Key takeaways
- ESADEX’s usual AI and platform project budgets range from €25,000 to €200,000, depending on scope.
- Integrations and compliance scope move the price more than the model does.
- Get a written scope with a fixed estimate before any build starts.
- Start with one workflow and measure it, and budget for running costs as well as the build.
Budget ranges
Our usual project range is €25,000 to €200,000. These are ESADEX planning bands, not a survey of AI prices across Spain. Your estimate depends on the workflow, integrations, hosting and required controls. The tables show illustrative scopes, not quotes or guaranteed prices.
| Range | Typical scope |
|---|---|
| < €25k | Website, landing pages, small internal tool |
| €25k to €75k | One AI workflow (RAG assistant or STT pipeline) with standard integrations |
| €75k to €200k | Multi-workflow platform, SSO, on-prem deployment |
| €200k+ | Multi-tenant SaaS or enterprise-wide rollout |
What drives cost
- Number of systems to integrate with
- On-prem vs cloud deployment
- Risk class under the EU AI Act
- Languages and accessibility requirements
How to scope a first version
- Pick one workflow with a clear owner and a way to measure success.
- List the systems it must connect to and the data it will touch.
- Decide where it must run and which rules apply to the data.
- Define what a good result looks like in numbers you can check.
- Plan the first increment to be small enough to ship in two weeks.
What compliance adds, and what it saves
Compliance work is real cost: a risk classification, an impact assessment where one is needed, audit logging, human review steps, documentation and redaction of personal data before any model call. Built in from the start, these are a modest share of the project. Added after the first audit, they usually mean rebuilding parts of the system.
| Item | Built in from the start | Added later |
|---|---|---|
| Audit logging | A design decision | Rework of every component that handles data |
| Human review step | A screen in the workflow | A new process and often a new screen |
| Redaction before the model | One pipeline stage | Changes to storage, search and prompts |
| Documentation | Written as the system is built | Reconstructed from memory and code |
What you pay after launch
- Hosting and model usage, which follow volume, or hardware if you run on-prem
- Monitoring and support, so that someone notices when answers get worse
- Updates to models, dependencies and security patches
- Content upkeep: documents, glossaries and rules change as your business does
Questions to ask any vendor
- What exactly is in the fixed estimate, and what would change it?
- Where will the data be stored and processed, and who are the sub-processors?
- Which risk class does the system fall into, and what does that require?
- What do hosting, model usage and support cost each month after launch?
- Who owns the code and the trained assets, and can we move away?
- How will we know it works, and what will you measure?
Red flags in a quote
- A price with no written scope or list of integrations
- No mention of personal data, hosting location or logging
- A promise of 100% accuracy
- No running-cost estimate and no handover of code
Send us your workflow, integrations and hosting requirements. We will scope the first version and provide a written estimate, usually in about five days.
Request an AI project estimateFAQ
Is there a cheaper way to start?
Yes. Scope one workflow, ship it in a few increments and measure the result before you expand.
Why is the range so wide?
Because scope varies. A single assistant over one document set and a multi-tenant platform with SSO and on-prem deployment are different projects, even if both are called an AI system.
How long does a project take?
A scope and architecture document arrives in about five days. Websites usually ship in weeks, while AI systems and platforms are delivered in two-week increments, with timelines agreed in the scope.
Sources
- ESADEX project budget ranges
- Regulation (EU) 2024/1689 (AI Act), risk classification, Official Journal of the EU
About the author
Erik Salo Akimov · Co-founder · Engineering
Erik runs engineering at ESADEX. He designs the platforms, infrastructure and security controls that sit behind client systems.
