AI & agent consulting for marketing teams
We build marketing operating systems: your knowledge, your data and working AI agents in one place, run by your own people. We answer where to start, what to automate and what to build, then we build it with you.
Two ways in. Pick the one that sounds like you.
“We want to do this, but we don’t know where to start.”
2 weeks · starting from €4,000
An audit of your marketing workflows and data, a use-case map, and a build order with owners. You leave knowing your first three moves and what they cost. If you are not ready for the bigger build, we say so and tell you what to fix first.
“We know roughly what we need. Nobody here has the time to own it.”
monthly · scoped in the intro call
One person from our team, in your week, whose job is your AI progress: building, training and sparring in the same person. Keep them as long as it is useful, have them train your own people into the role, or have us help you hire your own.
Where you are
Manual marketing gave way to algorithmic, algorithmic to AI-assisted. The fourth era is agentic: systems that hold your context and do real work. Most Finnish marketing teams are still somewhere between eras two and three, and that is a normal place to be starting from.
Who does the work
Chief AI Officer
CMO of Barona, one of Finland’s largest employers, until August 2026. Built the marketing operating system his own 20-person team ran every day, and showed the build publicly while it was happening. Leads strategy, transformation and adoption.
Chief Technology Officer
Ten years building marketing data infrastructure: GA4, BigQuery, Snowflake, pipelines, BI. Built Snowflake-native AI agents in enterprise environments and develops AI-native daily. Leads architecture, data and engineering.
Things you can look at today
Built by Joni and run daily by a 20-person group marketing organisation at one of Finland’s largest employers, documented publicly as it happened. Future builds are better versions of this.
The same structure stripped back to a starting point, on GitHub, with no strings. Read it, take it, and get going without us if that is what you need.
github.com/jhelm-mktg/marketing-os-starter-kitEleven people in January 2026, over twenty by August, the whole company through an 8-week AI sprint. Behind the two of us: senior specialists in search, social, data and creative, working agent-native daily.
The work
Both doors lead into the same builds. Take one piece, or take several in the order that suits your year.
Campaign variants, ad copy, product and category content, localisation across markets. Produced at volume in your brand voice, under your team’s review rather than handed to a content shop.
Routine reporting handed to agents that surface what matters. Your analysts move from collecting to deciding.
Five working sessions on your real cases, homework between them, feedback per person. Capability that stays after the trainer leaves.
A working agent environment on your knowledge and data: context layer, first skills live, team onboarded.
We pick up from wherever you are: new builds, deeper training, or a second opinion when you want one. Capability reviewed each quarter. The OS stays yours if you stop.
One job, built properly and built to last: campaign QA, competitor monitoring, brief generation, lead routing, whatever is eating your week. Bought on its own, with an honest read on what it will and will not compound into.
Multi-team organisation? Transformation programmes are scoped separately.
What this looks like in practice
If nobody has told you what an AI transformation actually means inside a company like yours, this is the shape it usually takes. Six layers, each with a job. In most marketing organisations the right first move is to build these foundations rather than to buy more tools, because each layer is what makes the one above it work.
Your brand, offering, audiences and tone, structured so agents can use them. Plain files, versioned, yours.
Your analytics, warehouse and search data, connected and clean. This layer is where output quality is decided.
Agents and skills producing real work: briefs, campaigns, content, analyses, reports.
Review gates and human sign-off wherever judgment matters.
EU data zones and model routing by sensitivity, agreed with IT and legal, enforced in the build.
Every use feeds back, so the system and the team both get sharper monthly.
Where this can go
Optional, and it comes later. Once your system is running and we also run channels for you, the two can talk to each other. You set what crosses, and you can close it at any time.
Yours. Runs on its own.
You set what crosses
Ours. Grows with every build.
Optimisation stops waiting
Campaign performance and search-intent data in one conversation, so account changes get made sooner.
Reporting comes from your side
Your team asks your own system. It fetches the numbers and the reasoning from ours.
You inherit every other build
A skill written for someone else’s problem plugs into your context. Capability crosses the bridge, data does not.
None of this is a condition. Your system runs on its own, and plenty of the companies we work with will never want anything more.
How to choose
Ours included. Take them into every vendor conversation you have this year. Each one takes a minute to ask, and the answers tell you most of what you need to know before you sign anything.
The brand, offering and audience knowledge the agents actually run on. Ask where those files live and whose account they sit in.
Ask to see the workflow itself. If it only runs inside the vendor’s environment, ask what happens to it on the day you leave.
Ask them to list it: the system, the files, the skills, the people who know how it works. Then ask what it costs to keep.
Which region, which models, and who signed off. Get IT and legal into the room at the sketch stage rather than at the contract stage.
Ask for a name and the plan that gets them there. If nobody can be named yet, that is the first thing to solve, and we will say so too.
FAQ
Thirty minutes. We show you a working marketing OS rather than slides about one, and give you an honest read on where you are. If the answer is that something else should come first, we say so and tell you what.