Skip to content
Amsby

AI solutions

AI that does real work for your team

We build AI agents and automation that take over repetitive research, writing and analysis work. Built on your own data, delivered in the tools your team already uses, with a human approving anything important.

Not sure where to start?Request a free AI audit

  • AI agents live with clients
  • Human always in the loop
  • Dutch BV & contract

What we build

From a single task to a complete AI workflow

We start with the work your people now do by hand. Where that work is predictable but time-consuming, we let an agent do the groundwork.

  • Research and sales agents

    Research accounts, find and verify decision-makers, and write a personalised first message for each account.

    See case: Control of Work software company

  • Content agents in your house style

    From keyword to publication-ready article: research, plan, copy and an automated quality check, in your own tone.

    See case: Online marketing agency

  • Knowledge bases and RAG

    Your templates, case studies and documents made searchable for AI, so every answer is grounded in your own material and stays current.

    See case: No-Hassle DigitalSee case: Online marketing agency

  • Audits and data pipelines

    Collect data from several sources, normalise and score it, and turn it into a prioritised action list.

    See case: DataClicks

  • Workflow automation

    Connections between forms, sheets, documents, CRM and APIs, so work flows through without copying and pasting.

    See case: Infinity Teams

  • AI in your own product

    AI features built into your existing application, by the same developers who already work in your codebase.

How an agent works

Not a chatbot, but a process with checkpoints

A good agent follows the same playbook as your best employee, and checks its own work before anyone sees it.

  1. 1

    Trigger

    A form, a new row in a sheet, a file in Drive or a set schedule.

  2. 2

    Collect

    Live search results, APIs and your own knowledge base, often in parallel.

  3. 3

    Reason

    A language model analyses, picks the right approach and turns everything into structured data.

  4. 4

    Check

    Fixed rules and a second AI reviewer score the result. Too weak? It gets rewritten.

  5. 5

    Deliver

    In Google Sheets, Docs or your own system, with a log of every run.

  6. 6

    Approve

    Your specialist reviews and sends. The agent does the groundwork, a person decides.

Our principles

AI you can build on

  • Human in the loop

    Anything that goes out is approved by a person first.

  • Your data, your tools

    We build on your own material and deliver where your team already works.

  • Built-in quality control

    Measurable rules, scores and automatic revision, instead of hoping for the best.

  • No silent failures

    Errors are caught and logged, so you always know what happened.

  • The right model for the job

    We choose models and tools based on the task, not the hype.

  • Tested on ourselves

    We run many approaches in our own business every day before we use them with clients.

How we work

From idea to working agent

  1. Map the process

    Together we map the current work: inputs, decisions, exceptions and what “good” looks like.

  2. Working pilot

    We first build one working flow on real data, so you quickly see what it delivers.

  3. Measure and refine

    With your specialists we tune prompts, rules and quality checks until the output is right.

  4. Run and extend

    We run and extend it, or hand it over to your team, with or without a dedicated Amsby developer.

Technology

What we build with

Proven building blocks we use in production. Already using other tools or models? We fit in with those.

  • n8n
  • Google Gemini
  • Gemini Embeddings
  • RAG
  • Pinecone
  • Serper (Google Search API)
  • Google Sheets API
  • Google Docs API
  • Google Drive API
  • Search Console & PageSpeed API
  • JavaScript
  • Prompt engineering

FAQ

Good to know

Do we need to get our data in order first?

No. Most agents start with what is already there: a spreadsheet, a folder of documents or a few API connections. Mapping the process shows which data is really needed.

Which AI models do you use?

In our current cases mainly Google Gemini, with n8n for the workflow and Pinecone or another vector store for knowledge bases. We choose what fits each project best, including when you already use a particular model or platform.

Will AI take over our team’s work?

It takes over the groundwork: searching, collecting and writing first drafts. Your people move from producing to reviewing and approving, and keep the decisions in their own hands.

How do we work together?

Usually as a project: we deliver all or part of the agent. If you want to keep building afterwards, you can do so with a dedicated developer or team, through the same Dutch BV and a Dutch contract.

What work would you like to hand over?

Tell us which process takes up too much time today. We'll help you work out what an agent can take over and what should stay with your team.