STRAIGHT ANSWERS
34 answers across 8 chapters — money, scope, risk, data handling, your team, timeline, how I work and the practical details. Every answer is on this page in full. Link straight to any one of them.
01 — CHAPTER
Every system I build must save more than it costs — that’s the only rule. For most clients the math is: system cost ÷ average client value = number of clients needed to pay it back. If that number is under ~20 clients or under 6 months for your business, we’re in green territory. If it isn’t, I’ll tell you on the first call and I won’t take the project.
Then the system pays back over more clients and a bit longer — usually 3-6 months instead of 2-4 weeks. The math scales down cleanly. What matters isn’t the size of one job, it’s how many are slipping through your pipeline every month. A 30% lift on a €1,000 average job is the same profit shape as a 30% lift on a $10,000 average job.
Because I don’t build throwaway automations. Everything I ship is documented, monitored, has error handling and fallbacks, and belongs to you fully. A $500 Zapier setup breaks in three months and nobody knows why. I build systems that survive API changes, staff turnover, and a bad Monday.
Systems don’t maintain themselves. OpenAI changes a model, Gmail changes an API, your workflow evolves, edge cases surface. Care Plan means 24/7 monitoring (I know it broke before you do), same-day fixes on business days, monthly improvements as your business changes, and a plain-language monthly report in dollars and jobs saved — not tech jargon.
Standard terms: 50% at signature, 50% at launch. For larger builds we can split into 3 payments across milestones. Care Plan is monthly, cancel anytime. For clients outside the US, I invoice in EUR or USD — your choice.
After the first 30 days I send a plain-language report showing actual saved leads, hours, or errors. If the numbers don’t cover the Care Plan, we adjust the scope or you cancel — no argument. I’ve never had a client where the math didn’t clear within 60 days, but I don’t promise magic and I don’t lock people in.
02 — CHAPTER
Full system design, code, integrations, error handling, monitoring dashboards, deployment to your cloud, complete documentation (a real README, not a PDF), 30 days of post-launch support, and a handoff call where I explain every piece. You own the source code, credentials, and infrastructure from day one.
That’s what custom scoping is for. The modules are the common patterns — 80% of clients need 2-3 of them. If yours is different (permits, subcontractor payments, warranty tracking, insurance claims, whatever), we scope it in Discovery and I quote it upfront. Same delivery model, same guarantees.
No. The system runs behind whatever you already have — Gmail, QuickBooks, Buildertrend, JobTread, Housecall Pro, ServiceTitan, Jobber, Google Workspace, Microsoft 365. If a tool I can’t integrate with is central to your workflow, I’ll tell you in Discovery before we sign anything.
Yes — most clients do. Start with 1-2 modules (usually Lead Ops + Estimate Ops), watch the ROI for 30-60 days, layer in the rest as it makes sense. Each additional module is a separate scope and price. No pressure to buy the whole stack upfront.
The core promise is that your team learns nothing new — the system runs underneath what they already use. But if there’s a piece someone needs to touch (approving flagged items, reviewing weekly reports), I record a 5-minute Loom for each and we do one live walkthrough. That’s it.
03 — CHAPTER
Every critical action has a human-approval layer by default. Estimates over your threshold, first-time client responses, anything money-related — the AI drafts, you approve, it sends. For low-risk actions (categorizing photos, filing docs, internal summaries) it acts on its own. You choose the line in Discovery.
This is the #1 real risk with LLM systems. I mitigate it three ways: (1) RAG grounding — the AI can only quote from your actual data, not from training memory, (2) DeepEval and grounding checks running in production catch hallucinated outputs before they reach a client, (3) fallback rules that trigger human review when confidence is low. It’s not zero-risk, but it’s engineered risk, not gamble.
Your data stays in your accounts — Gmail, your CRM, your Drive. Where I need a vector database (for company-wide search), it’s deployed to your cloud account (Supabase, GCP, Vultr) — you hold the keys, you can revoke access with one click. No data goes to my servers. No training of external models on your business. Written into every contract.
Care Plan includes 24/7 monitoring. Alerts hit me before they hit your inbox. Same-day fix on business days, next-morning on weekends. Every system has fallbacks built in — usually a “graceful degrade” back to your manual process, so a break rarely means downtime, just temporary manual work while I fix it.
Two layers of coverage. First, my delivery network — the specialists I already work with — can maintain any running system without me for weeks. They know the stack because they helped build it. Second, all systems use standard, well-known tools (LangGraph, Python, Postgres, n8n), fully documented, so any senior engineer can pick things up cleanly. You’re never dependent on my calendar for a system that’s already live.
Cancel Care Plan any month, keep the whole system running. You have the source code, credentials, docs, and cloud access from day one. No lock-in, no ransom fees, no “premium API keys” you have to keep paying me for. If you find someone cheaper to maintain it, hire them — I’ll do the handoff call for free.
04 — CHAPTER
Almost never in the first year. It removes the boring 40% of their day (data entry, follow-up chasing, filing) so they can focus on the 60% that actually needs a human (relationships, judgment, exceptions). Most clients keep the same team but stop needing to hire the next one as they grow.
The whole design goal is that they don’t have to. The system runs behind Gmail and their phone. If something needs their attention, it shows up as a normal email or text with two buttons: “approve” or “review”. No dashboards to learn, no logins to remember, no new app on the phone.
Sometimes, if they think it’s about replacing them. That’s why the first thing we do in Discovery is map out what stops being their job (the parts they hated anyway) versus what stays theirs. Clients who introduce it as “this is the tool that makes your job easier” get zero pushback. Clients who introduce it as “AI is replacing headcount” get exactly the pushback they earned.
05 — CHAPTER
2-4 weeks for most builds. 30-min discovery → 1-2 days architecture (you approve the blueprint before I write a line of code) → 1-3 weeks build with daily updates → deployment + handoff. Industry average is 8-16 weeks; I move faster because I don’t do committee reviews and I don’t context-switch across 10 clients.
A 30-45 min call where you describe how jobs come in and where they leak. I ask specific questions, take notes, and by the end I tell you: what’s worth automating, what isn’t, what it would cost, and how long. If we’re not a fit, I say so. If we are, I send a written scope within 24 hours. No pitch deck, no follow-up sales pressure.
Roughly 3-5 hours total across 2-4 weeks. Kickoff call (1 hr), read-throughs of the architecture doc (30 min), 2-3 short check-ins on decisions only you can make (15 min each), UAT before launch (1 hr), handoff (30 min). Everything else runs in the background.
By then the system is stable and you have your first month of numbers. We do a 30-day review call: what’s working, what to improve, what to add next. If you want to continue with Care Plan, we do. If not, you have everything documented and can maintain it yourself or hand it to any developer.
06 — CHAPTER
Neither. I’m the senior engineer and product lead on every project — you work with me directly, no account managers, no handoffs. Behind me is a small trusted network of specialists I bring in when a build needs them: a design assistant, a second developer for parallel work, a QA reviewer for critical launches. Upside: senior thinking on every decision, plus enough hands to move fast. Limit: I take 2-3 new engagements per quarter.
Construction firms, design-build studios, home-service companies (HVAC, plumbing, electrical, roofing, landscaping, cleaning, moving), interior designers, remodelers, small commercial contractors. Sweet spot: 5-50 people, owner still involved in day-to-day, average project between €500 and $500K, at least one thing is clearly broken in operations.
I work across US, UAE, and EU timezones with same-business-day response. In 2026-2027 I’ll be spending significant time in Dubai (AI events, Design Week) and Milan (AI Week). If you’re in the Middle East, we can meet in person.
Yes — I’ll share 2-3 depending on which module we’re scoping. Some client work runs under strict NDA and I can only give you screen recordings, not names. If direct references matter to you more than case data, tell me in Discovery and I’ll match you with someone who allows a call.
You work directly with the person who designs and writes the system — no account-manager layer, no junior engineers learning on your project. That means direct code review, one fixed point of responsibility, and an architecture built to be handed over: you own the source code, credentials and documentation from day one, so you’re never locked in. A large agency bills $50-500K over 6-12 months and splits the work across a team you never meet; here one senior builder owns the outcome and ships in weeks.
07 — CHAPTER
WhatsApp, LinkedIn, or Telegram — pick your favorite. All three are on my phone, same-day reply. Email works too. My response time in business hours is under 2 hours, worst case next morning in your timezone.
If you’re in Dubai, Milan, Singapore, or anywhere I’m attending an event — yes, glad to. Otherwise everything runs remotely: Loom videos for walkthroughs, WhatsApp for daily updates, Google Meet only when we really need it. Not because remote is trendy, but because it’s faster and clients get more done that way.
08 — CHAPTER
Yes. The modules aren’t country-specific. Same lead-to-invoice logic works anywhere — adapted to your CRM, currency, language, and local tools (WhatsApp Business, Bitrix24, Zoho, whatever). Systems have been built for clients across multiple regions and languages.
Yes. Every build includes a data-handling document listing where data lives, who has access, retention rules, and compliance mapping to your applicable regulation (GDPR for EU/UK, PDPL for UAE, CCPA for California, etc.). For heavily regulated industries (healthcare, finance) I bring in a compliance reviewer at cost.
Two options. (1) Book a Discovery call — it’s free, and by the end you’ll have a real scope even if we don’t work together. (2) Ask for a “pilot module” — I build the smallest useful piece for $1,500-3,000 in 5-10 days. If it works, we scale into a full build and credit the pilot against it. If it doesn’t, you have something small that still works and we part on good terms.
Send it over. You’ll get a direct answer — even if the answer is that I’m not the right fit.