JD Case Study · Headspace · Director, AI Enablement

The AI-enabled operator

Most candidates for an AI-enablement role can describe adopting AI responsibly at scale. This entire portfolio is me doing it. A dozen-plus sourced, cited analyses — researched, fact-checked, and shipped by one operator directing AI through a workflow I designed, with guardrails on every page. The proof isn’t a slide. It’s the site you’re standing on.

Remote (US) / SF hybrid · $138.6K–$250K · View the job description →

Jun 22, 2026

A dozen+

production analyses shipped — research → build → launch, end to end

Every figure

cited and dated; unverifiable numbers flagged, never guessed

4 + 20

deep competitor spotlights, on a 20-company market map

Public data only

an explicit no-internal-data integrity rule on every page

The operating model I run

AI enablement isn’t a tool rollout; it’s an operating model. Here’s the one that produced this portfolio — the same discovery → pilot → scale lifecycle the role is built around, run dozens of times until it compounded.

1

Intake & prioritize

Take an ambiguous goal ("show I can do this role"), turn it into a transparent queue of scoped initiatives, and sequence them by impact. You watched the queue move in real time.

2

Discovery

Fan out multi-agent web research against primary sources — SEC filings, the FTC, investor relations — to assemble a cited evidence base before a word of the artifact is written.

3

Pilot & verify

Draft the analysis, then adversarially verify it: cross-check every number, flag anything a source can’t confirm, and correct the narrative when the data disagrees with the hunch.

4

Scale & reuse

Ship to production, then templatize. The battlecard format is one data file per role precisely so the operating model compounds — discovery once, reuse many times.

Responsible by default

For a mission grounded in member safety and privacy, how you adopt AI matters as much as whether you do. The guardrails below aren’t an afterthought on this portfolio — they’re the reason it’s credible.

Cite or flag

Every figure carries a primary source and an as-of date. Where a source couldn’t confirm a number, it’s marked “not disclosed” — never invented to fill a gap.

Verify before ship

AI drafts; a human standard decides. Numbers are cross-checked against filings before they go live, and the model is corrected when it overreaches.

Integrity by design

Built entirely on public data, with an explicit no-confidential-data rule — the exact discipline a privacy- and safety-sensitive mission demands.

Honest framing

Where a story could overclaim (a competitor’s decline, a coincidence dressed as causation), the copy says so. Trust is the product.

The JD, answered

What the role asks forWhat this portfolio already demonstrates
Define & lead an enterprise AI-enablement strategy and operating model I designed a repeatable AI research→build operating model and ran it to ship a dozen-plus production analyses.
AI intake & prioritization — triage, size value, maintain a transparent queue I scoped, triaged, and sequenced a dozen initiatives against a single goal — the queue moved live.
Lead AI initiatives end-to-end (discovery → pilot → launch); own the outcome Every artifact went discovery (multi-agent research) → pilot (draft + adversarial verification) → launch (shipped, live, linked).
Define standards, frameworks, and responsible-AI guardrails One non-negotiable standard across all of it: cite or flag. Dated, primary-sourced, no fabrication.
Design, evaluate, and iterate on prompts for LLM / agent workflows The whole portfolio is orchestrated multi-agent research and build — structured prompts, iterated against output quality.
Define adoption & outcome metrics; comfort with dashboards (Looker / Tableau) I build the dashboards too — interactive models, scorecards, and ported Tableau work make up the rest of this site.
Executive-ready communication, recommendations, and reporting Every analysis ends in an executive narrative and a recommendation — not a data dump.
Responsible AI in a privacy- / safety-sensitive environment Public-data-only, integrity-flagged, honestly framed — the discipline a member-safety context requires.

A note on rigor

AI is the tool; the judgment, the sourcing standard, and the final call are mine. Every figure here was verified against a primary source before it shipped, and anything unverifiable was flagged, not fudged. That’s the difference between using AI and being used by it — and it’s exactly the discipline a safety-sensitive mission requires. I don’t pitch AI enablement. I ship it.

This whole site is one AI-enabled operator’s output, built responsibly and in the open. Let’s talk about doing it at Headspace’s scale, in service of the mission. — Paul Brown