HomeServicesAI Strategy & Data Readiness
Service 05 · AI Strategy & Data Readiness
Data infrastructure for the AI era
We prepare your data infrastructure for the AI era. From EU AI Act compliance to building first-party datasets for custom LLMs and predictive modeling.
First call within 48 h · 2 weeks · from €4,500 net
You're in the right place if:
Smart Bidding, Advantage+, or HubSpot AI run without use-case documentation.
Nobody knows how many AI components are already in the marketing stack.
There's no risk register for the AI systems in use.
Data ownership is unclear, PII sits in unclassified datasets.
What we build
Three pillars. One AI foundation.
Use-case assessment, data architecture, and governance, bookable individually or as a chain.
AI use case mapping & risk assessment
Custom Assessment Framework · EU AI Act checklistsWhich AI systems do you operate, knowingly and unknowingly? Many companies don't know how many AI components are already in their marketing tech stack.
Inventory of all AI systems including hidden ones like Smart Bidding, Advantage+, HubSpot AI, risk assessment per EU AI Act per use case, gap analysis, and a prioritised action plan.
First-party data architecture for AI
BigQuery · Snowflake · dbt · OneTrustAI models are only as good as the data they run on. Fragmented data produces poor outputs, regardless of how good the model is.
Audit of the data architecture for AI readiness, event-schema optimisation for ML inputs, consent-architecture check, recommendation for a data-lake setup if not in place.
AI governance & human oversight
Custom policy templates · EU AI Act frameworkWho decides what, and who reviews AI outputs before they inform decisions? These are process and organisational questions, not purely technical ones.
AI policy document under applicable EU law, human-oversight process per risk level, AI literacy workshop (half-day), documentation template for EU AI Act compliance.
How it works
From inventory to governance, in four steps.
First know what's running, then classify, then secure the data foundation and the processes.
Inventory
Capture every AI system in use, including the hidden ones like Smart Bidding, Advantage+, and HubSpot AI.
Classify
Each use case is rated per EU AI Act, with a gap analysis of compliance against current state.
Data foundation
Bring first-party data into an ML-ready schema, check the consent architecture, because AI only trains on consented data.
Governance
AI policy, human-oversight process, and documentation template, matched to each risk level.
Process
Four phases, fixed order.
always starts with phase 1 · no blind build
Audit Sprint
Five layers audited, findings ranked, effort estimated. The result is a report you could act on without us.
Architecture
Data contract, event design, target architecture. We fix where each number is produced and who guarantees it.
Build Sprint
Delivery in sprints, every module signed off on its own. Your team stays involved, not locked out.
Managed Evolution
Monitoring, release support, platform updates. Optional; plenty of clients run the setup themselves.
What you get
A report, not a workshop afterglow.
The Audit Sprint ends in a document: findings per layer, severity, effort, sequence. Not a slide deck full of recommendations in the subjunctive.
→ Findings with severity and reproduction path
→ Effort estimate per finding, in person-days
→ A draft data contract for the core events
→ An implementation plan another agency could execute
Deliverables · AI Strategy & Data Readiness
Structure taken from this page's scope of delivery. The concrete scope comes out of the audit.
Scopes
Three ways in, one starting point.
Audit Sprint
from €4,500 net
2 weeks
We audit what is wrong. Prioritised report + action plan.
Request an Audit Sprint →Managed Evolution
monthly
3-month minimum
Ongoing partnership. Analytics as a product.
Request Managed Evolution →Yes, Smart Bidding is an AI system that makes bidding decisions, so the EU AI Act applies. Under current guidance it doesn't fall into the high-risk category, but transparency obligations and internal documentation are still recommended. We clarify in the assessment exactly which use cases trigger which requirements.
When an AI system makes a decision, for example which segment sees which offer, a human has to be able to review that decision and override it when needed. The process doesn't have to be heavy; often a documented review routine is enough.
First-party data is data you collect directly from your own users, with consent, on your own infrastructure. Unlike third-party data (cookies from outside providers), it's available long-term, usable under GDPR, and qualitatively more reliable. AI models trained on first-party data are more stable and produce better results than those trained on aggregated third-party data.
The Act applies in stages. Prohibited practices have been banned since February 2025, and General-Purpose-AI obligations since August 2025. The Article 50 transparency obligations have applied since 2 August 2026. The high-risk obligations were deferred by the Digital Omnibus to 2 December 2027, and to 2 August 2028 for AI embedded in regulated products. Companies that move now avoid last-minute compliance pressure.
Next step
EU AI Act: where does your setup stand today?
Strategy call about compliance, first-party data, and custom-LLM foundations. Full-cycle implementation together with Saloid.
