AI that actually
gets used.
I'm Sture Skalstad, based in Oslo, Norway, and a recent graduate of NLA University College (2026). Over the past year I have worked on exactly that: I built a measurement instrument for AI visibility from scratch, with no coding background, got a client to first place in Perplexity one day after launch, and automated the document workflow of a company. I'm now looking for an advisory role in AI and digitalisation.
In teams I often end up taking the lead. I led the work and the supervision meetings on my thesis. In Nairobi I was given the lecturer's own case and designed the track myself. On the master's course HSM302 I proposed a new workflow and was given responsibility for the pedagogical development of all the lectures.
- Name
- Sture Skalstad
- Location
- Lives in Oslo, Norway
- Education
- Bachelor's degree in innovation, marketing and management from NLA University College, 2026Exchange semester at Griffith University, Australia, autumn 2025
- Thesis
- On AI adoption in small businesses, grade A
- Projects
- Built his own measurement instrument for AI visibility (GEO)Developed the video lectures for the new master's course HSM302 at NLA University CollegeAutomated quotes and document workflow for a travel company
- Work
- Student assistant at NLA University College since 2024On-call care worker, Midt-Telemark municipality
- Looking for
- An advisory role in AI and digitalisation in OsloAvailable at short notice
- Contact
- sture@skalstad.no
GEO: measuring AI visibility for businesses
Own project, June 2026 to date
GEO (Generative Engine Optimization) is about being found and recommended by AI assistants, in the way SEO is about search engines. I built a measurement instrument that shows how ChatGPT, Claude and Perplexity recommend businesses, and points out concrete measures a business can take to become more visible in AI answers.
The instrument was built from scratch in under three months, without a coding background. The code was written through AI-assisted development: I wrote the requirements for what the system should do, made the key decisions along the way, prioritised, and quality-checked every delivery through around 1,200 automated tests. In practice, product owner work. I have measured three Norwegian industries, and the instrument can be demonstrated live in a meeting.
For the pilot, I built a completely new website for the client from scratch, written so AI assistants can read and cite it. It launched on 1 October 2026.
One day after the launch, the client ranked first in Perplexity on every test question, and Perplexity quoted text verbatim that only exists on the new site.
In December a new full measurement shows how all four AI assistants have responded to the new website.
Four findings
- Cited is not the same as recommended. Perplexity can draw information from a company's website and still recommend its competitors.
- AI that does not search finds no sources. AI assistants not set up to search the web gave zero source references across 1,266 measured answers. So I measure the effect in the assistants that search.
- The assistants change on their own. When an AI assistant gets a new model or changes how it searches, results for a whole industry can change overnight. So I always measure a comparable company at the same time, so such changes are not mistaken for the effect of the measures.
- A good-looking website can be invisible. Many websites are built in a way that AI assistants cannot read the text on them. The business then cannot appear in their answers, however good the site looks.
More projects
-
Quote and document workflow automation for a travel company
October 2026, ongoing
I mapped every internal and outgoing document in the company and assessed twelve possible solutions. The result is a program where the owner sets up a tour with a few choices and gets emails, day programmes and quote letters where prices, lists and overviews agree with each other. He can also describe the tour in his own words to any AI assistant, attach a short template for how the AI should write its answer, and paste the answer straight into the program to get every document he needs. That makes the administrative work far more efficient and less time-consuming for the owner. The program itself contains no AI, customer data stays out, and prices are always calculated in Excel or code. Version 1 is complete.
-
Pedagogical development of the new master's course HSM302
NLA University College, August to September 2026
The lecturer, Maryna Vakulenko, knew me from before, among other things as a student assistant in statistics and research methods, and NLA asked me to record the lectures for the new master's course HSM302 Leading Responsible Innovation, planned at 20 hours. When we went through the scripts, we agreed they were too long: students quickly lose focus when a digital lecture runs longer than 15 minutes. She asked me to cut them down, but that broke the original plan of four five-hour recording sessions, because the time I had set aside to prepare now went into rewriting the scripts. So I proposed a new workflow and, on my own initiative, recorded a demo showing how it would work: I shortened a lecture, Maryna approved the script, I recorded it at home, and the sound engineer edited and finished it while I moved on to the next. Both Maryna and the sound engineer agreed to the new workflow after seeing the demo. When Maryna saw what I could contribute, I was given responsibility for the pedagogical development of all the lectures, and the assignment grew from 20 to 86.5 hours. I compressed 27 lectures from 30 to 15 minutes. Result: 7.5 hours of finished master's-level course material. In her letter of recommendation, Maryna writes that my input was valuable to the work on the master's programme and contributed positively to the development of the project.
-
Entrepreneurship Laboratory, Nairobi
NLA University College, course IML221, October 2024. Grade A.
The lecturer, serial entrepreneur Yngve Dahle, had planned to coach an entrepreneur far ahead of the rest of the class himself. He felt the standard programme would not challenge me enough and gave me the case to see how I would handle it. While the others followed a track set by the school, I had to design my own. I held eight meetings in two weeks and scaled the business model down to something that could actually be launched, by laying out what each working day would contain on the path he was stuck on.
Experience
-
Student assistant, NLA University College
Oslo, August 2024 to date
Teaching and supervision in HSM125 Entrepreneurship and Innovation, HSM126 Introduction to Statistics and Research Methods, HSM121 Organisation and Management and IML225 Sustainable Product and Service Development, and from November 2026 in IML222 Digital Marketing. Several of the courses have audiences of around 150 students. In most of them I follow up smaller groups closely, and I also give a fair amount of individual help with the material.
When a coursework deadline approaches, the lecturers often ask me to be available on campus and on Canvas so students can bring their questions to me. I then answer on my own on behalf of the course, and students use it because they know they will get an answer on the spot. The lecturers have asked me back every year since 2024.
-
Care assistant, Grauteknapp supported housing
Midt-Telemark municipality, May 2022 to date
More than 1,300 hours of social care, medication handling, record keeping and support for people with intellectual disabilities, alongside my studies and during holidays, and now on call. It showed me how big the difference is between technology that is easy to use and technology that is not, for the people who need it most.
Education
-
Bachelor's degree in innovation, marketing and management
NLA University College, Hauge School of Management, Oslo, 2023 to 2026
Grade A in the thesis, the Entrepreneurship Laboratory and Digital Marketing. Exchange semester at Griffith University, Australia, autumn 2025.
-
Thesis: AI adoption in small businesses, grade A
Spring 2026
Current AI use in small businesses is explained by perceived usefulness alone: those who use AI do so because it helps them, not because it is easy.
Intention to use more AI in the future, by contrast, is also shaped by how easy the tools feel and by business form, with sole proprietorships showing lower intention than companies with employees.
This asymmetry between current use and future intention was the thesis's main contribution, and it suggests that ease of use and business form matter more for the next wave of AI adoption than for the one that has already happened.
Quantitative study (N = 60), written in Norwegian. I had main responsibility for theory, method, analysis and writing, and led the supervision meetings.
Recommendations
"I immediately understood that Mr. Skalstad was the driving force among the two students, and that he was the obvious leader despite no formal authority."
Magne V. Aarset, CEO of Statcon AS, thesis supervisor. 19 October 2026.
"He takes initiative, thinks creatively and works purposefully, while being one of the key people in a team."
Maryna Vakulenko, Associate Professor, NLA University College. 13 October 2026. Translated from Norwegian.
Contact
Email is the fastest way to reach me, whether it's about an advisory role, a GEO measurement or a collaboration.