COS Interview Playbook

0. Speaking Scripts

Double Service — when she asks about experience

At Double Service, a B2B SaaS serving twelve municipal clients, I focused on turning messy customer problems into work that actually got resolved. Issues came in scattered across group chats, emails, and screenshots. I pulled each one into a single record — status, priority, evidence, client impact, and next action — and issue traceability went from near zero to about ninety percent. Then I made every handoff evidence-ready, with repro steps and checks attached, so engineers could start with usable context instead of rebuilding it from scratch; average resolution time dropped about forty percent. And when I noticed recurring needs across clients, I turned them into six reusable specs the team could build from. That's what I'd bring here: catch the small thing before it becomes the big thing, follow it through, and turn what repeats into a fix.

Migration — when she asks how you'd handle it (her biggest hiring risk)

I haven't led a customer platform migration before, so I want to be clear about that boundary. My closest experience is mapping and validating more than 4,200 records from Sheets into PostgreSQL, and tracking customer issues through status, exceptions, and next actions. At COS, I'd treat migration as an account-by-account customer transition. First, I'd clarify what completed means. Then I'd group accounts by readiness and risk, give each customer a clear next step, and track blockers, owners, and follow-up. I'd start with a small batch, reconcile the outcomes, and scale from there. Any recurring blocker should become an FAQ, a training step, or a product flag so the next migration gets easier.

AI — when she asks how you work or stay organized

I use AI in three steps: delegate, refine, and gate. In my application review and CRM handoff workflow, AI assembles source context and flags missing or duplicate information. When it misses something, I add that exception to the SOP so the next review gets better. Before anything goes out, I verify it against the original source; AI never sends on my behalf. At COS, I'd use the same loop for migration tracking and quiet-customer follow-up. AI helps me move faster and catch more details, while I remain responsible for the judgment and final action. In a business where a small mistake can mean real legal consequences, AI should be an extra layer of checking — never one less.

1. JD — Read It from Victoria's Seat

One line: She spent two years rebuilding the platform, and the company's next phase is riding on the migration. She's hiring someone who can take the first line of support and the migration off leadership's plate — with reliable follow-through.

Her three fears — every answer should land on at least one

Tone: warm and steady, not a process robot — she wrote "treats people like people, not tickets." Talk about the customer's situation first, then how you close it out.

Three memory points (what she should remember after the call)

1. This candidate understood the migration is our bet — not a generic CS opening.
2. She has real, numbers-backed issue-triage experience (Double Service).
3. She asks questions like an owner.

Ask only two questions

1. "For this role, what is the definition of a completed migration — a successful login, the first notice completed, or sustained adoption over time?"
2. "What is the biggest blocker in the migration today — customer readiness, data, training, or product exceptions?" (This gets her talking about her own headaches — a CEO remembers the person who listened.)

2. Double Service — One Story, Two Fears Covered

One line: At Double Service I was a customer issue-to-resolution operator — turning problems customers couldn't quite articulate into work the team could judge, engineers could pick up, and that actually got resolved.

The only numbers-backed direct proof in the room — lead with it whenever experience comes up. Three beats:

1. Vague to clear — issues from 12 municipal clients scattered across chat, email, and screenshots became one card each: status, priority, evidence, next action; traceability rose to ~90%.
2. Clear to hand-off ready — repro steps and evidence attached, so engineers start with usable context; resolution time fell ~40%.
3. Repeats to specs — recurring needs became 6 implementation-ready specs.

⚠️ Never open with Notion, SQL, or Postman (it reads as data analyst). The protagonist is always: "how I take a customer problem from vague to something the next person can act on."

3. AI — A Card to Play, Not an Opening

This JD is not buying an AI identity. Bring it up at most once, and only as support for "how I keep the work reliable." One loop, done in 30 seconds:

1. Delegate — repetitive work with stable judgment gets written into an SOP that AI runs: assembling current state, flagging gaps, marking quiet or stuck cases.
2. Refine — when AI gets it wrong, I don't just fix that output; I write the missing exception rule back into the SOP so the next case is judged better.
3. Gate — before anything goes out, I go back to the original source and verify it myself; AI never sends on my behalf.

Point 3 must close the loop — in a business where a small mistake can mean real legal consequences, the human gate makes AI one more layer of checking, never one less.

One-line claim

Closing Anchor