Why we charge by scope, not by hour, for AI projects

TL;DR

• AI projects fail on scope drift, not hour counts.
• Hourly billing rewards ambiguity. Scope billing forces clarity up front - the expensive part.
• Our projects are fixed-price after a scoping call + written assessment.
• Retainers are the exception: monthly capacity for ongoing iteration.
• Change orders are documented, not hidden.

---

The problem with hourly on AI work

Hourly looks safe. You pay for what you get. Reality:

1. The scope isn’t defined. Most AI briefs start as “we want AI to do X.” Scoping X takes 30 to 50 hours of thinking. On hourly, that’s 30-50 hours of billable ambiguity - and the client is footing it before anything ships.
2. The vendor is rewarded for slow clarity. Every hour the scope stays fuzzy is another hour billed. That’s not malice - it’s incentive design.
3. The client can’t compare vendors. “100 hours at €120/hour” vs. “fixed €10K for the same deliverable” - which is cheaper? You don’t know until the project is over.
4. Timeline slips silently. No fixed end-date means no pressure to protect the ship date.
5. Post-launch tuning gets conflated with build. “We’re still tweaking” turns into month three of billable tweaks.

For AI projects specifically, scope drift is the main failure mode. The technology is new. The workflows are new. The data is often dirtier than the first call suggests. Hourly structurally amplifies this. If you’d rather hire AI developers on a fixed shape than an open clock, this is the structural reason.

What fixed-price forces

Writing a fixed price on day zero forces us to write the scope down on day zero. That conversation - “what exactly are we building, what are we not, and what does done look like?” - is the work AI projects usually skip and then fail on.

A fixed price also shifts risk the right way. The vendor is best positioned to estimate vendor work. Clients shouldn’t carry schedule risk on our internal velocity.

How we do it

Step 1 - Free scoping call (30 min)

You describe the problem. We ask blunt questions. You walk away with a one-page written summary within 24 hours, whether or not we engage further.

Step 2 - AI Readiness Audit (optional, paid)

For larger projects, we quote a 1-2 week audit (from €2,500). Deliverable: a 10-15 page assessment with a phase-one build spec, ready to quote. You’re free to take it to another vendor.

For larger programs - an agentic systems build, a multi-domain data cleanup, a full custom product - the audit is usually how we de-risk the fixed quote that follows.

Step 3 - Fixed-price build

After audit (or after the free scoping call for smaller projects), we quote:

• Scope - what’s in, what’s out.
• Method - how we’ll build it.
• Timeline - weeks, not “we’ll see.”
• Price - fixed.
• Deliverables - named artifacts.
• Change-order process - written upfront.

Step 4 - Change orders, documented

If you change the scope mid-build, we quote the delta. Clear, written, separate from the main invoice. No surprises.

Step 5 - 30 days of post-launch fixes

Included in every project price. Anything that’s in the original scope and doesn’t work: on us. Anything new: change order.

The one exception - retainers

For ongoing work (agent tuning, data stewardship, site iteration, feature additions), fixed-price doesn’t fit. The shape is “keep shipping,” not “ship this one thing.”

So we run retainers at €990 / €1,990 / €3,990 per month (20 / 40 / 80 engineering hours). Retainer hours aren’t generic - we still scope each workstream. You see what’s in-flight, what’s queued, what’s complete. Hours roll over up to 25%.

A typical client arc: fixed-price project (weeks 1-8) → retainer (month 3 onward). The project ships the system. The retainer iterates it. Most of our ongoing custom AI development work lives in the retainer half of that arc.

What fixed-price doesn’t work for

• True R&D. If the question is “can this even be done?” we’ll scope it as a time-boxed spike (usually 1-2 weeks, fixed fee, explicitly no deliverable guarantee). Rare in SMB work; common in moonshot work we don’t take.
• Scope that can’t be written down. If after two scoping conversations we can’t commit to a shape, we won’t quote fixed. We’ll recommend an audit first, or decline the project.

Why this is good for you

1. You know what it costs. Before you commit.
2. You know when it ships. Before you commit.
3. You know what you’re getting. Before you commit.
4. You can shop it around. Our spec is yours to take elsewhere.
5. Our incentives align with your outcome. We’re rewarded for shipping, not for hours.

Why this is good for us

1. We only take work we can scope. Filters for quality briefs.
2. We get better at scoping every project. Written specs compound into templates.
3. We stop billing for ambiguity. That’s someone else’s business model.
4. We ship. Which is what we actually enjoy.

---

Got a project that needs a scope? Scope a project →

See the full pricing sheet → /pricing