Business understanding. Applied AI.

AI that survives contact with your business.

We connect business thinking with technical delivery. Turning open AI questions into clear decisions and good ideas into something that works every day.

Let’s talk it through How we connect the pieces

First questions, a stalled pilot or live operations: we start wherever you are today.

01 Budget & value “Where is
the ROI?”Recurring fees. Unclear returns.
02 Rules & control “Who is allowed
to do what?”AI in use. No shared rules.
03 Tools & pilots “Another week.
Another tool.”Lots of activity. Little progress.
04 Teams & ownership “Everyone does
their own thing.”Isolated wins. No common ground.
05 Data & shadow AI “Where does
our data go?”Quickly shared. Hard to trace.
06 Quality & context “That’s not how
our business works.”Good answer. Wrong context.
arctiq

We connect
the pieces.

Business × Technology

Our role

AI needs connection.
At every level.

Trained in business, hands-on in technology: we translate between leadership, business teams and IT. And work alongside you wherever the connection is missing.

Where are you
getting stuck?

An open business question, a team without clear guidelines or a solution stuck in pilot: that is our starting point.

We keep the whole picture in view and work on the concrete problem.

Let’s talk it through
ArctiqWe connect

Business & value

What should improve for your business?

Priorities, a business case and measurable value.

A decision that makes business sense.

People & ownership

Who is involved and who decides?

Collaboration, clear rules and knowledge in the team.

A solution people use every day.

Technology & operations

How will it work reliably in your systems?

Data, integration and AI we build alongside you.

Something that runs and stays with you.

Use cases

Six places where this work pays off.

Each one can start as a small, self contained piece inside your own tools. The first result makes it easier to decide whether a larger build is worth it. Every use case has a short demo you can open and try.

Process · inside and at the edge
Remove before you automate
Use case 01

Rethinking how the work flows

Before anything gets automated, we look at how one process really runs today, inside the company and at the edge towards clients and partners. Steps often disappear once you look at them closely.

The path it actually takes

We follow one request from the first email to the delivered result and write down every hand over. The version in the handbook and the version people use are rarely the same.

Steps that only move paper

Typing the same data into a second system, forwarding, chasing. These steps cost days and add nothing. They go first.

What has to stay

Approvals and checks stay in place. Good automation makes them arrive sooner and keeps them visible to everyone.

Then the tooling

Only what is left gets a tool built around it. That is usually a third less work than the original plan assumed.

Margin scenario · live
Shared assumptions, visible results
Use case 02

Decision models instead of another spreadsheet

Analyses that come back every quarter and spreadsheet logic that only one person understands become a tool of their own. Everyone can open it, move the assumptions and see what happens.

One version everyone uses

The logic lives in one place instead of in eleven files with the same name and different numbers.

Assumptions you can move

Price, volume, cost, churn. The model shows which of them the result really depends on, so the meeting can argue about that one.

Made for the room

It runs in the browser, opens in a second and needs no installation. That is what gets it used in front of a board.

The result is the clarity

The number was never the deliverable. Being able to show where it comes from is.

Mail · Calendar · CRM
Action with a checkpoint
Use case 03

Connecting the systems you already run

Value shows up where the work already happens: the mailbox, the calendar, the CRM, the ERP. An assistant reads what it needs, does the task and writes the result back, with a checkpoint in front of anything that matters.

Reading first

The first version only reads. That is enough to see whether the assistant understands the work before it is allowed to touch anything.

Permissions people can see

Every system it may reach is listed, with the reason it needs to be there. Access is a decision your team makes, not a setting hidden in a config file.

A checkpoint before it counts

Writing to a real system waits for a person. One click, with the full context on screen, and it goes through.

Everything leaves a trace

What was read, what was written, who approved it. Six months later that record is the only reason anyone can answer the question.

Curated corpus
Sources · units · container
DOCX XLSX PDF MAIL
containerv4
Use case 04

Knowledge that knows what is still true

Contracts, guidelines, offers and the mail thread where the exception was agreed. Scattered sources become answers with a source attached, and the system knows what replaced what and when.

One curated corpus

Documents get sorted, versioned and connected. The old version stays available and is clearly marked as replaced.

Every answer shows its source

You see the clause the answer came from, when it took effect and who owns it. That is what makes an answer usable in writing.

Conflicts stay visible

When two documents disagree, the system says so instead of picking the more convincing one.

Served where people work

In the tools your team already has open, so nobody has to remember a new place to look.

House voice · written down
Terminology · tone · templates
Use case 05

Output that sounds like your company

The model knows the language of the internet. It knows nothing about your terms, your tone or the way your best people write. Once that is written down, the difference is obvious in the first paragraph.

Your words, on paper

The terms your industry and your company use, with the ones you avoid. This is a short document, and it does most of the work.

Context from your systems

The customer history, the open ticket, the last visit. Facts only you have are what make a letter yours.

One quality bar

New colleagues write like the team on day one, and the team keeps the level on a busy Friday.

Route by sensitivity
One door, every model
eu api own server on device
Use case 06

Governance and model routing

One governed entry point decides which model handles which workload, based on how sensitive the data is, where it may be processed and what it costs. What ran where stays traceable.

One door for every model

Teams ask one place. Behind it sit the public model, the business API, your own server and the device.

Sensitivity decides the route

Public information takes the cheap route. Health data and personal data take the strict one. The rule is written once and then applies to everyone.

A log you can hand over

Every request leaves a line: what was processed, on which route, under which rule. That is the part an auditor asks for.

Shadow AI loses its reason

People use unapproved tools when the approved way is slower. Make the approved way the fast one and the problem shrinks on its own.

How I work

Not every engagement starts in the same place.

Four routes. Each answers a different question, and each ends with a clear next decision instead of another workshop.

Shape

01

Where should AI change the business?

How we work together
  • Leadership interviews
  • Exposure mapping
  • Business model work
  • Roadmap workshop

Prove

02

Will it work here?

How we work together
  • One bounded question
  • Success defined upfront
  • Built on real data
  • Fund, revise or stop

Scale

03

How does it become dependable?

How we work together
  • Integration
  • Access and controls
  • Exceptions and monitoring
  • Handover

Advise

04

Which decision needs a second view?

How we work together
  • Retained context
  • On call for decisions and inputs
  • Small follow on builds

I sell no platform and no model licence. When the honest answer is that you can wait, that is the answer you get.

Moritz Blattner
Zürich, Switzerland · German and English

About

Moritz Blattner

Forward Deployed AI Architect

I build AI systems where they will run: inside existing enterprise environments, with real data, real permissions and the legacy systems that came with them.

Before Arctiq I spent five years at BearingPoint and led agentic engineering internationally. I watched good demos fail because ownership, access and operations were left open. The model was rarely the reason.

That shapes how I work today: decide with leadership, build alongside the team, and leave the system, the documentation and the know how behind.

I work from Zürich in German and English. If something at your company is half built or half decided, that is usually a good time to talk.

FAQ

The questions that come up before we start.

Arctiq is a small consultancy. That shapes the answers below, so they are worth reading before the first call.

What does forward deployed mean in practice?

I work inside your environment instead of writing about it from the outside. That means access to a real system, a seat in the team that owns the process, and code that runs in your setup. Most of the work happens where the data already is.

Who actually does the work?

I do. Arctiq is one person with a network, so you always talk to the person who builds the thing. For larger builds I bring in people I have worked with before and I say clearly who does what and what it costs.

How does an engagement start?

With a call of about thirty minutes, free of charge. If it looks like a fit, the next step is usually a short assessment of one process with a fixed price. At the end of it you have a written recommendation and enough clarity to decide whether to build.

How long until we see a first result?

A working piece you can click on usually takes two to four weeks. Daily use in production takes longer, because access, exceptions and operations take the time they take. I say upfront which of the two we are aiming at.

How do you charge?

Small pieces of work have a fixed price. Longer builds run on a day rate with an agreed budget and a monthly summary. Advisory work runs as a small retainer. There are no licence fees and no platform on top.

Do we need an AI platform or a strategy first?

No. One finished use case teaches you more about your own readiness than a strategy paper does. The strategy gets much easier to write afterwards, because you know what the work really takes.

What happens to our data?

Your data stays in your environment. Where a model is involved, we agree beforehand which route it takes: a business API with a contract, your own server, or a model on the device. Everything gets logged, and that log stays with you.

What do we own at the end?

The code in your repository, the documentation, the tests and a short handover with the people who will run it. You can keep working with me afterwards. You should not have to.

Do you work with our IT and our existing partners?

Yes, that is the normal case. Your IT sets the rules for access and operations, and I work inside them. If an existing partner already runs part of the landscape, I take their constraints as given and agree the interfaces with them.

What if the honest answer is that we should wait?

Then you get that answer, with the reason and with what would have to change for it to become worthwhile. A project that starts on a weak case costs more than a delayed start.

Contact

Tell me what you are trying to build.

Send me a few lines about the process and where it gets stuck. I answer within a couple of days and the first thirty minute call is free.

Most useful: who owns the process, which systems are involved and what a good result would look like. Please keep credentials and customer data out of the first message.

Direct moritz@arctiq.ch LinkedIn