Leveling narrative and reject postmortem
North-star loops fail for predictable reasons. I prepare the leveling story before I need it. I write the postmortem the same day as the reject — memory lies after a week.
What “Principal scope” means
Titles vary. Org reach doesn’t.
| Level-ish | Who inherits the work | Story that lands |
|---|---|---|
| Senior | My team | “I led the project and we shipped.” |
| Staff | Adjacent teams / the domain | “I set technical direction for this area; adjacent teams adopted pieces.” |
| Principal | Org / multi-org | “I set the approach for X across programs; peer teams and leadership bought the standard; outcomes moved org metrics.” |
Self-test on the last 18 months: if my strongest stories end at “my team shipped,” I’m solidly Staff-shaped. If they end at “four other teams adopted the pattern” or “I made the architectural case and leadership signed,” I’m closer to Principal-shaped. Lucid’s title may say Sr Staff. The scope narrative has to carry the rest — without inventing a title I didn’t hold.
Leveling narrative (Sr Staff title → Principal-scope work)
Claim: Lucid title may read Sr Staff; scope was Principal AI — multi-program architecture, org standards, 0→1 AI, measurable outcomes.
30-second pitch I’d actually say
Lucid titled me Sr Staff. The work sat across programs: AI from zero on 20k-supplier EDI, multi-agent and RAG in production, standards other teams reused, $10M licensing eliminated and $7M net I can walk as two numbers — not a mashed vanity total. Your ladder may call that Staff or Principal — I’m happy to be leveled here. What I won’t do is shrink it into a single-team feature lead.
Evidence pillars (pick three per panel — don’t dump five)
- Org reach — multiple programs, ADRs, a standard another team reused
- Production AI — agents, RAG, eval gates, HITL
- Business outcomes — $10M licensing + $7M net + $1M Kafka, spoken as separate numbers
- Inspectable proof — AegisAI + Enterprise RAG + technical-review
- Charter wins on the current seat (when real)
Anti-patterns
- Apologizing for the title (“I know I’m only Sr Staff…”)
- Inflating titles I didn’t hold
- Catalog counting (“I have 17 demos”)
- Passive voice that hides who decided what
- Sanitized multi-year arcs with no mid-course correction (sounds fake at Principal)
Mid-course correction beat
Strong Principal stories include: original bet → what evidence changed → what I revised → who I had to re-align. If every story is a straight line to glory, interviewers assume inflation.
Scripts
“Are you Principal or Staff?”
Lucid’s ladder put this work under Sr Staff. Multi-program architecture, standards other teams reused, production outcomes. I’m happy to be leveled by your bar. What I won’t do is pretend it was a single-team feature lead.
“Tell me about influencing without authority”
Kafka/MSK. Infra director didn’t want to touch production. I didn’t win a meeting — four-week matched pre-prod, SMT rebuilt as Lambda, dual-write. He sat in pre-prod and presented the win. Resume outcome is $1M+ annual. Influence was uneven on residual paths — I won’t claim everyone converted. Happy to go into the disagreement.
“What would you do differently?”
I would’ve put the negative tenant test in CI a quarter earlier. We caught a class of bugs in review that a deny-path eval would have caught automatically. “Nothing — it went great” is not an answer I’ll give.
“You’re a strong Staff. Not Principal here.”
Help me see the bar here — is it org-wide standards adoption, multi-program outcomes, or a title ladder I don’t have yet? I’ll take Staff if the work is Principal-shaped and the path is dated. I won’t take a Senior-shaped backlog with a Staff sticker.
Mid-course correction (have one ready)
Original bet: one highly autonomous agent that just completes the EDI case. Evidence that changed: irreversible writes with no risk tier and evals that only covered the happy path. What I revised: specialists plus HITL plus a promotion gate for autonomy. Who I had to re-align: delivery peers who wanted the demo path. If every story is a straight line, I sound inflated.
Common reject causes → counters
| Likely cause | What it felt like | Counter next loop |
|---|---|---|
| Isolation depth | Follow-ups after the happy path; Demo/Strict confusion | Whiteboard threat model cold; mention negative evals; deep path |
| Language-fit anxiety | “How’s your Go / Java day-to-day?” | Lead Spring/Java enterprise + Python agents; never overclaim daily Go |
| Demo farm | Eyes glaze during catalog tour | Two systems only; limitations first; Lucid outcomes before URLs |
| Level mismatch | “Great Staff, not Principal here” | Leveling pitch + charter wins; ask what Principal looks like here |
| Referral gap | Slow recruiter black hole | Double intro quality; stop relying on cold apply |
| Eval science thin | “How do you know it works?” → vibes | Artifact A/B; golden-eval CI; false-confidence examples |
| Values / judgment (labs) | Rehearsed “responsible AI” lines | Practice updating mid-conversation; name a refusal boundary |
| Coding rust | Timed round wobble | Coding refresh; weekly reps |
After two rejects with the same root cause, change one input (artifact, intro source, or rehearsal focus). I do not build a new platform to soothe the ego.
Reject postmortem (offline)
Copy per reject. Do not commit employer-confidential notes.
Company / role:
Stage reached:
Date:
What they probed hardest:
Where I was weak (my words, not theirs):
Evidence I lacked:
Was this level mismatch, skill gap, or process/referral gap?
Counter for next loop (one action only):
Keep / kill / pause this target:
Emotional residue (name it so it doesn’t drive the next build):
Same-day write-up beats a polished essay next month.
Mock panel rubric (self or Practice Arena)
Score 1–5. Want ≥4 average before a north-star onsite:
| Dimension | Score | Notes |
|---|---|---|
| Pitch clarity (governed agent platforms) | ||
| Multi-tenant threat model → layers → evals | ||
| Gateway / HITL story | ||
| Access-aware RAG story | ||
| Lucid outcomes without fluff / honest provenance | ||
| Honest limitations | ||
| Leveling narrative without apology | ||
| Questions asked to them | ||
| Sounded like a peer, not a brochure |
Record one mock. Brochure-voice is obvious on playback.
If this chapter ends without north-star
90-day correction plan (still no new brands):
- Fix the top reject cause from postmortems (one cause)
- Keep 2 high-quality intros/week for 12 weeks
- Revise the same signal artifact (same option letter) — deeper, not wider
- Re-enter chapter 4 gates only when mock rubric averages ≥4
- Consider a dated bridge upgrade if ownership and leveling are real — climbing from a stronger seat beats grinding pride