The Manager & Agent Exclusion

Who may not share

The pooling page defined the inclusive edge — the chain of service, which in California reaches from the server to the kitchen. This page defines the exclusive edge, which is at least as important and far more frequently violated: the bar on agents. Section 351 forbids not only the employer but its agents from taking gratuities, and including an agent in the pool routes the employees' money into the supervisory tier — a taking, not a redistribution. The two edges work together: a lawful pool captures everyone in the chain of service and excludes everyone who is an agent, and a pool that gets the inclusive edge exactly right can still be unlawful if it gets the exclusive edge wrong by admitting a single agent.

The reason the agent exclusion deserves its own analysis is that it is where pools most often fail, and where the failure is hardest to argue away. Restaurants routinely include a shift lead, a working supervisor, or an assistant manager in the pool — often for understandable reasons, because that person serves customers, helps the team, and seems to belong. But if the person is an agent under section 350(d), inclusion is a taking regardless of how sympathetic the rationale, and the violation is typically systematic, affecting the pool every shift the agent works. This page sets out the agent definition, the core rule that an employee cannot be required to share with an agent, the narrow and contested exception some employers invoke, and the functional test for identifying agents in the roles a restaurant actually has.

Hiring, firing, or supervising — any of them

Section 350(d) defines an agent as every person, other than the employer, who has the authority to hire or discharge any employee or to supervise, direct, or control the acts of employees. The structure of the definition is disjunctive, and that is the first point to internalize: hire-or-fire authority is sufficient on its own, and, entirely independently, so is the authority to supervise, direct, or control the staff. A person need not have both; either makes the person an agent. This means the definition reaches well beyond owners and titled managers to capture any working supervisor, lead, or coordinator who in fact directs the staff — assigns stations, sets break order, instructs on service, corrects performance — even if that person has no power to hire or fire.

The second point is that the inquiry is functional, not titular, which cuts in both directions. A grand title does not make someone an agent if the person exercises no actual authority over other employees — a "lead server" who simply has seniority and a name tag, but who does not supervise or direct anyone, is not an agent and may share in the pool. Conversely, a modest title does not save someone who in fact exercises supervisory authority — a "senior server" who in practice runs the floor, assigns sections, and directs the other servers is an agent despite the unassuming label. Because the test looks to what the person actually does, the analysis requires examining the real allocation of authority in the restaurant, not the org chart, and the recurring trap is a working supervisor whose day-to-day authority over the staff makes her an agent even though everyone, including the employer, thinks of her as just another member of the service team.

Federal law reinforces this exclusion but does not define it as broadly. The 2018 Consolidated Appropriations Act bars managers and supervisors from keeping any portion of an employee’s tips regardless of any tip credit, yet the federal “manager or supervisor” definition tracks the executive-exemption duties test — a primary management duty, the regular direction of two or more employees, and hire-or-fire authority — and is therefore narrower than section 350(d)’s disjunctive reach. A working lead who directs and controls the staff can fall outside the federal definition while sitting squarely inside section 350(d), so satisfying the federal bar does not establish compliance with the broader California rule; section 350(d) governs, and it is the more demanding test.

Either hiring-and-firing or supervising-and-directing makes a person an agent — and the test is what the person actually does, not what the title says.

The core rule

The core rule that the agent definition serves was stated in Jameson v. Five Feet Restaurant: an employer violates section 351 when it requires an employee to share a tip given to that employee with the employer's agent. In Jameson, the restaurant required servers to share tips with floor managers who were agents, and the court held that this violated the statute, because it allowed the employer's agents to take a portion of gratuities that belonged to the servers. The principle is a direct application of section 351's text — no employer or agent shall take a gratuity — to the pooling context: the employer may redistribute gratuities among employees, but it may not use the pool to route a share to its agents, because that is the agents taking the employees' property.

Jameson supplies the rule that makes the agent definition operational, and it frames the violation precisely. The wrong is not that a supervisor happens to receive money; it is that the employer requires an employee to surrender part of a gratuity for the benefit of an agent. That framing matters because it identifies the structure courts scrutinize — a mandatory pool that channels employee tips to a supervisor — and it is the structure that the narrow exception in the next section is said to fall outside of. For planning purposes, the Jameson rule is the firm baseline: a restaurant designing a pool must keep agents out, because requiring the service staff to share with an agent is the paradigm section 351 violation, and the exposure runs to the diverted gratuities for the full limitations period across everyone in the pool.

Not the safe harbor it is taken for

The exception employers most often reach for comes from Chau v. Starbucks, which reversed an eighty-six-million-dollar restitution award and is sometimes read to mean that supervisors may share in tip pools. That reading is too broad and, used as a planning basis, dangerous. Chau turned on a specific structure: customers placed tips in a collective tip box that, the court found, they intended for the entire service team — baristas and the shift supervisors who also served them. Because there was no policy requiring a barista to surrender a tip given to her to a supervisor, the court reasoned, the Leighton-Jameson rule against routing an employee's tip to an agent did not apply; the employer was merely dividing a collective box equitably among the service staff the customers intended to tip. On those facts, the court held the shift supervisors could share even if they were agents.

Two features make Chau a poor foundation for including supervisors in a restaurant tip pool. First, it is narrow and fact-bound — it depended on a collective tip box left by customers for the whole team and on the absence of any policy redirecting an individual's tip to a supervisor, a structure quite different from a typical restaurant pool in which servers' tips are collected and reallocated. Second, it is in evident tension with Jameson, and commentators and the courts have noted the conflict; an employer relying on Chau is relying on a contested holding that a later court, or a different fact pattern, may not extend. The defense-prudent conclusion is that Chau does not license including agents in an ordinary tip pool, and a restaurant should not design its pool in reliance on it. The conservative posture — exclude anyone with genuine supervisory authority — is both clearly lawful and far cheaper than litigating whether a given pool fits Chau's narrow facts.

Barred, eligible, or fact-dependent

Each role is tested against the functional § 350(d) standard. Select a role to see whether the person is an agent barred from the pool, a non-agent who may share, or a fact-dependent case:

Fig. 1. Agent status by role under § 350(d). The test is functional — actual hire-fire or supervisory authority, not the title. Jameson; Chau (174 Cal.App.4th 688). The fact-dependent case is the working supervisor; the conservative course is to exclude anyone with genuine authority rather than rely on Chau's narrow facts.

The sympathetic supervisor in the pool

The agent-inclusion error is the most common tip-pool violation, and it almost always arrives wearing a reasonable justification. The working shift lead serves customers, hustles like everyone else, and is resented if excluded from the pool; the assistant manager covers the floor on busy nights and "earns" a share; the senior server who runs the section is the best server on staff. Each rationale is understandable, and each is beside the point, because section 350(d) asks only whether the person exercises hiring, firing, or supervisory authority — not whether the person works hard, serves customers, or deserves the money. If the person is an agent, inclusion is a taking, and the sympathetic facts that motivated it do not change the characterization. The error is also durable: once a supervisor is built into the pool, the violation repeats every shift, accruing across the limitations period for every employee whose tips were diverted to the agent.

The defensive answer is a conservative, functional audit of the pool's membership against the agent definition, conducted with the understanding that the close cases should be resolved by exclusion rather than by reliance on Chau. For each participant who has any supervisory dimension — a lead, a coordinator, a working supervisor — the question is whether the person actually directs or controls other employees or has hire-fire authority; if the answer is yes, or is genuinely uncertain, the safe course is to exclude the person from the pool and compensate the supervisory role another way, such as through wages or a bonus the employer pays from its own funds. Relying on Chau to keep a borderline supervisor in the pool trades a clear, cheap compliance step for the risk of an O'Grady-scale dispute over whether the pool fits a narrow, contested exception. The exposure and remediation page develops how to size and unwind a tainted pool; the discipline here is to keep agents out in the first place, and to resolve the doubtful cases against inclusion.

Identify agents functionally, exclude the close cases, and do not rely on Chau