Field Notes - Nov 6, '25

Executive Signals

• Connectors are the new dependencies: contract-first adapters keep delivery moving safely before connectors land
• Evidence beats status: clipped outcomes and simple deltas arm sponsors to advocate
• Invisible automation, visible outcomes : end-to-end proofs replace walkthroughs for trust with timestamps
• Reliability is product: first-class failure states, alerts, and runbooks reduce blast radius
• Meetings to momentum: weekly show-and-tell plus async tracker beats exec dailies
• Quality gates, not vibes: ship only when code, vulns, and telemetry clear

CEO

Replace Exec Dailies With Weekly Show-and-Tell

Executives do not need daily status if a weekly demo plus retro exists and an async tracker carries the narrative. This gives leaders back roughly half a day per week while raising signal quality with evidence over updates.

• After the first successful demo day, sunset exec attendance at dailies; keep weekly demo and the tracker
• Keep the tracker read-only and visual with phases, owners, and ETA bands
• Feed wins into existing exec 1:1s so momentum compounds asynchronously

Turn Demo Day Into a Wins Engine

Sponsors advocate better with artifacts than with Jira links. Stand up a lightweight wins library that anyone can forward in two clicks: a single page with a prominent progress bar, a milestone countdown, and short clips of end-to-end runs. Pair with one or two simple before/after deltas to make impact legible.

• Publish 20–60 second recordings of full runs with expected outputs
• Add defensible before/after deltas for one key task
• Ship one new artifact per cycle to keep advocacy fresh

Model the First One, Then Hand It Off

If teams or partners are new to the cadence, run the first session yourself to establish the pattern, then immediately delegate. Pre-flight the agenda, timebox to 60 minutes, publish the recording and clipped highlights, and name the next facilitator.

• Dry-run 24 hours ahead; ban “it worked on my machine”
• Timebox to 60 minutes; publish clips for easy sharing
• Document rules on one page and assign the next facilitator

Product

Separate Submit From Check Status

Ship “Submit to OEM portal” as its own deliverable. Treat status polling and ingestion as a separate flow per OEM, since names and semantics differ. Copy the common 95 percent, isolate the bespoke 5 percent, and track them as distinct epics to unblock value earlier.

• Track “Submit” and “Check” as separate epics with owners and SLAs
• Maintain a regression harness across all OEM adapters
• Stand up a pre-submit test card using known dealer fixtures

Engineering

Freeze Scope on the Legacy Platform

When a legacy system enforces hard caps, stop adding scope. Route new patterns to the target stack and keep delivery moving with stubbed inputs via a JSON contract. Expect at least one week for the connector to be usable; do not place critical paths on earlier dates. Plan migrations later to avoid increasing the legacy blast radius.

• Declare a moratorium on new objects and tables in the legacy tool
• Define a JSON contract; build adapters against it until the connector lands
• Migrate existing workloads later; do not expand the legacy blast radius

Instrument Failure Paths Before You Ship

Reliability is a feature. Add a first-class failure state in the CRM for compliance records, emit Slack alerts on errors with enough context to act, and queue manual intervention with a simple runbook. Hold a tight triage SLA: acknowledge within 15 minutes, resolve or workaround within 24 hours.

• Page to Slack on any non-2xx or timeout with OEM, payload hash, retry count
• Update CRM status atomically; no silent failures
• Maintain a manual queue with a concise runbook and SLAs

Define Go-Live Gates Early

Bake security and observability requirements into an infrastructure ticket jointly owned by Engineering and IT. Block launch on code quality, vulnerability posture, and telemetry so day-one reliability is predictable.

• Code quality: no critical issues; line coverage at or above 70 percent
• Vulnerabilities: zero critical or high at release in Snyk or equivalent
• Observability: error tracking with alerting, deploy health checks and runbooks

Demo Invisible Automation With Evidence

Backend and automation work has no UI. Demo outcomes. Enforce a rule: every contributor demos something that runs end to end on production or production-like every two weeks. No code walkthroughs, just timestamped evidence. Keep “error postcards” internal for learning and send only wins upward. Timebox each demo to 3–5 minutes.

• Show the automation executing with expected outputs and guardrails
• Maintain error postcards internally; publish only wins upward
• Timebox to 3–5 minutes to force crispness and shipping

Merge Retro + Demo Into an AI Center of Excellence

Use the same session to ship and to teach. After demos, run a tight start/stop/keep retro, then a micro-teach on how AI tooling accelerated real work. Capture only high-signal patterns and track adoption like a product metric.

• Require one “AI pattern of the week” with a measured delta
• Maintain a living one-pager of patterns; prune aggressively
• Track adoption metrics such as percent of PRs touched by AI and review time