aiweb

Honka

AI Configurator & Personalized Spec Sheets

KotiConfig is an AI-native home configurator we built for Honka, the Finnish log home manufacturer founded in 1958. Branching pricing flow, A/B/C/D lead tier scoring, and Claude-generated personalized spec sheet PDFs that weave the founders’ own sales pitches into each prospect’s words — embeddable on any partner site.

Client
Honka US
Industry
Premium custom homes
Our role
Design + build
Timeline
6 weeks to prod
Stack
Next.js 16 + Supabase + Claude
Status
Live, generating leads

The Challenge

Pilotship.io partnered with Honka US, the American arm of Honka — Finland’s largest log home manufacturer, founded in 1958 with 85,000+ buildings delivered across 50+ countries. Mark and Tom run the US business, selling premium custom log homes (most starting at $1M+) to American buyers through a HubSpot marketing site they’d built campaign-by-campaign over years.

Two things weren’t working. The configurator they were renting from a third-party vendor was generic, hard to modify, and gave them no way to qualify the prospects coming through. And every personalized spec sheet that went out had to be written by hand — by Mark or Tom — every time. With a long sales cycle and seven-figure builds, every prospect deserved a personalized answer. The system meant only the strongest leads got one.

They needed a tool that did what they’d been doing manually — but at scale, in their brand voice, in the time it takes Claude to read a JSON object.

You dug right into the project — you took it on as if it was your business. What if we did it this way? You actually improved upon what we had. You can’t ask for any more than that from a professional service. That’s like off the charts.
Mark·Honka US

Tom · Honka US

“What I saw when you guys finally put it together was a system that did work.”

What we built

KotiConfig is an AI-native home configurator. Prospects step through a branching flow with live pricing. Mid-flow, a lead-capture wall. On submit, Claude generates a personalized PDF spec sheet — using Mark and Tom’s own sales language — and the lead is scored A/B/C/D so they know who to call first.

KotiConfig — Walkout home style selected, with photography, description and live running total

Feature 01

Branching configurator with live pricing

Every step branches based on what the prospect picked the step before. Selecting Walkout swaps the next five steps for a walkout-specific path; selecting Ranch loads a different one. Pricing updates as they go — so by the time they hit submit, the number on screen matches the PDF they’re about to receive.

“Get Your Personalized Quote” lead-capture modal asking for name, email, build location, timeline, budget, and land ownership

Feature 02

Lead capture at the right moment, not the front door

No splash page. No paywall on the home page. The prospect gets to explore the configurator and watch the price form — and only then, mid-flow, hits a clean capture wall that asks for the information Mark and Tom actually need to qualify: location, timeline, budget, whether they own the land.

The modal is configurable from the admin panel — Mark and Tom can move the wall earlier or later in the flow without writing code.

Investment breakdown page — donut chart showing $10M total split across Base Honka Components, Soft Costs, and Site Development

Feature 03

Investment breakdown they can actually understand

After the configuration step, prospects see exactly where their money goes — base Honka components, soft costs, site development, upgrades — as a donut chart with an expandable accordion. No spreadsheet. No surprise line items at signing.

Personalized spec sheet PDF, page 1 — “Your Honka Home Specification”, branded with Honka colors and Playfair Display serif

Feature 04

Personalized spec sheet PDFs, generated by Claude

On submit, Claude reads the prospect’s selections plus their answer to the anchor question — “what must your new home give you that your current one doesn’t?” — finds the matching Honka pitches (fire safety, energy efficiency, legacy, healthy indoor air), and writes a branded PDF that opens with the prospect’s own vision in their own words. Brand fonts. Brand colors. Same PDF Mark or Tom would have written by hand — in under 10 seconds.

Download a sample spec sheet (PDF) →

Inside the AI

The founder pitch weave

The most distinctive thing in the codebase isn’t a model or a framework. It’s an array. Mark and Tom’s best sales pitches — the ones that actually close — live as code. When a prospect answers “yes, that matters to me,” the same pitch shows up inline in the configurator and gets woven into their personalized spec sheet PDF, in their own words.

Configurator step asking “Is fire safety a concern of yours?” with Yes selected, showing the Honka campfire-secret pitch text inline

Same source, two surfaces

The pitch text the founders wrote, exactly where the customer is

Ask a prospect ‘is fire safety a concern?’ — they answer Yes — and the Honka pitch shows up inline, framed as the founder would have said it on a call. The same line gets woven into the personalized spec sheet PDF an hour later. One source, two surfaces, zero copy-paste.

src/lib/ai/prompts.ts
View on GitHub
// src/lib/ai/prompts.ts
const QUALIFYING_PITCHES = [
  {
    match: "fire safety",
    title: "The campfire secret",
    pitch: "Thick log walls are actually one of the safest ways to build in fire-prone areas. LA County codes specifically list log wall construction as approved for Very High Fire Hazard Severity Zones.",
  },
  {
    match: "next 100 years",
    title: "Reimagine your legacy home",
    pitch: "A Honka Home lasts for centuries. It is the most enduring legacy we can leave for future generations.",
  },
  {
    match: "energy efficiency",
    title: "Energy efficient naturally",
    pitch: "Honka log systems provide a naturally energy-efficient home through thermal mass properties and prevention of air infiltration.",
  },
  // ...3 more pitches, each from Mark and Tom's own sales playbook
]

// For every "Yes" answer in the prospect's qualifying questions:
for (const a of input.answers) {
  const pitch = QUALIFYING_PITCHES.find((p) =>
    a.question_label.toLowerCase().includes(p.match)
  )
  if (pitch && a.value.toLowerCase() === "yes") {
    qualifyingLines.push(
      `- "${a.question_label}" → Yes
  Honka message: "${pitch.title}" — ${pitch.pitch}`
    )
  }
}
The founders aren’t in the room. The system is. Same pitches Mark would deliver on a call — attached to every prospect’s spec sheet, automatically, the moment they submit.

The outcome

What changed

6 weeks
Kickoff to lead-generating prod
Including admin CRUD, brand-guide editor, and HubSpot iframe.
4 tiers
Leads auto-scored on submit
A/B/C/D from `src/lib/lead-tier.ts` — pure, testable, founder-editable.
6 pitches
Founder-approved messages woven by Claude
Live in `src/lib/ai/prompts.ts`; dormant pitches stay commented for cheap reactivation.
< 10s
AI spec sheet generation, end-to-end
Claude Haiku p95 latency for our prompt size + PDF render.
I think we benefited more than you did. Based on what we’ve done and the dollars we’ve spent with past professional services vendors, we definitely benefit here big time.
Mark·Honka US

Mark · Honka US

On choosing a small, hungry team over a big agency.

Decisions that mattered

Claude Haiku, not Sonnet, for the PDFs

Why

Spec sheets need to feel instant — every extra second feels like the page hung. Haiku hits sub-2-second latency on our prompt size with the same JSON-schema adherence we’d get from Sonnet at a fraction of the cost.

Trade-off considered

Slightly less stylistic flair in long-form copy. Acceptable because the founders’ voice carries the soul — we just needed the model to weave it in faithfully.

Explicit branching, not tag-based routing

Why

Every Honka home style (Ranch, Walkout, 2-Story) needs its own unique step sequence. We modeled it as explicit "this step → that step" jumps on each option, so the flow stays auditable and Mark and Tom can change branches from the admin UI without code.

Trade-off considered

Steps duplicate where two paths converge. Worth it for editability and for keeping the founders unblocked.

Iframe-embed-first, not standalone

Why

Honka’s traffic already lived on the HubSpot marketing site they’d built over years. Forcing prospects to leave for a separate configurator domain would have leaked conversions. Iframe + postMessage SDK keeps the prospect in their existing funnel.

Trade-off considered

A few CORS gotchas and a slightly more complex auth model for the admin panel. Worth it to preserve their analytics and SEO equity.

Tech Stack

Next.js 16
Supabase
Claude Haiku
React PDF
Tailwind
TypeScript
React
Vercel
Would I recommend you for something like this? Absolutely. Would I recommend anybody else we work with? No.
Tom·Honka US

Tom · Honka US

26 seconds. The clearest answer in the whole conversation.

Want a system like this for your business?

Here's how to figure out if it's the right move — without talking to a salesperson.

1

Spec it in Mission Control

Answer a few questions and our AI drafts a build spec for your idea — free, no signup.

2

Pressure-test it anywhere

Paste the spec into Claude, ChatGPT, or any AI for a free second opinion on feasibility and cost.

3

Talk to us if it's a fit

When you're ready, we'll scope, price, and ship — same studio that built this.