08 · AI, processes, CRM and sales
AI solutions
AI that does one job well inside your process, with clear limits and a person in the loop where it matters.
In short
I work with AI agents every day: they draft content under house rules, fix code from scan findings and read my dashboards through an API. What that has taught me is that AI works when the job is narrow, the rules are written down and the output is checked.
That is how I would build AI into your business: one job at a time, connected to your systems, with a person deciding wherever the risk is real.
What's included
Customer service agents
An assistant that answers from your own documents and hands over to a person when it is unsure or the customer asks.
Lead qualification
An agent that asks the questions your sales team would, scores the answers and books or routes the lead.
Internal knowledge assistants
Answers across your manuals, policies and past tickets, with a link to the source of every answer.
Document processing
Invoices, orders and contracts read into structured data and checked against rules, with uncertain fields sent for review.
Custom AI tools and integrations
AI steps added to the systems you already use: your CRM, inbox, helpdesk or spreadsheets.
How I run it
Pick one job
The task, its volume, what a good answer is and what a wrong one would cost.
Data and rules
Which documents the AI may use, which rules it must follow and when it must hand over.
Prototype on real cases
Run it on past examples and compare with what your team actually did.
Integrate
Connect it to your systems, with every answer logged.
Watch and tune
Review samples, fix the rules and widen the scope only once it is reliable.
Content Monitor · Site Scanner · Polaris CMS
Run on my own tooling
My own systems are built for AI agents to work in. The same patterns carry over to yours.
Content pipeline
From a keyword round to a tracked post
Agents under house rules
My content pipeline drafts with AI agents under written rules, then runs automated checks for style and SEO and flags facts for review.
Site Scanner
A 98-item launch checklist, a crawler and Lighthouse
Findings as agent prompts
Scan findings become a ready prompt for an AI coding agent, so the agent starts from an exact list instead of a vague task.
Content Monitor
The dashboard every live site is run from
An API for your own agent
A personal API token lets your AI agent add sites and read the data, within the same limits as your login.
Recommendations as prompts
Chosen link suggestions and growth actions are exported as a ready prompt for an AI coding agent.
Polaris CMS
My own CMS, this site included
API keys for AI agents
Agents publish through the CMS API with their own keys, under the same 11-check SEO score as a person.
Proof
- Project PolishPal Polish audio made with an AI voice model forced to speak Polish, then checked by a native speaker.
- Blog post How I built PolishPal, and how I run it now How AI coding agents work on PolishPal from rules kept in the repository, and what had to be checked by people.
Where this stands Customer-service agents, knowledge assistants and document processing have no public case here yet; the approach above is how I would build them.
Questions
Will an AI agent make things up?
It can, so the design assumes it will try: answers come from your documents with a link to the source, unclear cases go to a person, and every answer is logged for review.
Where does our data go?
That is decided per project, before anything is built: which AI provider, which region, what is sent and what never is. I write it down so your team can check it.
Do we need a big project to start?
No. The best start is one narrow job with enough volume to matter, such as sorting incoming emails or reading one type of document.
Related services
All services →Tell me what you need
A few lines are enough: what you want to build, fix or automate, and any date that matters. I reply in English or Polish.