Now booking Q1 2027
Now booking Q1 2027Start a project

17:00 — this craft's hour on the studio wall

AI automation and LLM agents for startups and product teams

We automate the repetitive work around your team with language models and plain workflow tools: reading and routing email, updating your CRM, drafting replies, answering questions from your own documents. Where a mistake would be expensive, a person approves the step.

Start an automation projectEstimate the cost

What we help with

  • Hours lost to copying, pasting and chasing

    Data moved by hand between email, spreadsheets and your CRM. We map the steps, automate the predictable ones and leave a clear queue for the rest.

  • Your knowledge is buried in documents

    An assistant that answers questions from your policies, product docs or past tickets, built with retrieval so every answer shows where it came from.

  • An AI pilot nobody trusts

    We add evaluation sets, logging and guardrails, so you can see how often it is right, catch where it goes wrong and improve it on evidence.

  • Support and sales replies that eat the day

    Drafts and summaries that a person reviews and sends, so your team handles more without handing its voice to a machine.

What you get

  • A map of the workflow as it runs today, and which steps are worth automating
  • Agents and workflows connected to your tools, such as Gmail or Outlook, HubSpot, Salesforce, Slack and Notion
  • Approval steps wherever an action costs money or reaches a customer
  • An evaluation set, and a dashboard showing how the automation performs
  • A log of every action, so any result can be traced and explained
  • Prompts, code and runbooks handed over to your team

What we build with

The simplest tool that does the job, and a model you can swap later.

LLM APIs
OpenAI, Anthropic, Google or open-weight models, chosen per task on quality, cost and where your data may go.
Retrieval (RAG)
your documents indexed in a vector database, so answers come from your sources rather than the model's memory.
n8n
visual workflows your team can read and adjust, self-hosted if you need it.
Python
for the parts that need real code, tests and version control.

How an engagement runs

Most projects move through these in order. You can stop after any of them and keep what was made.

  1. Scope sprint

    2 weeks · fixed fee

    Turn an idea into a plan you can trust before you commit to a build.

    We map the workflow, prototype the riskiest step on your real data and quote the build.

    • A clickable prototype
    • A list of risks and how to retire them
    • A fixed quote for the build
    • The fee is credited against the build

    Sample price: from $3,000

    Start with a scope sprint

  2. Build

    6–12 weeks

    We design and ship the product, with something real to try every Friday.

    Agents and integrations shipped one workflow at a time, each with its own evaluation.

    • Weekly builds you can open on your phone
    • Design, engineering and launch
    • Code, models and data handed over
    • 30 days of fixes after launch

    Sample price: typically $15,000–60,000

    Plan a build

  3. Ongoing team or ML pilot

    monthly

    A small senior team that keeps improving a product you already run.

    Tuning prompts, adding workflows and keeping up as the models change.

    • A fixed team, no rotating juniors
    • A monthly plan and a weekly demo
    • Pause or stop with a month's notice

    Sample price: from $4,000 a month

    Talk about a team

Prices here are samples, to give a sense of scale. A scope sprint ends with a fixed quote for your project. For a rough range before you write, try the cost estimator.

Common questions

Is it safe to let AI act on our systems?

Only within limits you set. Each automation gets the narrowest access it needs, every action is logged, and a person stays in the loop for anything that moves money, messages a customer or changes a record you can't easily undo.

Which AI model do you use?

Whichever fits the task. We compare models on your own examples for quality, speed and cost, and build so the model can be swapped later without rewriting the workflow.

What happens when the model gets something wrong?

It will sometimes, so the system is designed for it. Low-confidence results go to a person, mistakes are logged, and the evaluation set grows from real errors so the next version does better.

Can it work with the tools we already use?

Usually, yes. Most business tools have an API or an n8n integration. List your tools in the brief and we will flag anything that would need a workaround.