Capstone - The Mock Account Loop

One fictional account, four deliverables, one honest scorecard. This is the practical case the loop puts in front of you, simulated so you can fail it cheaply now instead of live.

Read the brief once, then treat it as fixed. Every later section is a timed artifact you produce against these facts - a 30/60/90, a stalled-rollout rescue, an EBR and an expansion play. The account is deliberately uncomfortable: a real Cursor deployment in a 2,000-engineer org never arrives clean.

Your charter on this accountThe three numbers

An ADM's number is not a vibe. On Meridian you own three measurable outcomes over the first 12 months and every deliverable has to ladder to one of them.

The constraints you have to plan aroundWhat makes this realistic

Give yourself 25 minutes and build the plan as if you're presenting it to the VP Eng next week. Phases, metrics per phase, the Copilot story and your top risks.

The trap in a 30/60/90 is listing activities that sound busy. A strong plan reads the other way: each phase opens with the number it moves and the activities exist only to move it. On Meridian, phase one is gated by the security review, so the first 30 days are as much about pre-empting the blocker as starting the pilot.

Metrics and leading indicators per phaseMake every phase measurable

The Copilot narrativeCoexist, then earn the switch

Do not open by trashing Copilot. Skeptical engineers have muscle memory there and an attack reads as insecurity. Run them side by side in the pilot and let the depth difference show - Cursor's codebase-aware context, Agent doing multi-file work, .cursor/rules/ encoding the team's conventions. The switch is something the pilot data earns, not something you assert.

Top three risks and mitigationsName them before the sponsor does

It's month four. Activation has stalled at 40% and depth-of-use is sliding. You have 20 minutes to diagnose from the data and prescribe - not to reach for “more training.”

The reflex answer to a stall is a blanket re-enablement push. That's the answer a panel is watching to see you avoid. A stalled account is almost never uniformly stalled. Segment first, then prescribe per segment.

Read the signal before you actDiagnose from telemetry

For the dead team, resist guessing. Pull the telemetry and then talk to two or three engineers directly. A flatline usually has a specific cause: a security setting blocking the codebase index, a tech stack where the starter rules don't fit or a vocal senior engineer who tried it once and wrote it off.

What you escalate vs. solve yourselfUse the account team

Design the re-engagementA visible win plus the champion

Re-engagement needs a fresh, high-visibility win and a re-activated champion in the same motion. Pick one painful, common task - a flaky test suite, a migration, a slow review queue - and land an Agent-driven win on it with the healthy team. Then have the champion, not you, present it at the next eng all-hands.

Build the EBR you'd run with the VP Eng at the six-month mark. It opens with their goals, pairs data with real developer stories and surfaces the renewal before they ask.

An EBR is not a usage dashboard read aloud. The sponsor is skeptical of AI ROIReturn on Investment. The value gained versus what it cost, the language an economic buyer funds deals in., so the structure matters more than the slides. Open in their language, prove value with paired data and stories and make the renewal a natural next step rather than a separate ask.

The EBR arcSix moves, in order

1. 1Open with their goals. Restate the outcomes the VP Eng signed up for, in their words. “You wanted faster onboarding for new hires and less time lost on boilerplate review.” You're grading against their scoreboard, not yours.
2. 2Show progress in their metrics. Map your telemetry onto those goals. Activation and depth-of-use are your proof, but framed as “three quarters of the pilot org now uses Agent on real tasks weekly,” not as raw chart dumps.
3. 3Pair each number with a story. For every metric, bring one concrete, named workflow win. The data proves scale; the story makes it real and memorable to a CTO.
4. 4State value honestly - no inflated ROIReturn on Investment. The value gained versus what it cost, the language an economic buyer funds deals in.. Talk about time reclaimed and friction removed without claiming a fabricated productivity multiplier. Credibility with a skeptical sponsor is worth more than a big number you can't defend.
5. 5Bring the forward plan. Show the next two quarters: the wave to new BUs, the depth targets and where you see expansion. The sponsor should see momentum, not a victory lap.
6. 6Surface the renewal early. Name the renewal timeline now and tie it to the realized value, so it's a continuation of a working program rather than a year-end surprise.

Pairing data with workflow storiesThe credible-evidence pattern

Answer the hardest question head-onPre-empt the skeptic

A VP Eng who doubts AI ROIReturn on Investment. The value gained versus what it cost, the language an economic buyer funds deals in. will ask some version of: “How do I know this is real productivity and not just developers feeling faster?” Have the answer ready and lead with the honesty.

The pilot org is thriving and the EBR landed. Now build the expansion: where you grow, when you pitch, how you quantify it and how Sales stays in lockstep so Meridian sees one team.

Expansion is earned, not pushed. The fastest way to torch trust with a technical buyer is to pitch more seats before the first ones are deep. The right expansion anchors on a proof point the beachhead already produced and times itself to a visible win.

Pick the expansion vectorWhere the next dollar comes from

On Meridian, the cleanest first move is replicating the platform team's success into the next business unit, anchored on the EBR proof point. The MCPModel Context Protocol. A standard that lets an AI agent pull in context from outside the repo, like Jira tickets or internal docs. and Agent-at-scale story is the value-deepening play you bring once the second BU is live.

Time the pitch and multi-threadWho hears it and when

1. 1Wait for a visible win. Pitch expansion in the wake of a result the customer can see - the EBR, an all-hands demo, a backlog the beachhead drained. A win is the only credible pretext for growth.
2. 2Map beyond the single sponsor. You have one executive sponsor today. Find the economic buyer for the next BU - a peer VP or director - and build a relationship before you need their signature. Single-threaded accounts churn when the sponsor moves.
3. 3Let the champion carry it inward. A peer recommendation from the platform lead to the next BU's lead beats any pitch you deliver. Arm the champion with the proof and let them open the door.

Quantify in the buyer's own metricsSpeak their scoreboard

One account team with SalesNo surprises, no turf

The expansion is a commercial motion, so Sales has to be in it from the start - not handed a closed deal at the end. Brief the account executive on the proof, the target BU and the timing before you talk numbers with the customer. Meridian should never sense daylight between you and Sales.

Four deliverables done. Now grade them the way the panel would - across five dimensions, with a hard look at anything you overclaimed.

Self-scoring only works if you're honest about it. The whole point of this capstone is to find your weak dimension while it's free to fix. Score each one 1–5, then read the consistency check, because a loop that contradicts itself across deliverables fails even if each piece looks strong alone.

The five-dimension rubricScore each 1–5

The overclaim auditFind your own hand-waving

Go back through all four deliverables and flag every place you asserted something you couldn't defend with the brief's facts. That's the exact instinct Cursor's “no AI in interviews” and truth-seeking culture is built to screen out.

Find your weakest dimension and drill itOne drill, highest impact