← Blue Butterfly for salons
Under the hood · for builders

The pet-styling app is the easy part.

Blue Butterfly is a multi-tenant, deterministic-first, AI-optional platform — built by a company that runs roughly 80% on AI agents and 20% on humans, 24×7, under real governance. v1.1 shipped June 2026: trust layer live — JWT auth, role-based access, row-level security enforced at the database. We're not bolting AI onto a CRUD app. We're proving a new way to build and run a software business. We call the reusable model Chrysalis. Pet styling is the proving ground.

~80/20
agents / humans — 24×7 build pace
DB-enforced
tenant isolation lives in the database itself, not in app code
0
AI required for the core to work
GO / NO-GO
a quorum votes before anything ships

The agentic org, watching itself.

Real internal dashboards from the running platform — the agentic team holds itself to live OKRs (the same way it holds salons to theirs), a release tracker fed by git + CI, and a quorum that votes before anything ships. This is the control plane, live.

Preview: Blue Butterfly internal dashboards — agentic OKR scorecard, release tracker, and live signals

Agentic OKR scorecard · live-signal release tracker · the staff console + pet-parent portal it ships for.

Opinions we'll defend.

The interesting part isn't the feature list — it's the stances. Here's how we think. (The specific gates, schemas and the secret sauce stay behind the curtain; this is the philosophy.)

deterministic-first

AI is a fast-follower, never a dependency

The platform is fully functional with zero AI. Every number a salon trusts is computed deterministically and grounded in real data — no model in the money path. AI is opt-in, metered, and layered on top, so it can fail without taking the business down.

Why it's hard: resisting the urge to LLM-everything, and drawing a hard line where determinism ends and intelligence begins.
multi-tenant

Isolation enforced at the database, not in app code

Tenants are isolated at the data layer with row-level security running under a non-privileged database role — so an app bug can't leak one salon's data into another's. The gate is the database itself, not developer discipline. Shipped in v1.1.

Why it's hard: making isolation self-maintaining and provable on every change — not a thing you hope someone remembered. Most teams skip this until it's a headline.
agentic delivery

We build with teams of agents under governance

Features are built by coordinated AI workstreams, integrated by a tech-lead orchestrator, and gated by deterministic checks — readiness, deployment, drift. Then a multi-discipline quorum reviews the work and casts an explicit GO / GO-WITH-CONDITIONS / NO-GO vote before anything is promoted.

Why it's hard: velocity without an adversarial check is how regressions reach prod. The governance is the product as much as the app is.
own our primitives

We don't glue over third-party shapes

When complexity accretes around a vendor's API, we collapse it into a clean, first-party primitive we own instead of accreting glue. Providers (payments, accounting, comms, AI) sit behind ports, gated per sub-capability, so a swap is config, not a rewrite.

Why it's hard: the discipline to refactor toward a primitive under delivery pressure — and to keep the seams honest.
portable by default

Built local-first to land cleanly on the cloud

Everything we build locally is designed to migrate to managed cloud infrastructure as a config change, not a port — infrastructure-as-code, the same governance gates, the same isolation guarantees from a laptop to production.

Why it's hard: local/cloud parity is a continuous tax most teams skip until it's a rewrite. We pay it up front.
Chrysalis · business-in-a-box

The bigger bet: a reusable agentic company

The real experiment is the 80/20 model itself — codifying the agents, roles, skills and governance as portable assets, dogfooding them to build Blue Butterfly, then packaging the whole apparatus as Chrysalis by Blue Butterfly: a reusable way to stand up or transform any software-driven business. We proved it on ourselves before we sold it.

Why it's hard: it's two products at once — the pet-styling platform, and the machine that builds and runs it.

Every change runs a gauntlet.

Agents do the heavy lifting; deterministic gates and a human-approved quorum keep it honest. No change reaches the trunk on vibes.

decompose parallel workstreams green gate readiness quorum vote deploy + drift watch

Fortune-500 agentic AI, poured into something we love.

We didn't wander into AI. We've built it at scale for the world's biggest companies — and then chose to bring all of it home to a craft, and a community, we're crazy about. We were pulled into the pet-styling world by Kat's passion for dogs, and we're here to transform it — one real, educational, joyful experience at a time. It's really that simple.

Co-founder

Salle Ingle — AI Strategist & Data Architect

25+ years in technology leadership. Formerly Principal Architect at McKinsey & Company, where she led generative-AI agentic-workflow teams for Fortune-500 enterprises and ran global pre-sales for McKinsey's GenAI practice — multi-cloud, multi-lingual, on-prem LLMs at scale. Named to Built In's Top 100 Most Influential Women in Tech; advisory board, UCCS Strategic AI.

Connect on LinkedIn →

Co-founder

Chad Ingle

26+ years across technology and business — a career spent building, scaling and operating real systems and real teams. The steady hand that turns big ideas into things that actually ship and run.

Connect on LinkedIn →

The juicy details stay off the page — the ideas, we'd love to share.

The exact gates, the data model, the governance internals, the agent roster — that's the moat, and it's not going on a marketing page. But the thinking behind it? We love talking about that. If you build platforms, care about multi-tenancy and AI governance done right, or just want to geek out and learn from each other — we'd genuinely love to meet you. We're also (quietly) looking for a few exceptional people who want to build something that matters.