SBAI · Kuba Parol

AI that works.A team that gets it.

I advise and ship production AI for companies of 30-300. Then I leave behind a team that can grow it - without me.

human-in-the-loopproduction AIaudit · RBAC · EU hosting
20+years in enterprise software
3AI products built
millionsinvoices / month on systems I've architected
27,000sensitive contracts / month in OCR + RAG pipelines

01 · Problem

Sound familiar?

Everyone's talking about AI, but I have no idea where to start in my own company.
Someone sold us a chatbot for EUR20k. It's gathering dust.
I'm scared the competition will figure out AI before we do.

All three are solvable - but not with a deck, a workshop, or yet another PoC.

02 · Services

Three things I'm good at.

Production AI builds

RAG, OCR, agents, mail bots, search.

I build solutions for a specific problem - not platforms. EU hosting, audit log, RBAC, compliance-ready from day 1.
  • 4-12 weeks brief to production
  • Your stack, your data
  • Full docs + runbooks

AI-coding adoption

Your team keeps shipping after I leave.

I onboard your team into AI agents: Cursor, Claude Code and custom workflows.
  • 2-6 weeks hands-on
  • Reusable standards + templates
  • Sparring + code review

03 · Process

Four steps. No theater.

01

I listen

90 minutes on Zoom, free. I ask 20 questions and check whether I can actually help.

02

I map

1-3 days of workshops and analysis. Concrete scope, estimate, risks, fixed fee.

03

I build

4-12 weeks. Weekly demos on real data and model decisions visible from the start.

04

I teach and leave

Knowledge transfer, runbooks, monitoring and 30 days of availability after handoff.

04 · Case studies

Not logos on a slide. Working systems.

01

Global SaaS · Accounts Payable · NDA

Agentic AI assistant for AP teams.

Multi-agent architecture for invoice status, payment terms and escalations. Event-driven on Azure, human-in-the-loop and production observability.

  • ~70% auto-resolved
  • minutes to first response
  • 10+ languages
02

Insurance · OCR + RAG · GDPR

27,000 sensitive contracts a month.

OCR, extraction, vector index, RAG and PII redaction before LLM calls. Audit log and RBAC down to clause level.

  • 27,000 contracts / month
  • ~90s to indexing
  • full audit trail
03

Own product · SignalBoy.ai

AI-powered content triage and semantic knowledge platform.

My own lab for agentic architectures: multimodal RAG, conversational search and automated content triage.

  • live product
  • multimodal RAG
  • agentic pipeline

Lab

What I build when I'm not working for clients.

I don't just write about AI. I use it every day.

SignalBoy.ai

AI knowledge management. Save once, actually use it later. Live and shipped.

YagaAi

Multi-agent runtime on Azure for autonomous software dev. Research grade.

VibeRails

Standards and guardrails for teams working with AI coding agents.

I'll help if...

  • You're a 30-300 person company with concrete operational pain.
  • You have an internal tech team I can pair with.
  • You want a working system, not another PoC for the drawer.
  • You value honesty over 'yes, of course'.

I won't help if...

  • You're chasing the cheapest quote.
  • You want AI transformation without use case or ROI.
  • You're a 2-person company and ChatGPT Plus is enough.
  • You need someone in-house on payroll.

About

Kuba Parol.

20+ years in enterprise software. Recently Head of AI / AI Architect for a global SaaS in Accounts Payable.

Before that, 12 years as co-founder and tech lead of a financial controlling SaaS used by leading Polish companies.

I started SBAI because SMEs in PL and EU need production AI, not demos - and someone who can build the system and leave a capable team behind.

FAQ

Questions I hear most often.

How much does it cost?

Audit + strategy: EUR2-6k fixed. Production build: EUR15-60k. AI-coding adoption: EUR6-20k.

How long does implementation take?

4-12 weeks from brief to production. First demo on your data usually in 2-3 weeks.

Is my data safe?

Yes. EU hosting, PII redaction, zero data retention models, NDA and DPA as standard.

Do you train models on our data?

No. Never. Models run in no-retention mode.

What happens after handoff?

Docs, runbooks, monitoring and 30 days of contact stay with you. Longer SLA is separate.

Let's talk

30 minutes. No deck.

Tell me what you want to achieve. If I can help - I'll tell you how. If I can't - I'll tell you exactly why.