Run support for every client — under their brand, not yours

If you run an agency — e-commerce builds, performance marketing, web development — you already operate most of your clients' digital presence. Their store, their ads, their email flows, their landing pages. And then there's support: the one function clients keep asking you to take on, and the one most agencies either decline or do badly through a shared Gmail inbox with no records, no metrics, and no brand.

Declining is leaving money on the table. The Gmail version is worse: unbillable chaos that makes your agency look exactly as improvised as it is. Here's the third option — running support as a productized, branded, reportable service line — and the mechanics that make it work.

One client, one workspace

The foundation is isolation. Each client gets their own Iris workspace: their own conversations, their own contacts, their own help center, their own settings, their own reports. Nothing bleeds between clients — not data, not branding, not team access. If a client ever leaves, their workspace is a clean, self-contained thing.

This sounds obvious, but it's exactly what the shared-inbox-with-labels approach fails at. Labels leak. Metrics blur. And you can never hand a client a report that says "here is your support this month" when their tickets live interleaved with four other companies'.

Creating a workspace is self-serve — a signup form, not a sales process — so provisioning client number six works the same way as client number one.

Their brand on every surface the customer sees

The whole point of the service is that the client's customers never encounter your tooling — or your agency. Three surfaces, all whitelabel:

The messenger. Their logo, their brand colors, their welcome copy, light or dark to match their site, the launcher positioned where their design wants it. The "powered by" line can be turned off. What their customers see is their support widget, indistinguishable from something they built.

The help center. Published articles become a public, SEO-indexable help site on the client's own subdomain — help.theirbrand.com, connected with a single CNAME record from a settings screen. Their customers google "theirbrand refund policy" and land on their domain, not yours and not ours.

Email. With the email channel, the client's existing support address forwards into the workspace, and replies go out from support@theirbrand.com — their domain, properly authenticated, via a few DNS records you publish once. You're usually already managing their DNS; this is a ten-minute change with a verification poll that goes green on its own.

Behind all three surfaces: your team, in one place.

One login for your whole team

The operational trick that makes multi-client support sane: your account managers sign in with one email. If that email is an agent on several client workspaces, Iris asks which workspace to open — a picker, not a password-per-client spreadsheet. Your team hops between Client A's queue and Client B's queue without juggling credentials, while each client's workspace only ever contains its own people and data.

Onboarding a client in an afternoon

The realistic checklist, with honest timings:

1. Create the workspace — 5 minutes.
2. Branding — logo, colors, theme, welcome message, launcher position — 20 minutes with their brand assets on hand.
3. Install the snippet — two script tags on a site you already manage — 10 minutes including deploy.
4. Seed the help center — 15 articles from their existing FAQ page, policy pages, and the questions their old inbox answers weekly. This is the real work: about 90 minutes of pasting, tightening, and publishing.
5. Connect email + help domain — publish the DNS records, start the verification — 20 minutes of your time, plus DNS propagation on the calendar, not the clock.
6. Turn the AI on. Every published article is immediately part of what the AI agent can answer, in the client's brand voice surface. Where it isn't confident, it stays silent and the thread routes to your team — which is precisely the behavior you want wearing a client's logo.

Total: one focused afternoon per client, most of it step 4. And step 4 is also where the deflection comes from, so it's the afternoon that pays for itself.

Pricing it into the retainer

Two numbers: a setup fee for the afternoon above ($300–500 is defensible for the audit-and-seed work alone), and a monthly line item for operating it ($300–800 depending on volume and hours of coverage).

The margin math is what makes this a service line rather than a favor. A small D2C client generates maybe 150–250 conversations a month. With a seeded help center, the AI resolves the recurring majority — order status, returns policy, sizing, shipping times — and your team handles what's left: perhaps 30–50 conversations a month needing human judgment. At a few minutes each, that's 2–4 human-hours per client per month against a recurring fee. It stacks: five clients is one part-time queue for one account manager, not five part-time jobs.

Two rules keep the margin honest. First, watch the knowledge gaps list in Insights weekly — every gap you turn into an article converts future human minutes into AI resolutions, which is your margin compounding. Second, don't promise "the AI answers everything." Promise coverage and first-response times. The AI's restraint — silent when unsure, human when it matters — is a selling point to any client who's been burned by a chatbot that improvised their refund policy.

The monthly report writes itself

The retainer's renewal case is the Reports screen: conversations handled, the resolution funnel, first-response and time-to-close medians, CX score — scoped to that client's workspace, because everything is. Export the story monthly: "412 conversations, 61% resolved automatically, median first response 4 minutes, CX 4.6."

That's a paragraph no shared Gmail inbox has ever produced, and it's the difference between a client wondering what the support line item is for and a client asking what else you can run for them.