AI products forreal business problems.
I'm Thor Matthiasson, an AI product builder with a background in performance marketing and analytics. My work connects business workflows, practical AI products and questions about what works.
How I work
Start with the problem
Define who needs the workflow and what a useful result looks like.
Evaluate the result
Check the evidence
Inspect the output, question the assumptions and compare against a baseline.
Make it useful
Plan for adoption
Consider the people, support and measures needed to use a workflow in practice.
The approach
A clear problem.
A useful next step.
Connect AI capabilities to the work people do.
Define, test and learn.
Start with a workflow, agree what success means and use the result to decide what comes next.
AI Value + Risk Brief
10 business days
Align leadership on priorities, risks, and what to build first.
AI Workflow Pilot
Scoped together
Define a useful workflow, test its outputs and plan how your team will use it.
AI Product Advisory
Ongoing
Priorities, product decisions and adoption planning as the work develops.
How the work runs
From a business question to an informed decision.
Identify the problem, examine the workflow and test the result. Use what you learn to decide whether to continue, change direction or stop.
01
Align
What outcome — and who decides?
02
Assess
Are workflow, data, team, and controls ready?
03
Prioritize
What do we build first — and what do we not?
04
Design
How does the work itself change?
05
Pilot
What can we learn from a bounded trial?
06
Scale
How does this become repeatable?
Scope and timing depend on the workflow, available data and team. Agree engineering ownership and support before a pilot moves into production.
Before wider adoption, review integration, access, evaluation, monitoring and ownership.
Integrated
Connected to real tools, data, and the way people actually work.
Permissioned
Access follows role, data, and security constraints.
Evaluated
Outputs tested against known cases and explicit success criteria.
Monitored
Usage, failures, drift, cost, and exceptions are observable.
Owned
A named business owner and a named technical owner, on your side.
Speaking & advisory
From panels to executive tables.
Hybrid Threats panel, featured panelist
Reykjavik, Iceland
Premier Partner Event, opening keynote
Amsterdam, Netherlands
AI product showcase
New York, United States
Selected projects
Products and workflows to explore.
AI products / Publishing
Pagenta
An AI-generated page needs a stable link, access controls and a way to keep it useful after the conversation ends.
AI / Decision support
Plurality
A single polished answer can hide uncertainty. Comparing contributions makes assumptions and disagreements easier to inspect.
Finance / Research
Hybrid AI Finance
Equity research spans financial statements, valuation assumptions and competing interpretations. The useful question is what changed and why.
Meet the builder: Thor Matthiasson. Read the project case studies, product decisions and agent evaluation research.
Publication · AI Without the Hype
A closer look at AI claims.
AI Without the Hype is my AI publication, exploring developments in AI and the evidence behind them. It is another setting for questions about useful information, editorial judgment and AI-assisted workflows.
The questions that matter: what is the claim, what supports it, what remains uncertain and what does it change in practice?
Visit AI Without the HypeClaim
What is being asserted?
Evidence
What supports the claim?
Limits
What remains uncertain?
Use
What changes in practice?
Products, workflows and adoption
Let's talk
through the work.
Tell me about the problem, project or role you have in mind. We can explore the fit and what the work would involve.
A conversation about the problem and the next step.