// Julien Casse 🇫🇷

I find where AI can move your business, and I build it to last.

Not a demo. Not a chatbot. Something that works on Monday. And is still working a year later.

Request your auditThe work
// The Reality

You don't need more AI. You need the right AI.

Every business I look at has the same pattern: a handful of genuinely valuable automations, surrounded by a dozen shiny demos that'll never pay back. The work isn't building more. It's seeing clearly.

01

Everyone wants AI. Nobody knows where.

You feel it's coming for your business. You're not sure which 20% of the work is the 80% of the leverage.

02

Demos don't survive production.

Prompt chains that worked in a notebook quietly fall apart when real customers, real data and real edge cases arrive.

03

Tools sprawl. Nothing connects.

Notion, spreadsheets, raw CSVs, a Zapier bill. Work lives between the cracks and nobody owns the seams.

04

Spend goes up. Outcomes don't move.

ChatGPT Pro across the team, two SaaS tools with 'AI' bolted on, a couple of workflows nobody's measuring. The invoice is clear. The lift isn't.

// What an AI win looks like

Not chatbots. Not co-pilots. Leverage you can measure on Monday.

Wins are boring and specific. They remove a human step. Connect two systems. Collapse a queue. That's it. Five shapes. They cover most of what I see:

◆Lead capture & qualification

Leads enriched, scored, and first-drafted, before you open your inbox.

Form submissions, prospect lists, outbound replies: all pass through enrichment and fit-scoring, with first-contact drafts ready to personalise. You see the leads worth answering, already prepped.

BeforeManual scoring, inbox overwhelm, leads going coldAfterPre-scored, pre-drafted, inbox as a prioritised queue
◆Client & service ops

Booking, follow-ups, notes: the stuff that slips when you're busy.

The repetitive layer of high-touch service (scheduling constraints, client-specific memory, follow-up loops) done in a tool with triggers, not by hand.

BeforeSpreadsheet + calendar + memory + Post-itsAfterOne tool with constraints, memory, and triggered follow-ups
◆Targeted outbound

Prospecting that doesn't live in a spreadsheet.

Public data (registries, social graph, web) wrapped into research your team can actually use. Targeted contact lists with context built into each first message, not another list of names.

BeforeStatic lists, manual research, generic messagingAfterResearched targets, context-aware first messages
◆Always-on front desk

The call you can't take. The DM at 11pm. Caught, qualified, booked.

An AI receptionist at the edge of every channel: calls, DMs, after-hours forms. Qualifies intent, books the slot or passes the message on with context. Works the hours and channels you can't.

BeforeVoicemails, "I'll call you back", deals going to the next name on the listAfterEvery inbound caught, qualified, slotted or routed with context
◆Internal platform

One tool replacing five.

Spreadsheets, Notion, a CRM trial, three Zapier zaps and a side script. Consolidated into one internal platform with AI baked in. One login, one source of truth.

BeforeFive tools, three Zapier zaps, a spreadsheet, no source of truthAfterOne internal app, one login, AI in the workflow
// The work

Five projects. Same methodology.

Before-and-after for each one. Swipe through, or open a case for the full story.

case · internal os

Internal OS for an AI-native dev agency

Before

Lead management and project management both eating the week. Raw notes first, then Notion + Linear, with client comms scattered across messages and emails. A decision in one tool would contradict a thread in another. No single point of truth.

After

An internal OS, not a CRM with a project tool bolted on. CRM and project management in one system, leads and projects properly linked. Then layered with AI: logged interactions, lead scoring at intake, drafted client responses, smart sprint generation.

Read the full case
case · pet sitter

Booking + memory for a pet-sitting service

Before

Handwritten notes. Intro days double-booked. You can't host two new dogs arriving the same hour. Follow-ups slipping. Owners re-asked the same questions stay after stay.

After

Booking with intro-day constraints baked in. Pet-specific notes auto-attached to each reservation. Stay-event follow-ups starting to replace memory.

Read the full case
case · lead gen

Lead generator for a freelancer

Before

Manual lead hunting. Cold outreach without context. A half-finished CRM of spreadsheets and copy-pastes.

After

A private app that finds and enriches prospects from public sources, scores them for fit, then drafts the first contact, ready to personalise and send. A lead channel that didn't exist before.

Read the full case
case · online presence

Content engine for a local service

Before

Cold outreach and phone calls doing all the work. No inbound channel. The website was a placeholder, just enough to hold a domain.

After

Site rebuilt with a content engine running behind it. Articles drafted, reviewed, published on a calibrated cadence. A simplified CMS shaped for his actual workflow: review, refine, approve. Inbound from search keywords nobody else was targeting.

Read the full case
case · saas prospecting

Prospecting engine for a SaaS

Before

An Excel spreadsheet. Manual research per target. Slow, error-prone, impossible to scale.

After

A dashboard wrapped around Pappers lookups, Facebook mapping and OSINT-style enrichment. Targeted outreach with context built in. The spreadsheet era ended.

Read the full case
// How we work together

Two paths.

One problem to solve, or a standing seat at your side. Both use the same methodology. The scope is what changes.

mode_01

The Project

You bring a problem. I ship the answer.

One discovery, one scoped build, one shipped system. Documented, handed over, yours. What most agencies sell, done properly.

When you know the shape of the win, and you need it built.
mode_02

The Residency

An AI expert at your side. Standing.

Ongoing access to me as your embedded AI expert. Almost always, we build a dedicated application tailored around your business, and keep evolving it as you do. Artisanal AI: one client, one system, alive.

When AI is continuous in your business, and the system needs to grow with you.
// How we work

Four steps. No discovery theatre.

Same rhythm whether it's a Project or a Residency. What changes is the scope, not the how.

01 /
Conversation
You tell me what you're running on, and where you want more: time, money, leads. I listen. No discovery script.
a call
02 /
Research
I go deep on your context, your stack, your data. I come back with the AI moves that are actually yours.
~a week
03 /
Scope
Final questions, a written draft, a quote. You see what we'll build before anything is built.
days
04 /
Ship
First version in days. We iterate from something working, not from a Notion page about what we'll do someday.
days → live
// About

A decade of shipping. Now artisanal AI.

Julien. A decade of shipping software where mistakes had real consequences: regulated finance, a scaled SaaS, institutional clients. Then I went deep on AI inside real codebases and found the pattern: without foundations, AI creates silent chaos; on top of discipline, it's a multiplier. Now I point that discipline at business workflows and build the AI worth building. Artisanal, not assembly-line.

// Also founder of itzworking.io, where the same discipline ships full projects with AI.
Read the full story
// Questions

The honest answers.

Any business that wants AI to actually move a number: margin up, cost down, more shipped, less manual work. SMBs running real operations, agencies of any flavor, founders past day-zero, operators running something they care about. Not pre-product founders chasing funding, not enterprise RFPs.

You don't need a spec. The Conversation + Research phases are built for that. You come with the business and its frictions; I come back with the AI moves that fit.

First version in days, full build in weeks. The Residency is ongoing, continuous shipping. Nothing takes months unless it shouldn't be done at all.

No. For a Project, I ship and hand over with documentation. For a Residency, your team can plug in or stay out of it. Both work.

See all questions
// Your move

Find your win. Then ship the one that matters.

One conversation. One scoped plan. Something working in days. Project or Residency. Same rhythm.

Request your auditHow we work
→ Apply to start→ EU-based→ EN / FR→ < 24h reply