Skip to main content

AI agents for routine tasks

AI agents are language models given access to tools so they can solve a task in several steps. I build them bounded, with a log of every step and a human involved where it matters. The experience comes from 400+ projects since 2010.

  • I usually reply the same day
  • Written estimate before work starts
Mikkel Tschentscher

star reviews

Read every review
  • Carries the project all the way through

    When you work with Mikkel, you can be certain the project gets carried 100% of the way.

    Simon Bille Hjort MartinsenColleague at Bigum&Co(translated from Danish)
  • The most solution-oriented developer

    In Mikkel's world there are challenges rather than problems.

    Michael Roscoe PedersenEntrepreneur(translated from Danish)
  • Competent, precise and a huge support

    He is simply a pro. Highly competent and precise, brings good input, and is a huge support.

    Carrsten Johan Thessen(translated from Danish)

AI agents

What agents are good at

Agents work best on tasks where the steps are known but the order depends on what turns up along the way. A person would check three places in whatever order makes sense given what appears, and an agent can imitate that.

They work badly on tasks with no right answer, or where a mistake only surfaces much later. Judgement calls, prioritising between customers and anything requiring discretion should stay with people, and I say so from the start.

Typical tasks for an agent

  • Customer service: investigating a case across the order system, carrier and email history, and preparing a draft reply for approval.
  • Document handling: filing, naming and routing incoming files based on their content.
  • Internal tools: maintaining data by finding duplicates, filling gaps and flagging what looks wrong.
  • Integration with existing systems: working through APIs in your CRM, accounting system or webshop.

Approval and control

The most useful setup is rarely the most autonomous one. Often it is the one where the agent finishes the work and queues it for approval, so a person spends two minutes instead of forty.

That model is also easier to start with, because nobody has to let anything loose on day one. Trust grows as the log shows what the agent does and why.

Examples of AI agents in a business

AI agents can be used well beyond customer service. The examples below are grouped by department, and what they have in common is that the agent gathers and prepares while a person makes the decision. Use them as inspiration when you look for your own first task.

  • Sales: preparing a customer meeting by pulling order history, open cases and recent correspondence from the CRM and email, and writing a short overview.
  • Finance: reconciling incoming payments against open invoices, finding the ones that do not match and preparing a reminder for approval.
  • Purchasing: comparing order confirmations with purchase orders and flagging discrepancies in price, quantity or delivery date.
  • Logistics: keeping an eye on shipments that have stalled with the carrier and preparing a message to the customer before they ask.
  • Ecommerce: finding products with missing descriptions, images or specifications and proposing corrections based on the supplier's datasheets.
  • IT support: doing the first investigation of a support ticket by checking the user's account, known issues and earlier tickets before a technician takes over.
  • HR: making sure a new employee's accounts and access are ordered and the introduction meetings booked, and following up on whatever is missing.
  • Marketing: gathering numbers from ad platforms, web analytics and the newsletter for the weekly report and drafting the commentary.
  • Contracts: reviewing incoming contracts against your standard terms and flagging the points a lawyer should look at.

What is an AI agent?

An AI agent is a language model given access to some tools and permission to use them several times in a row, until it judges the task complete. That is the whole difference from an ordinary chatbot, and there is no hidden intelligence underneath. What it does open up is tasks where you cannot write the order of steps down in advance.

Agents are relevant for companies where staff regularly investigate cases across several systems, gather information from scattered sources or maintain data that has been keyed in over many years. If the answer to a question is always spread across the order system, the inbox and a spreadsheet, there is an obvious agent task.

Many people can explain what agents are, but the hard decisions live in the actual building: which tools the agent may use, what requires approval, and how you test something that can take different routes to the same goal. I have built these in practice and have been a developer since 2010, so those decisions rest on experience rather than theory.

My approach is to give each agent one responsibility. An agent meant to do everything becomes impossible to test, because the number of possible paths explodes, and impossible to debug, because you cannot see which step went wrong. Several small agents are duller to talk about and far easier to live with in production.

How an AI agent works, step by step

An AI agent works from a goal, for example finding out why a customer's order has not arrived and preparing a reply. It has a list of tools it may use: look up the order, query the carrier's system and read the email thread. Each tool comes with a description of what it does, and the agent chooses for itself which one to use first.

Then it runs in a loop. It uses a tool, reads the result and decides the next step based on what it found. If the order system shows the parcel has shipped, it asks the carrier. If the carrier shows a delay, it reads the email thread to see what the customer has already been told. Along the way it remembers what it has seen in the current task, and that memory is what stops it asking the same thing twice.

The loop stops when the agent judges the goal reached, or when it hits a limit on the number of steps. The result is typically a draft or a proposed action queued for approval, and every step is in the log with its reasoning. Memory across tasks is something you build deliberately, for example as notes in your own system. It is not something the model has on its own.

AI agents for companies across Denmark

From my base in Glostrup near Copenhagen I build AI agents for companies across the country. Clients sit in Aalborg, Aarhus and Odense, a few abroad, and most of the collaboration happens over video and in writing.

An agent project needs access to your systems, not to your office. So it works fine at a distance: we clarify the task in a video call, I get API access, and you follow a log of what the agent does while it is being tested. If you are in the Copenhagen area, I am happy to meet in person along the way.

The projects I build most often are agents for case handling in customer service, for routing incoming documents, and for cleaning up product or customer data across systems.

The advantage of a freelancer over an agency is that you talk directly to the person building. Decisions about scope and approval flows require many small clarifications, and those move fast when there is no project manager in between.

How trust in an agent gets built

Nobody should let an agent loose on their order system on day one. The sequence that works is the same every time: first the agent may only read, and the result is shown to a person. Then it may propose an action that gets approved with one click. Finally it may carry out the actions where a mistake is cheap to fix.

Along the way the log grows into something you can make decisions from. After a couple of weeks you can see how often the agent was right, where it was unsure, and which case types it consistently fails on. Those case types get routed to a person instead of trying to make the agent good at everything.

In certain places the approval step should never disappear. Anything sent out of the building to a customer, and anything that moves money, belongs there. That is a decision about who answers for the outcome, not a technical limitation.

What you get

Concrete deliverables, not a report of recommendations someone else has to find time for.

  • One bounded agent

    An agent with one clear remit, a step limit and a way to stop it in the middle of a task.

  • A tool layer

    Access to your systems, split between reading and writing, where write access only arrives once the read side has proven itself.

  • A full log

    Every step, every tool call and the reasoning behind it, so you can debug and always answer what happened.

  • An approval flow

    Where the consequences are large, the agent queues its work for approval instead of carrying it out itself.

How it works

Four steps. You know what happens when, and you can stop after any of them.

  1. 01 / 04

    Pick the process

    Together we find the piece of work that repeats often enough for automation to pay off. The more contained the task, the faster the first version reaches production.

  2. 02 / 04

    Set a baseline

    We record how long the task takes today, how often it occurs, and how many errors slip through. Without those numbers, nobody can tell afterwards whether the solution was worth the effort.

  3. 03 / 04

    Build against a test set

    I collect real examples from your day-to-day and test every prompt and model change against them. That keeps quality stable when the solution is tuned later on.

  4. 04 / 04

    Production with logs and a kill switch

    The solution is wired into your existing systems with full logging and a switch that can turn it off. I document the setup and train someone on your side to maintain it.

Brands I've worked with

  • MT Højgaard Danmark
  • Egmont
  • NNIT
  • Visma
  • Lomax
  • Pascal
  • able.
  • OOONO
  • Novo Nordisk Fonden
  • Energii
  • Maersk Tankers
  • Nordkysten Entreprenørfirmaet

Frequently asked questions

What is an AI agent, briefly?

A language model with access to tools, allowed to use them across several steps until the task is done. It can look things up in systems, gather information and propose or carry out actions. The difference from a chatbot is that it does work instead of only answering.

What is the difference between an AI agent and an AI assistant?

An AI assistant answers when you ask it, typically in a chat, and you decide what happens with the answer. An AI agent carries out a task on its own across several steps and systems, within the limits it has been given. Both are built on the same kind of artificial intelligence. The difference is what it has access to and is allowed to do.

How does an AI agent work in practice?

The agent is given a goal and a set of tools, such as access to look things up in the order system or read email. It uses a tool, reads the result and chooses the next step, until the task is solved or it hits a limit on the number of steps. The result is typically a draft or a proposed action that a person approves, and every step is logged.

What can an AI agent be used for?

An AI agent can be used for tasks where someone today gathers information from several systems before a decision can be made. Typical examples are investigating a customer case, reconciling payments against invoices, following up on delayed deliveries or cleaning up product data. The agent handles the gathering and presents the result, which frees up time because the person only has to decide on it.

Can we use Copilot agents from Microsoft?

If you already have Microsoft 365, Copilot agents can be a good place to start. They suit standard flows inside Microsoft's own apps such as Teams, Outlook and SharePoint. If the agent needs to work across your own systems, such as your order system, accounting system or webshop, that usually calls for an agent built for the task. The two can easily be used side by side.

How do you build an AI agent?

You start with the task, not the technology: what should the agent do, which systems does it need access to, and when should a person approve? Then a language model is connected to the tools it may use, such as your CRM or email system, with clear limits on what it may change. The most important step is testing on real examples before it is allowed to work on its own.

Which task should our first AI agent take on?

Start with one case type that comes in many times a week and requires lookups in two or three systems. At first, let the agent only read and present a draft, and measure how much time it saves. Once the log shows it gets things right, the next step is letting it propose actions that are approved with one click.

Which AI agents are best?

There is no single best one. The large language models from OpenAI, Anthropic and Google are all good enough for most tasks, and the difference lies more in how the agent is built: which tools it has, which limits it works within, and how well it has been tested. An off-the-shelf agent from a platform can be fine for a standard flow, while tasks in your own systems usually need one built for them.

What does an AI agent cost?

The price depends on what the agent needs to do. What matters most is how many systems it works in, whether those systems have an API, and how much approval flow needs to be built around it. Send a description of the task and I will come back the same day with clarifying questions and then a fixed quote.

What if the agent does something wrong?

Then the mistake has to be detectable and reversible. That is exactly why I start with read access and add write access afterwards, one action at a time, with logging on everything.

Which systems can an agent work in?

Anything with an API: Shopify, Google Workspace, HubSpot, e-conomic, Jira or your own system. If the API is missing, a layer can often be built in between, but that belongs in the estimate because it is frequently the largest part of the job.

What about GDPR when an agent has access to customer data?

The agent only gets access to the fields it needs, and everything it does is logged. The log in itself helps with documentation requirements. Personal data can often be masked before it reaches the model, and I walk through the setup with you before anything goes live.

Do you build agents for companies outside the Copenhagen area?

Yes, most of the work happens remotely anyway. I have clients in Aarhus, Aalborg and Odense whom I mostly meet over video. If you are near Copenhagen, I am glad to come out for the kickoff meeting.

How long does it take to build an agent?

A bounded agent on one system is typically two to four weeks including testing. Working across several systems, it is the integrations rather than the AI part that set the timeline.

Can the agent work outside office hours?

Yes, and that is often the point. Cases that arrive in the evening can be investigated and prepared by the time someone starts in the morning. The decision itself can still sit with a person.

Your next step

Get a quote for AI agents

Tell me about your project. I’ll get back to you with questions, options and a concrete next step.