one-third contractual. Getting the clause right matters. Relying on the clause alone does not work, because a client who has decided you add nothing will eventually find a lawful way to act on that.
What the clause actually needs to say
The relevant protection has two halves that people conflate.
Non-solicitation covers people: neither party will hire the other's staff, or the client's, for a defined period after they have worked together. It should be mutual, it should name a period that a court would consider reasonable in the governing jurisdiction, usually somewhere between six and eighteen months after the engagement ends, and it should have a stated consequence, commonly a fee, because a clause with no consequence is a request.
Non-circumvention covers the commercial relationship: the partner will not contract directly with a client you introduced, for services of the type covered by the engagement, for a defined period. This is the one that addresses disintermediation, and it needs three things to be usable: a defined list of covered clients, a defined scope of covered services, and a defined term.
Two details are worth insisting on. Make the covered-client list a living schedule that is updated in writing when a new account is introduced, rather than a vague reference to clients of the agency, which is unenforceable and unfair in equal measure. And carve out the honest exception: if the partner already had a relationship with that client before you introduced them, they should say so at the point of introduction, not litigate it later.
Make it mutual, and mean it
Agencies often send a one-sided version and are surprised at the reaction. A partner is exposing its own people and its own methods to you, and has a reciprocal fear: that you will meet its engineers, learn how it recruits, and build the same capability yourself next year having used its business as a training exercise.
A mutual clause is easier to sign, easier to defend, and sets the tone for everything that follows. It also protects you in the direction you were not thinking about, which is your own account managers taking a partner's people with them when they leave.
IP and confidentiality across three parties
Ownership of the relationship is one problem. Ownership of the work is a separate one, and it is where the paperwork most often has a hole.
The chain has to be unbroken. Rights in what an individual creates must vest in the partner, whether by employment terms or a contractor assignment; the partner must assign to you; and you must assign to the client under your own agreement. If any link is missing, the client does not own what they paid for, and the discovery usually happens during due diligence on their funding round, at the worst possible moment and with your name attached.
Confidentiality flows the same way, and it flows down as an obligation rather than up as a promise. Whatever you committed to the client, you must have committed to the partner in writing. Where personal data is involved, that means a data processing agreement between you and the partner that is at least as strict as the one between you and the client, naming the same purposes and the same sub-processors. Some client contracts require the client to approve sub-processors by name; if yours does, the partner is one, and skipping that notification is a breach even when nothing bad happens.
The part the contract cannot do
Clauses expire, and they only bind the parties who signed them. What keeps you in the relationship is being genuinely necessary, which comes from doing things the partner does not do.
You hold the commercial relationship: the contract, the pricing, the negotiation, the invoice, the credit risk. You hold the domain understanding: what this client's business actually needs, which is the thing a delivery team rarely acquires and never acquires quickly. You hold direction and prioritisation: deciding what gets built next and why. And you hold accountability, which is what the client is really paying a premium for.
A client leaves for the delivery partner when the delivery partner is the only thing they can see. The remedy is to be visible, not to be litigious.
The visible version of that is unglamorous: your lead runs the client meetings, your name is on the reporting, the strategic conversation happens with you, and the client's questions get answered by you even when the answer came from the partner. None of that is deception. It is an accurate reflection of who is accountable.
Habits that create the risk
Three ordinary conveniences quietly hand the relationship away.
- Letting the client and the partner run a shared channel that you are in but do not lead, until the daily working relationship is theirs and yours is ceremonial.
- Passing through partner reporting unedited, which tells the client that your contribution is forwarding email.
- Stepping out of technical conversations entirely, so the only person who can discuss the system with the client works for someone else.
None of these are worth solving by keeping the partner hidden or forbidding direct contact. Enforced separation makes delivery slower and worse, and clients notice. The fix is presence, not walls.
If it happens anyway
Sometimes a client approaches the partner directly. The partner's response in that moment is the best information you will ever get about them, and it is worth agreeing in advance what it should be: tell you immediately, decline to discuss commercial terms, and refer the client back. Ask a prospective partner what they would do in that situation before you sign anything. The ones who have thought about it will have an answer ready, and the answer will be that one.