How I AIHow I AI

Quests, token leaderboards, and a skills marketplace: the elite AI adoption playbook | John Kim

Claire Vo and John Kim on sendbird’s AI adoption playbook: quests, leaderboards, and internal marketplaces.

John KimguestClaire Vohost
May 5, 202642mWatch on YouTube ↗
Marketing as builders (swag store, Easter eggs, Stripe)Automators platform and “quests”AI agents generating PRDs and codeSecure internal app templates and compliance-by-defaultAI Engineer for Internal Operations and cross-functional task forceCompany-wide skills/plugin marketplaceToken consumption dashboards, tiers, and enablement
AI-generated summary based on the episode transcript.

In this episode of How I AI, featuring John Kim and Claire Vo, Quests, token leaderboards, and a skills marketplace: the elite AI adoption playbook | John Kim explores sendbird’s AI adoption playbook: quests, leaderboards, and internal marketplaces Sendbird empowered non-engineering teams (especially marketing) to ship production-grade experiences—like a full swag store with Stripe—in days using AI-assisted building.

At a glance

WHAT IT’S REALLY ABOUT

Sendbird’s AI adoption playbook: quests, leaderboards, and internal marketplaces

  1. Sendbird empowered non-engineering teams (especially marketing) to ship production-grade experiences—like a full swag store with Stripe—in days using AI-assisted building.
  2. The company created an internal “Automators” quest system where employees post automation needs and others (humans and AI agents) build and deliver reusable workflows.
  3. A dedicated “AI Engineer for Internal Operations” function partners with CTO/InfoSec to provide a vetted, secure default stack, templates, and daily-updated guides that make shipping safe and easy.
  4. Adoption is actively managed via token-usage dashboards and a tiered leaderboard (from beginner to “AI god”), used for coaching and expectation-setting rather than performance punishment.
  5. Beyond work tooling, Kim demonstrates personal AI knowledge systems (e.g., “Gardener” for notes and auto-generated learning wikis) to highlight AI’s leverage for continuous learning and curiosity.

IDEAS WORTH REMEMBERING

5 ideas

Make AI adoption a product, not a policy.

Sendbird built explicit UX for adoption—quests, marketplaces, rewards, templates, and dashboards—so using AI feels like participating in a system designed to help, not a mandate from leadership.

Let non-engineering teams ship real production value.

The marketing-built swag store illustrates a core outcome: when creative teams can build and deploy safely, “fun” and culture-forward experiences become feasible without fighting for roadmap priority.

Use a quest marketplace to bypass traditional prioritization bottlenecks.

Quests convert internal needs into well-scoped, volunteer-friendly projects with fast feedback loops, enabling “micro-projects” to get done outside rigid sprint planning.

Standardize a secure “happy path” to production.

Templates with authentication, environments, and pre-vetted tooling reduce risky shadow IT and stop employees from deploying sensitive tools to the public internet just to get something working.

Instrument usage, but frame it as enablement—not surveillance.

Token dashboards and tiers help managers tailor coaching (beginner → intermediate → expert) while explicitly avoiding the “lines of code” trap and keeping it out of performance reviews.

WORDS WORTH SAVING

5 quotes

This is an internal platform where anyone in the company can raise their hand and create what we call the quest. When there's a quest, AI can actually read through the specification, create PRDs, and start actually coding.

John Kim

It's taking someone's super creativity and giving them powers to, to deliver it to your customers.

Claire Vo

Basically a marketplace of AI needs and AI builders inside your company where anybody can just pop in and say, "Oh, I, I think I know how to do that."

Claire Vo

We measure AI gods as somebody who spend more than 100 million tokens a day.

John Kim

This is a beautiful time to fail forward and still get up and run faster than the others, right? So use more examples of that, and people bring out their confidence, so you have to really build energy around those people.

John Kim

QUESTIONS ANSWERED IN THIS EPISODE

5 questions

How exactly does a “quest” get specified—what fields (risk, weeks saved, data touched) are required before work can start, and who approves them?

Sendbird empowered non-engineering teams (especially marketing) to ship production-grade experiences—like a full swag store with Stripe—in days using AI-assisted building.

What does the new flow look like when “AI can read the spec, create PRDs, and start coding”—which parts are automated vs. reviewed by humans before merging?

The company created an internal “Automators” quest system where employees post automation needs and others (humans and AI agents) build and deliver reusable workflows.

How do you prevent token metrics from becoming a vanity contest (e.g., people “burning tokens”) while still keeping the leaderboard motivating?

A dedicated “AI Engineer for Internal Operations” function partners with CTO/InfoSec to provide a vetted, secure default stack, templates, and daily-updated guides that make shipping safe and easy.

What are the concrete components of your secure internal app template (authN/authZ, logging, secrets management, data access boundaries), and what’s the minimum you’d recommend other companies implement first?

Adoption is actively managed via token-usage dashboards and a tiered leaderboard (from beginner to “AI god”), used for coaching and expectation-setting rather than performance punishment.

In the skills marketplace, what’s the unit of reuse (prompt, tool, workflow, code package), and how do you version/maintain skills so they don’t rot?

Beyond work tooling, Kim demonstrates personal AI knowledge systems (e.g., “Gardener” for notes and auto-generated learning wikis) to highlight AI’s leverage for continuous learning and curiosity.

Chapter Breakdown

John Kim’s AI-first ambition: AI as part of the workforce

Claire Vo frames the episode around treating AI adoption like a product, not a policy. John Kim explains Sendbird’s goal: making AI a true workforce partner by empowering every function—not just engineering—with tools, access, and enablement.

Marketing builds a full swag store in 1–2 days (with payments and Easter eggs)

John demos a marketing-built Delight.ai swag store (“Big Ass Energy”) created without engineering support. The example showcases what happens when non-technical teams can ship real, production-quality experiences fast—including Stripe integration and playful details.

Why “fun” used to lose: the old roadmap and prioritization bottleneck

Claire and John contrast today’s build velocity with the “before times,” when marketing ideas competed for scarce engineering cycles. They argue AI makes fun, creativity, and experimentation cheap enough to prioritize—and that changes culture and customer experience quality.

The Automators platform: internal quests as a marketplace for builders and needs

John introduces the Automators platform, where any employee can create a “quest” describing an automation or tool they need. Other employees (or AI agents) can pick up quests, collaborate, and deliver reusable workflows/skills—creating a lightweight internal build marketplace.

Quest mechanics that drive adoption: feedback loops, rewards, and visibility

The quest system is gamified to keep momentum: contributors earn experience points and rewards, and teams demo wins in weekly company standups. John emphasizes the fast feedback loop (real internal users) and the motivational “dopamine hit” from shipping useful tools.

AI builds alongside humans: from specs to PRDs to code

John explains a new layer: quests can be handed to AI to generate PRDs and start implementation. This positions AI agents as additional “builders” that help deliver internal automation faster while humans guide, review, and iterate.

Safe-by-default shipping: guides, templates, and pre-vetted infrastructure

To prevent insecure one-off deployments, Sendbird provides internal docs, Git/GitHub guidance, and an application template with authentication, environments, and compliance baked in. This creates a secure ‘happy path’ so any function can build and ship inside guardrails.

The ‘AI Engineer for Internal Operations’ and the cross-functional task force

John describes a dedicated role/team responsible for accelerating AI transformation, reporting to him and the chief of staff. The group partners with CTO/engineering and InfoSec and meets regularly to unblock compliance, logging, and tooling decisions—removing friction for the rest of the company.

Company-wide skills marketplace: turning expertise into reusable plugins

John demos a marketplace where employees publish ‘skills’ and ‘plugins’ (collections of skills) by function—sales, recruiting, design, etc. The goal is reuse and co-evolution: avoid teams rebuilding the same capability in silos and encode institutional knowledge into shareable components.

Driving real adoption: curiosity, champions, and some top-down pressure

John shares that adoption required both organic pull (curious people exploring) and executive push (leaders setting expectations). Managers even have direct conversations with low-usage employees to understand blockers—positioned as support, not punishment.

Real wins in practice: marketing’s internal ‘mini-SaaS’ and the Buzz Board campaign tool

John highlights concrete outcomes: marketing built a full internal portal of tools (planning, ABM, competitor review) and a ‘Buzz Board’ for campaign creation and social sharing. Teams can generate posts (e.g., billboard campaign assets), track engagement, and run daily workflows without buying or waiting for external software.

“SaaS isn’t dead”: internal rebuilds and the renaissance of internal tools

Claire and John argue SaaS isn’t disappearing, but many teams will increasingly build bespoke internal software first—optimized for their workflow and culture. John notes internal tools are newly exciting because AI makes them faster, better designed, and less resource-starved than in the past.

Token tracking dashboard and leaderboards: measuring adoption without shame

John demos a company-wide token dashboard tracking usage by model (Claude Code vs Codex), team, and individual—plus a tiered leaderboard from beginner to ‘AI god.’ They explicitly avoid tying this to performance reviews; the dashboard is used to set expectations, tailor enablement, and monitor overall adoption health (including ‘smoothing the curve’ with autonomous AI work).

Personal AI workflows: the ‘Gardener’ for notes + building a personal learning hub

John shares personal use cases: an open-source ‘Gardener’ that enriches and organizes markdown knowledge bases (Obsidian/Logseq style), and AI-generated learning maps for topics like neuroscience. Claire emphasizes how AI can deepen learning by reshaping information into custom, navigable structures.

Lightning round: how to get a company to do this + leadership signaling and mindset

John’s advice: find internal champions with curiosity and agency, spotlight their wins, and encourage ‘fail forward’ iteration. He underscores leadership modeling (top token consumers are senior leaders) and maintaining a cooperative stance with AI tools—building long-term habits and energy around building.

EVERY SPOKEN WORD

Install uListen for AI-powered chat & search across the full episode — Get Full Transcript

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome