Splitting the account manager and delivery lead roles

An account manager owns the client's experience of the relationship: what gets communicated, what gets committed to, what happens when the client is unhappy. A delivery lead owns whether the work is right: sequence, dependencies, quality, and the unglamorous business of telling people their estimate was wrong.

In a five-person shop the same person does both, and that is correct. The overhead of a handoff would cost more than the conflict creates. The problem is that nobody notices when it stops being correct.

The conflict is specific and it always resolves the same way. When the delivery answer is unwelcome and the relationship is under strain, a person holding both jobs will soften the delivery answer. Not dishonestly. They will find the version of the truth that keeps the call pleasant, because they are the one who has to be on the next call too.

That is not a character flaw, it is the predictable result of putting both incentives in one head. The split exists to make the delivery answer come from somebody whose week does not get worse when the client dislikes it.

Do not look for a headcount threshold. Look at hours, and use your own calendar rather than a rule of thumb.

For each account, measure two figures over a typical month. Call C the client-facing hours: scheduled calls, the email that follows them, the reporting conversation, the informal check-ins, the escalations. Call D the delivery-coordination hours: sequencing, unblocking, chasing dependencies, reviewing work before it goes out, and rewriting the plan when something slips.

Across a book of N accounts, one combined person needs the sum of C plus D for every account, plus the fixed internal load that is not attached to any client. Most people find the total exceeds a working week somewhere between four and eight accounts, but the number is yours and it depends entirely on how heavy each account is.

The failure signal is more useful than the total. When C and D compete in the same week, one of them loses, and it is nearly always D, because C has a person on the other end of it asking. Delivery decisions get made late, by default, or not at all. If your delivery slips cluster in the weeks with the most client meetings, you already have the answer.

The common mistake is to split by rank: the senior person keeps the client, the junior person runs the work. That produces a delivery lead who cannot say no, which is the same failure with an extra salary attached.

Split by which decision each role is allowed to make alone. Atlassian's responsibility assignment matrix guidance is explicit that a task should have exactly one accountable individual, and that accountability should not be delegated. Applied here, it means every recurring decision on an account has one named owner, and that owner is not always the person the client speaks to most.

The related discipline is separating decision-making from role definition. The DACI framework exists to name who drives, approves, contributes and is informed on a decision, which is a different question from who does the work. Atlassian's own roles and responsibilities play runs the exercise as a workshop rather than a memo, which matters, because a split announced by email is not a split.

The wider version of this discipline is covered in a RACI that survives a real project. What is specific to the account manager split is that one of the two roles is commercial and one is not, and the boundary has to be drawn where money starts.

Vagueness here is expensive, because the two people will discover the boundary during an argument in front of a client. Write it down before that happens.

The clean line: the account manager owns anything that changes the commercial relationship, and the delivery lead owns anything that changes the work. Price, scope commitments, renewal, and escalation are commercial. Sequence, estimate, quality gate, and go or no-go on a release are not.

The two genuinely shared items are the date and the bad news. A date is a delivery fact with a commercial consequence, so it is set by the delivery lead and communicated by the account manager, and neither of them changes it alone. Bad news follows the same rule, which is why telling a client you will miss the date needs a written protocol rather than a judgement call.

Service commitments belong in the same document. An SLA is a plain-language agreement defining the services delivered, the responsiveness expected, and how performance is measured, which makes it the natural place to record which role answers which kind of request.

Two people on an account is more expensive than one, and the cost is not only salary. It is the handoff, the duplicated context, and the client learning two names instead of one.

Price it before you decide. If C plus D for an account is twelve hours a month and the handoff overhead is two, you have added roughly seventeen percent to the account's unbillable load. On a large account that is trivial. On a small one it can be the difference between profitable and not, which is the same calculation behind setting a minimum engagement and holding it.

That gives you a sensible middle position. Split the roles on the accounts where the arithmetic supports it and keep them combined on the rest. A partial split is not a failure of nerve, it is the correct answer for a mixed book.

The pattern is visible in programs that already run this way. A migration with 154 data sources and three parties holding pieces of the plan needed a delivery lead whose entire job was sequencing and coordination, separate from whoever owned the commercial relationship, because no single person could have held both at that size.

Introduce both people to the client at the same time, with the boundary stated out loud. A delivery lead who appears three months in, after something went wrong, is read as an escalation rather than a structure.

Then enforce it in the small moments. When a client emails the delivery lead about price, the delivery lead forwards it rather than answering. When a client asks the account manager for a date on a call, the account manager says they will confirm it. Two weeks of that establishes the boundary better than any kickoff slide.

The failure mode to watch for is the account manager who keeps a private version of the plan. If both people are maintaining separate views of the same commitments, you have not split the role, you have duplicated it. One plan, two owners of different decisions about it.

Where an agency does not have the second person internally, the delivery half is the half that transplants cleanly, because it is defined by artifacts rather than by relationship history. That is the shape of fractional technical project management: an external owner of sequence, dependencies and the quality gate, with the client relationship staying where it already sits.

Take your last three unwelcome delivery messages to clients. For each one, write down the date the team knew and the date the client knew. If the gap is more than a few days and the person who knew was also the person who manages the relationship, the conflict is already costing you.

Then take your calendar for the busiest week of last month and mark every hour as client-facing or delivery-coordination. If coordination is under a fifth of the total, the work is not being sequenced by anyone. That is the argument for the split, and it is made from your own data rather than from a benchmark someone published.