Sorting email and tickets with AI: when a small model beats GPT
How to design categories an AI can apply, when a small fine-tuned model beats a frontier model, how to route what it is unsure about, and what sorting a thousand emails costs.
Read →BLOG
Notes from production.
What I learn shipping agents, platforms and campaigns, written down while it is still true.
Voice AI agents for Greek businesses: latency, recognition and the law
What an AI phone agent can do for a Greek business, why a reply after 700 ms feels broken, the recognition gap for Greek, what a call costs and what the law requires.
Read →What a large language model (LLM) is, and what it cannot do
How large language models work in plain terms, why Greek text costs about twice the tokens of English, what LLMs do well, and the five things they cannot do reliably.
Read →myDATA and e-invoicing in Greece: what AI can automate in the books
B2B e-invoicing became mandatory for every Greek business in October 2026. What that means for myDATA, where mismatches come from, and what an AI reconciliation agent can do safely.
Read →From scanned invoice to validated record: AI document processing that holds up
How AI reads invoices and forms, why validation matters more than the model, which confidence signals work, what a page costs, and what e-invoicing changes in Greece.
Read →RAG explained: how to make AI answer from your own documents
How retrieval-augmented generation works step by step, when RAG is the wrong tool, what breaks it most often, why Greek needs different chunk sizes, and how to test that it works.
Read →AI use cases for small businesses: where AI pays back first
Where AI pays back first in a small business, department by department: sales, support, accounting, operations, HR and legal, with what stays human and how to pick a first project.
Read →What an AI agent costs to run: tokens, caching and failed runs
What goes into an AI agent's running cost, what one task cost on a real benchmark (about one cent), how much prompt caching saves, and how to estimate your own agent's cost before building it.
Read →Jev and structured outputs: when AI should choose, not write
Most business AI tasks are decisions, not text. How structured outputs and decision models like Jev work, what the first independent tests found, and how to design a decision AI can take safely.
Read →AI customer support for an e-shop: what a chatbot should and should not handle
What an AI chatbot can answer for an online shop, why the connection to orders matters more than the bot, what Klarna and Air Canada teach, and what the law requires.
Read →Choosing an AI agency or AI partner: the questions that expose a weak vendor
Ten questions to ask any AI agency before you sign: which model and why, how they test, what the agent may do alone, what Greek costs, and who owns the result.
Read →Agentic AI explained: what it is and what it changes for a business
Agentic AI chooses its own next step: it calls tools, checks the result and retries. How AI agents work, how they differ from chatbots and automation, and when one is worth building.
Read →When NOT to build an AI agent
I build agents for a living, which is exactly why I refuse half the agent projects that reach me. The four cases where an agent is the wrong answer, and what to do instead.
Read →Jev use cases: where a decision model fits in a business, and where it fails
A six-point fit test for decision models like Jev, the business tasks where they work, the ones to keep in code, the patterns that make them safe, and what the terms forbid.
Read →Jev vs LLMs: what the first evaluations show on accuracy, cost and speed
How Jev compares with large language models on accuracy, cost and speed: the vendor's workflow test, two independent papers, what survives a real system, and when to use both.
Read →What national prize-draw campaigns teach you about traffic spikes
A TV spot airs and your quiet site gets a month of traffic in twenty minutes. Campaign tech for brands like OPAP and Coca-Cola taught me lessons every e-shop should steal.
at digitall as Web Developer Read →GDPR without the drama: designing compliance in instead of bolting it on
Compliance bolted on after the letter arrives is expensive and ugly. Designed in from day one, it is mostly free, and it makes the product better. A builder's view, not a lawyer's.
Read →Lighthouse 96-100: what moves site speed, and what is a myth
The builds I still control hold 96-100. Getting there is not one trick. It is weight budgets, image discipline and refusing tag soup. Here is what moved the needle and what did not.
Read →Why your beautiful site earns nothing, and the number that fixes it
The industry's speciality is sites that win compliments and lose money. The fix is deciding, before any design, which single number the build must move.
Read →The three numbers an owner needs at 8am
Most dashboards are forty charts nobody reads. The ones that survive show three numbers the owner checks with coffee. How to find yours, and why fewer is a feature.
Read →Your spreadsheet is an ERP that is failing quietly
Every growing business runs on a heroic spreadsheet, until the day two people edit it at once. How to tell when it is time for a real system, and what custom actually buys you.
Read →AI automation: when a workflow is enough and when you need an AI agent
When a script or a workflow, in n8n for example, is enough, when a workflow needs one AI step, when only an AI agent survives, what each costs, and the test that decides.
Read →What an AI agent does at 3am, and why demos never show it
Every agent demo shows the happy path. Production starts where the demo ends: timeouts, malformed data, and nobody watching. Here is what surviving that actually takes.
Read →