Skip to main content

AI workflows your team can run

I help product and engineering teams put AI into their existing software and daily work. We start with one useful workflow, build a pilot and test it against your current process. Your team gets the working system and the documentation to run it.

Work directly with me, with the scope agreed before I start and the code in your repository.

Discuss an AI pilot

Two places to start

AI in your product

You have a product and a task users need help with, such as finding answers in documents or classifying incoming information. I build the feature and the controls around its model calls, including cost limits and evaluation.

AI in your engineering workflow

Your engineers are trying coding agents and need a shared way to use them. We put project knowledge and tool access around one workflow, then test how a real task moves through implementation and review with your team.

One workflow, checked before expanding

The first engagement has a written scope and a result your team can inspect. Here is what the pilot includes.

  1. Choose the workflow

    Look at how the task runs today, where time goes and what a useful result looks like. Agree on representative examples and the criteria the pilot must meet.

    A workflow map and written scope

  2. Connect the context and tools

    Give the workflow the documents and system access it needs. Define which actions need approval, limit access and spending, and make failures visible.

    A working pilot in your environment

  3. Check it against real work

    Test the output on your examples, including the awkward cases. Compare quality and cost with your current process, counting the time people spend reviewing and correcting it.

    A measured result and a recommendation

  4. Hand it over to the team

    Walk the team through using and maintaining the workflow. Agree who owns it and document the limits, failure handling and checks to run when something changes.

    Team training and a documented handover

The work behind the offer

Client AI engineering and systems I built for my own work, labelled by how I was involved. Each example links to the engineering decisions.

  • Classification and alerts

    MarketAlerts · client work · company closed

    I owned the AI classification layer and alert engine, turning financial filings and news into events users could follow. A cheap model handled the first pass, with harder decisions sent to a stronger model.

    Read the engineering example: Classification and alerts

  • Usage costs and memory

    AI tutoring platform · client work

    I replaced message-based limits with cost-based usage metering and built background memory for an AI tutor. The client is under NDA; the public example describes the engineering.

    Read the engineering example: Usage costs and memory

  • Retrieval over financial news

    Metis · my own platform

    My financial-news platform combines semantic search with agent tools. It gives me a working example of connecting models to a changing body of source material.

    Read the engineering example: Retrieval over financial news

  • Building with coding agents

    My own workflow

    The workflow I use on client work and my own products keeps project context close to the code, verifies changes and asks a separate agent to review them before my own review. The reusable skills are public.

    Read the engineering example: Building with coding agents

Before we start

Do we need custom software?

That depends on the workflow. The first step is to check what your existing tools can do. Configure them where that is enough, and build a connection or product feature where the task needs it.

How is the engagement scoped?

One team and one workflow first. We agree on acceptance criteria and operating costs, including the system access the pilot needs. Scope and price are set in writing before work starts, with a schedule agreed around your team. The pilot gives us evidence for deciding what comes next.

Can you help our engineers use coding agents?

Yes. We can adapt the context files and review process I use in my own work to your repository, then test them on a small change with your engineers. The walkthrough and documentation are part of the scoped engagement.

Who does the work and owns the result?

You work directly with me, Borut Balazek, through Corco Labs. You keep the code and repository, with documentation so your team can continue the work. Ongoing support can be agreed separately.

Bring one workflow you want to improve

Email me the task and the tools your team uses. I'll come back with an honest fit check and a proposed first step.

Discuss an AI pilot Book a Call