Playbooks
Practitioner playbooks on running customer service in the real world. Written from twenty years on the floor — contact centers, biopharma support, healthcare SaaS — not twenty years in a vendor boardroom. Built for the team of one to five who got handed CS and is trying to make it work.
During an outage your silence is data, and customers fill it with the worst plausible story. The three-sentence first message inside fifteen minutes, the cadence you keep even when empty, the grammar of honest not-knowing, and the postmortem that ends with more trust than uptime bought.
“Great idea, adding it to the list” is deferred honesty with interest. Splitting the problem from the proposed solution, the five-part honest no, the one-email script, the churn-warning tells, and the log that makes “it shipped” possible.
The first reply to a furious customer has one job, and it isn’t the fix — it’s proof somebody is listening. The ten-minute rule, the four-sentence structure, the three moves that pour gasoline, and the rails that make it safe to delegate.
If every refund needs the founder, you don’t have a policy — you have a bottleneck with feelings attached. Three questions that decide almost every case, a goodwill budget with a real number, and a refund log that tells you what your product is costing you.
Customers report feelings, not defects — and that's fair; precision was never their job. The extraction craft: the five facts a report needs, the two-round rule, reproduce-before-you-route, and the one-bug-one-record ledger that turns five vague complaints into one weighted fix.
Small teams copy enterprise response promises, miss them within a month, and train customers not to believe anything they say. Build SLAs backwards from real capacity instead: honest tiers keyed to impact, published windows with published hours, automation that buys time honestly, and a breach protocol for the weeks it goes wrong.
A generic 1-5 CSAT tells you the room is basically fine and hides the exact customer about to leave. The sharper post-resolution question, the three individual-level signals that actually predict churn — effort, the unresolved feeling, repeat contacts — and the weekly ritual of closing the loop on the detractors.
Your customers hit send at 11pm, midnight, Sunday morning — and the response clock starts then, not when you wake up. How to cover the after-hours queue without an overnight hire or a reckless bot: acknowledge instantly, auto-handle the bounded 60%, queue every judgment call for morning.
A third to half of your tickets are questions you've already answered. You don't have a content problem, you have a capture problem. The workflow that turns the support you're already doing into a self-serve KB — stop authoring, start harvesting.
The fear with AI support isn't that it answers tickets — it's that it confidently mishandles the one that mattered. The triage line: a stakes-by-confidence framework for what to auto-handle, what to draft for approval, and what must reach a human every time.
A template system your customers can't detect: the 70/30 rule, the three human seams, the five-template core library, quarterly rot review, and where AI fits — from twenty years in contact centers.
The all-caps email landed and your instinct is to defend yourself — that's the mistake. The de-escalation framework from twenty years running contact centers: acknowledge before you solve, the six moves that calm a furious customer, and the polite lines that quietly make it worse.
The actual operating model for solo customer service: 24-hour backlog rescue protocol, five systems to install, honest read on what AI does and doesn't do for the team of one.