Book Discovery Week
Investment · Priced by scope

Priced by scope. Not by hour.

You see the cost before we start. Each stage has an exit point. The audit fee is fully credited toward MVP if you proceed within 60 days. Most engagements land between €15k and €30k for the first agent live in production, plus a maintenance retainer thereafter.

01What you're actually paying for

An AI agent is not "one model with a long prompt." It is six layers, each designed, integrated, tested, and maintained for the agent to stay useful past month two. Every Hi-Ai engagement builds and operates these layers explicitly. The prices below buy you the layers — not just the base model.

L · 01FoundationWhat it is

Base reasoning model + system prompt. Defines role, capabilities, refusal boundaries. Picked to match the task — Claude, GPT, or open-weight.

ModelSystem promptRefusals
L · 02KnowledgeWhat it knows

Internal docs + source-of-truth systems + external lookups. Chunked, indexed, retrieved with permissions intact. Citations on every answer.

RAGPermissionsCitations
L · 03GuidelinesWhat it can do

Business rules, decision logic, escalation thresholds, scope boundaries. Encoded as evaluable constraints — not vague paragraphs.

PolicyScopeEscalation
L · 04Context memoryWhat it remembers

Session state + conversation history + long-term memory. What carries across turns and days — and what is explicitly forgotten.

SessionLong-termForget rules
L · 05Brand voiceHow it speaks

Your terminology, tone, formatting, and the things you'd never say. Pre-publish review against forbidden phrases and required disclosures.

VoiceFormatDisclosures
L · 06Eval harnessHow we know it works

Reference dataset of golden questions, weekly automated runs, drift monitoring. Without this, none of the rest is maintainable.

Golden setWeekly runsDrift alerts

Agency teams are multiple agents like this coordinating with each other — for example a sales triage agent calling a knowledge agent calling a CRM-update agent — under a shared harness that monitors the whole flow. Each agent has its own L01–L06; the team adds an orchestration layer on top.

02Stage prices

Discovery Week €0One week · Written roadmap
  • 15-min questionnaire
  • 30-min discovery call
  • Shadowed working session
  • Written roadmap, ranked by ROI
No commitment · ~33% decline
Foundation Audit €2,8002 weeks · Tech spec + integration scaffolding
  • Source-system integration
  • Permissions + observability
  • Finalised MVP technical spec
  • Hidden-debt surfaced early
Fully credited toward MVP within 60 days
Scale-Out €8–14kPer additional agent · Faster than first
  • Foundation + harness reused
  • Second agent ~60% of first cost
  • Composable system, not project
  • Per-agent scope & exit
Scope priced before each ship

03Maintenance & Optimization

Three tiers, all month-to-month with 30 days cancellation.

Light

€1,400 / mo

Production monitoring, error alerting, monthly evaluation harness re-runs, source-document refresh. Suitable for one stable agent that does not need active iteration.

Core

€2,400 / mo

Everything in Light, plus quarterly tuning sessions, prompt + retrieval iteration, and discovery work for the next agent. Most ongoing engagements sit here.

Plus

€4,800 / mo

Everything in Core, plus a half-time dedicated engineer assigned to your account, a shared Slack channel, and a 4-hour SLA on production incidents.

04Pricing policies

05What this looks like in year one

For a typical engagement — one MVP, then a Core retainer for the remaining 10 months of year one — the all-in cost lands at €36,000–€48,000 (€15–18k MVP + €2.4k × 10 months retainer + €2,800 audit, less the audit credit). Most clients see 3–5× this number in recovered hours, lift in pipeline, or reduced support cost in the same year.

For the math behind that calculation, see the field note The €40,000 question.