Engagement type: Iron Pine self-service product Current status: Live and free to use at check.ironpine.ai
The Problem
Every conversation with a mid-market executive about AI eventually lands on the same question: where should we actually start? The variant matters — sometimes it is "what should we prioritize," sometimes "what are other companies like us doing," sometimes "how do we avoid wasting six months on a pilot that goes nowhere." Underneath, the question is the same. They have the mandate, they have some budget, they have a team that could execute, and they are looking for a structured way to identify the highest-leverage opportunities for their specific business.
The paid answer to this question is an Integration Assessment. Four to six weeks of structured discovery, a complete operational map, an opportunity portfolio scored on impact and feasibility, and a phased implementation plan. It is the right answer for a company preparing to integrate AI at scale. It is also a commitment most executives are not ready to make on a first conversation.
The actual first move for most executives is lower-stakes: can we see what the analysis would even look like? Can we get a preview of how a structured assessment thinks about our business, before we decide whether to commission a full one?
The AI Health Check was built to answer that. It is a self-service AI readiness assessment anyone can run against their company URL, in roughly a minute, at no cost. Enter your company website, answer a branching questionnaire about your operating context and current AI posture, and receive a branded readiness report with scored opportunities, a readiness score, and an executive summary. It is not a marketing funnel disguised as a tool. It is a working, useful artifact in its own right — a lightweight version of the deeper assessment work Iron Pine does as paid engagements.
The Approach
The Health Check is the public-facing surface of a deeper internal system. Iron Pine's Intelligence Engine is the pipeline that generates branded strategic roadmap deliverables for paid clients — a ten-step analysis pipeline with operator review, completeness scoring, and inline editing of scored opportunities before the final branded document ships. The Health Check is a simplified, self-service derivative of that pipeline: the same architectural DNA, shallower depth, zero operator review layer, optimized for a 60-second response rather than a multi-week engagement.
Building the Health Check as a derivative product rather than a standalone tool gave it two important properties. First, it inherits the pipeline discipline of the internal engine — temperature zero on deterministic scoring steps, exponential backoff retry on every API call, partial save on pipeline failure, queue pattern for concurrent executions. These patterns were battle-tested in the internal Intelligence Engine before being adapted for the public tool. Second, it creates a natural progression for users who want to go deeper. A reader who runs the Health Check and finds the output useful has a clear next step — engage Iron Pine for the full assessment that produces the deeper analysis. The Health Check is built to be genuinely helpful on its own terms, not to be a walled garden that forces users into a sales conversation.
Three decisions shaped the Phase 2 build:
Inline styles with CSS variables, not Tailwind. The product needed full brand control and complete framework independence. The Health Check renders against Iron Pine brand colors and typography with zero Tailwind preflight conflicts — inline styles plus CSS variables keep the visual layer tight to the brand specification.
Hybrid sandbox build pattern. The product was built end-to-end via Chrome browser tools inside a single conversation with Claude — files generated, delivered as archives, extracted locally, and tested through Chrome MCP in the same conversation. The entire build took five working days. This was the first Iron Pine product to use this pattern end-to-end, and it informed subsequent pattern choices: for greenfield builds needing simultaneous iteration and visual verification, hybrid sandbox build outperforms the traditional terminal-driven flow.
Free tier as the product, Pro as a passing option. The core Health Check is genuinely free. A Pro tier exists for users who want ongoing monthly re-assessment, the full opportunity portfolio rather than a preview, and a downloadable branded report — but the design center is the free tier, not the conversion funnel. A user who never upgrades has received the value the tool was built to deliver.
The Architecture
The Health Check runs on the standard Iron Pine stack with a production-grade AI pipeline at its core.
Pipeline. Six-step Claude analysis, running sequentially with full resilience handling. Step one scrapes the submitted URL and extracts structured company context. Step two assembles operating context from the branching questionnaire. Step three maps relevant AI and industry trends against the company's profile. Step four identifies candidate AI opportunities. Step five scores each opportunity at temperature zero against a deterministic rubric. Step six produces the executive summary. Total execution time runs 55 to 65 seconds. Pipeline cost comes in at roughly $2.50 per report across six Sonnet API calls.
Resilience. Every Claude API call runs through a callWithRetry wrapper with exponential backoff. Pipeline partial-save on failure means a step-three failure does not lose the step-one and step-two output — the pipeline can be resumed from the failure point without starting over. A queue pattern caps concurrent executions at 10 via Supabase-backed status tracking to protect against traffic spikes overwhelming the API quota.
Auth and email. Unified login with email-first detection — the auth flow looks up whether the submitted email exists as a Supabase auth user and routes to login or signup accordingly, without forcing the user to choose. Account creation happens after the report is generated, not before, so a user can run the Health Check without committing to account creation on the front end. SendGrid handles report delivery, welcome email, and a five-email nurture sequence. Vercel Cron schedules daily nurture email dispatches and monthly report-update runs.
Operational layer. An admin dashboard at /admin provides real-time visibility into subscribers and report activity — full subscriber list with detail panel, readiness score inspection, and opportunity drill-down pulled from the nested JSON structure each report produces. A smoke test endpoint at /api/smoke-test runs six checks before every deploy: Supabase connectivity, Anthropic API reachability, SendGrid delivery, Stripe access, timeout behavior, and retry module health. The smoke test is part of the deploy checklist, not just a development utility.
Branding and SEO. The report output is a styled PDF-quality document rendered in-browser, with Iron Pine branding throughout. Every report page carries a 1200×630 branded Open Graph image and full meta tags so a shared report link produces rich previews on social and messaging platforms.
The Outcomes
The Health Check launched with the full pipeline in production: six-step analysis running in 55 to 65 seconds against any company URL, branded report rendering with scored opportunities and executive summary, auth and email infrastructure operating end-to-end, admin dashboard providing operational visibility, and smoke test verifying every integration on every deploy.
The build took five working days — sessions zero through nine of the Hybrid Sandbox Build pattern, from initial scoping through production launch. Operating costs settled at approximately $65 per month in fixed infrastructure (Vercel, Supabase, SendGrid), plus usage-based Claude API spend scaling with report volume.
The product is live at check.ironpine.ai and free to anyone who wants to run it.
You can use it yourself. Run the Health Check against your company URL and see what the output looks like. The report takes about a minute to generate and arrives in your inbox within another minute or two after that. No payment, no commitment, no sales follow-up unless you request one.
The Lessons
Five things from building and operating the Health Check now inform our approach to self-service and consumer-facing AI products:
Derivative products beat standalone products for go-to-market discovery. The Health Check derives from the Intelligence Engine — inheriting pipeline discipline, evaluation patterns, and branding. Building it as a derivative rather than from scratch meant the core pipeline was already battle-tested against internal paid use before the public tool shipped. For any Iron Pine product that might eventually want a public-facing surface, building the internal tool first and then deriving the public version is a pattern worth repeating.
Temperature discipline across a pipeline is not a detail — it is the product. The deterministic scoring step runs at temperature zero. The creative steps (trend mapping, opportunity identification, executive summary) run at default temperature. Mixing those up in either direction produces a product that feels either robotically wooden or frustratingly inconsistent. Getting the split right is what makes a multi-step AI pipeline feel like a professional analysis rather than a generic chatbot.
The free tier is the product, not the funnel. Products built as "free tier to drive conversions" produce experiences that feel transactional even when they technically work. Products built as "genuinely useful free tier that users may choose to extend" produce experiences that earn trust. The Health Check is the second kind. Many users will never upgrade. That is fine. The goal is not conversion rate. The goal is a Iron Pine-branded assessment that executives actually find valuable.
Pipeline resilience is not optional at any scale. Even at modest volume, an AI pipeline without retry logic, partial-save, and concurrency limiting will eventually produce user-visible failures at the worst possible moment. Building these patterns into the Health Check from the start meant that an API hiccup during a report generation degrades gracefully rather than stranding the user on a broken state. This is cheap to build in at the start and expensive to retrofit later.
Code injection near return statements is the bug pattern to watch for. During the Health Check build, three separate occurrences of this bug pattern surfaced — email trigger dropping the return keyword, welcome email code splitting a return in half, queue check injected inside a subscriber-not-found return. Each produced a silent failure mode. This pattern is now a standard code-review check on every subsequent Iron Pine build: after any code injection near a return statement, verify the return is still intact. Small discipline, significant downstream savings.
Iron Pine helps mid-market companies integrate AI into how they actually operate — grounded in your data, embedded in your workflows, adopted by your people, and operated with production discipline.
Try the AI Health Check · Talk to us about an Integration Assessment