With a SaaS, you pay per user, every month.
You pay for the system, not the seats.
A fixed investment up front and a predictable server cost. Growing doesn't raise the bill.
Artificial intelligence agency · Spain
AI agents and internal software, inside your own server.
We build AI agents, n8n automations, RAG systems, custom CRMs and dashboards, deployed on your own infrastructure. Fixed scope, delivered in 4-8 weeks.
- academy
- −€24,000/yr
- manufacturer
- −€39,000/yr
- to deliver
- 4-8 wks
NowTechnical agent · generating report-DR50.pdf
A SaaS rents. Your own system works.
Three things that change when the software is yours and runs on your infrastructure.
With a SaaS, your data lives in their cloud.
Your data stays at home.
Deployed on your server or with a European host you choose. GDPR by design.
With a SaaS, you adapt to their process.
The software adapts to you.
Stages, permissions and automations shaped exactly like your operation.
What we build
Internal software with AI deployed on your own server. Fixed scope, sovereign data and one method for everything.
—Architecture
Inside, layer by layer.
All our projects share the same structure. What changes is what sits in each layer, not where it lives: on your server.
01 · Email · ERP · CRM · sheets · PDF
Your sources
What you already use: email, ERP, CRM, spreadsheets, catalogues and documents. No painful migrations.
Gmail
Google Sheets
HubSpot
Odoo
WhatsApp
02 · n8n · webhooks · APIs
Ingestion and automation
Workflows that collect, clean and move data between tools, with retries and alerts when something fails.
n8n
Webhooks
FastAPI
03 · PostgreSQL · pgvector
Data and knowledge
Your own database and a vector index so the AI answers from your documents and cites the source.
PostgreSQL
pgvector
04 · LLM via API or private model
AI agents
Models that reason and use tools under permissions. Sensitive actions wait for your approval and everything is logged.
Claude
Google Gemini
Mistral AI
Ollama
05 · CRM · dashboards · reports · chat
Interfaces
What your team uses every day: a custom CRM, a dashboard, Monday's report or an internal chat.
Next.js
React
Telegram
Real cases, real numbers
Context → decision → result.
Education
Training academy: 3 SaaS tools → 1 internal system
- Context
- They paid for three subscriptions that almost covered student intake, payments and follow-up, but none of them did it well.
- Decision
- Cancel the three SaaS tools and build a custom internal CRM and dashboard, with the data on their own server.
- Result
- One system aligned with their process. A less fragmented team and a predictable fixed cost.
−€24,000/yr
saved
Manufacturing
Manufacturer: a technical RAG agent
- Context
- Hydraulic know-how lived in two or three people's heads. Quoting a tank meant hours of calculations and catalogue lookups.
- Decision
- A RAG agent over the catalogue and internal standards: it calculates, selects equipment and generates a PDF report.
- Result
- Answers in minutes, full traceability and zero data outside their infrastructure.
−€39,000/yr
saved
Simulation · 40-user SME
40-user CRM: SaaS vs custom software
- Context
- SaaS licences + consultants + the risk of price increases over three years: €160,200.
- Decision
- Build the internal CRM in 4-6 weeks, with full ownership of the data and the code.
- Result
- Estimated investment of €48-58K over three years. A difference of roughly +€100,000 in favour of custom software.
+€100,000
difference over 3 years
The 3-year maths
SaaS + consultants
€160,200
Custom software
€53,000
actual range €48-58K
| Item | SaaS | Custom |
|---|---|---|
| Licences / year | ~€35,000 | €0 (own hosting) |
| Consultants / setup | ~€20,000 | Included in the project |
| 3-year total | €160,200 | €48-58K |
| Data ownership | With the vendor | On your server |
| Price increase risk | High | None (you own the code) |
| Difference | +€100,000in favour of custom software in the 40-user scenario | |
The 4-week process
A fixed framework. Delivered in 4-8 weeks depending on complexity.
Week 1
Diagnosis
We map the real process, the SaaS tools you pay for and the pain points. We agree a fixed scope: what is in and what is out.
Week 2
Infrastructure and ingestion
Server, database, pipelines and connectors. Your data stays in your infrastructure.
Week 3
Agent / business logic
CRM, RAG or dashboard: the business logic, permissions and automations.
Week 4
Validation and deployment
Testing with your team, adjustments and go-live. Delivered with operational documentation.
✓ Fixed scope from day one. No "we'll see as we go".
When NOT to hire me
Sometimes SaaS is the right answer. Better to say it on the website than on the third call.
Your process is not stable yet
If you reinvent the workflow every week, a generic SaaS is cheaper while you experiment. Custom software freezes what already works.
Only 1-2 people need it
With a very small team, a SaaS licence usually beats investing in an internal system.
The SaaS already does 90% well
If you are only missing two fields, do not build a CRM. Configure or integrate. Custom software is not a vanity project.
There is no real project budget
Four in ten calls end with: do not commission custom software. Not yet. I would rather say it plainly.
Typical stack and connectors
- n8n
- PostgreSQL
- pgvector
Docker- FastAPI
- React
- Next.js
- OpenAI
Anthropic- Mistral
- Google Gemini
- Open-weight models
- Gmail
- Outlook
- HubSpot
- Odoo
- Holded
Notion- Slack
- WhatsApp Business
Stripe
Names and logos are trademarks of their respective owners. They are shown only to identify the technologies we work with and do not imply any partnership, sponsorship or endorsement.
- GDPR by design
- Data on your server or in the EU
- Code in your repository
- AI transparency (EU AI Act art. 50)
- Data processing agreement
Frequently asked questions
What exactly does InnoAIConsulting do?
We are an artificial intelligence agency that builds internal software for SMEs: AI agents, n8n automations, RAG systems that answer from your documents, custom CRMs and dashboards. We deploy it on your own infrastructure, with a fixed scope and delivery in 4-8 weeks.
Where is the data hosted?
On your server or with a European provider you choose. If a project uses external AI models, the diagnosis defines which data may leave and which may not; for sensitive data we use private models hosted on your infrastructure.
How long does a project take?
Between 4 and 8 weeks depending on complexity. Week one is the diagnosis and closes the scope: what is delivered, what is out and how it is validated. After that there is no "we'll see as we go" phase.
How much does it cost?
It depends on the scope, which is why we don't publish rates. After a 30-minute diagnosis call we give you a fixed price. If a SaaS tool would serve you better, we tell you: that happens in four out of ten calls.
Who owns the code?
You do. It is delivered to your repository with documentation and runbooks so your team or any provider can maintain it. No per-seat licences, no lock-in.
Are you GDPR and EU AI Act compliant?
Yes. We design for GDPR from the start, sign a data processing agreement and, when a system talks to people, it says so clearly as article 50 of the EU AI Act requires. Sensitive actions require human approval and are logged.
Tell me which SaaS you pay for and what your team does by hand
30 minutes. If custom software is not a fit, I will tell you.
Message
Reply within one business day.
Or book directly
Discuss your case in 30 minutes. No commitment.
Calendly · 30 min · diagnosis