Learning Goals Overview

Key Figures

20

participants

6.0

average score

9

highest score

63%

experienced (score 5+)

Score Overview

All participants, sorted from highest to lowest self-assessment (scale 1–10).

Bo Bergstedt

9

Christian Jessen

9

Pelle Lykkebo Mørk

8

Emil Gjerding Nielson

8

Peter Møller

7

Lars Apel

7

David Buch

7

Anne Fuglsang Borg

6

Sebastian

6

Mads Buur Bach

6

Lasse Bergkvist Jessen

6

Mathias Overgaard

5

Joachim Saxtorph

5

Franziska Weiss Lauritzen

5

Mikkel Fruerboel Secher

5

Sanne Lau Pedersen

5

Nanna

4

Mads Oxlund Petersen

3

Jakob Hohlmann Villumsen

3

Marie Møller Munksgaard

—

Score 3: 2 (11%) Score 4: 1 (5%) Score 5: 5 (26%) Score 6: 4 (21%) Score 7: 3 (16%) Score 8: 2 (11%) Score 9: 2 (11%) Median: 6 Range: 3–9

Participant Grid

9 Bo Bergstedt

AI toolbuilding pioneer

Building anti-AI-slop tools

9 Christian Jessen

Skeptical expert

Video/image verification

8 Pelle Lykkebo Mørk

Newsroom AI implementer

Systematic OSINT methodology

8 Emil Gjerding Nielson

Editorial AI implementer

Automated feedback & writing

7 Peter Møller

Research & verification

OSINT & image verification

7 Lars Apel

News scanning & agents

Global news monitoring

7 David Buch

Daily AI user

Prompting & critical thinking

6 Anne Fuglsang Borg

Academic sparring & data

Ethical AI implementation

6 Sebastian

Basic ChatGPT user

Workplace AI tools

6 Mads Buur Bach

News bulletin AI developer

Verification & translation

6 Lasse Bergkvist Jessen

AI text modifier

Fact-checking & understanding

5 Mathias Overgaard

PimEyes & OSINT user

Free OSINT research tools

5 Joachim Saxtorph

Regular ChatGPT user

OSINT & verification

5 Franziska Weiss Lauritzen

Liveblog research

Criminal network OSINT

5 Mikkel Fruerboel Secher

Research & transcription

Verification & prompting

5 Sanne Lau Pedersen

Python monitoring attempts

Scraping & monitoring tools

4 Nanna

ChatGPT & Copilot basic

People research hacks

3 Mads Oxlund Petersen

Chatbot research

Scraping & document analysis

3 Jakob Hohlmann Villumsen

FoI & data processing

OSINT people finding

— Marie Møller Munksgaard

Expert research

Extensive AI research

Learning Goal Categories

Five categories based on what the newsroom itself indicated. One participant can appear in multiple categories.

14 / 20

OSINT & Verification

The dominant goal. Nearly everyone wants image/video verification, people-finding, or open source intelligence. This is not a nice-to-have — it is the core demand of the newsroom.

Peter Møller

"I want AI to help me with research in journalistic sources... and to verify images and video." (translated from Danish)

Mathias Overgaard

"I would like to learn about more smart and free AI tools that can be used for research purposes (OSINT)."

Joachim Saxtorph

"I really want to use AI as a research tool, especially in the realm of OSINT/verification of images, videos, etc."

Pelle Lykkebo Mørk

"Get more systematic with AI-assisted OSINT — verification of images, identities and locations."

Franziska Weiss Lauritzen

Interested in criminal journalism OSINT, monitoring criminal networks on social media. (translated from Danish)

Christian Jessen

"If AI can help with video / images verification and identifying places on a map."

Mads Buur Bach

"Verifying pictures and videos... Trust worthy translation from foreign languages"

Mikkel Fruerboel Secher

"Tools to make it easier to verify the authenticity of images/videos."

Jakob Hohlmann Villumsen

"When trying to find people, that do not want to be found, I would love to be able to get more assistance from AI to do OSINT"

Nanna

"if you have any hacks on researching and getting information on specific people"

Lasse Bergkvist Jessen

"I would also like to be able to use AI to fact check and/or check images and video"

Sebastian

"Want to know how I can use it in my work"

Anne Fuglsang Borg

"I would like to learn how to use AI to gather information even better, as well as sort it."

David Buch

"train my critical sense towards AI"

12 / 20

Research & Source Finding

More than half the group wants AI for finding, scanning, and analyzing sources. From scanning global news to extracting key information from large reports — the newsroom sees AI as a research accelerator.

Peter Møller

Research in journalistic and scientific sources. (translated from Danish)

Mathias Overgaard

"streamline my searches and to sort through and extract the most important information from large reports"

Lars Apel

"helping me scan news across the world and potentially pitch me stories"

Marie Møller Munksgaard

"get even better at using AI for extensive research that otherwise would have taken me a long time"

Mads Oxlund Petersen

"using AI to read documents and stuff like that. Like using AI to make a resume of 100 pages"

Sanne Lau Pedersen

"I would like to have better research tools fx scraping websites"

Pelle Lykkebo Mørk

Already uses AI in breaking news research — wants to sharpen methodology.

Franziska Weiss Lauritzen

"Search for information to use in articles." (translated from Danish: "søge information frem til brug i artikler")

Emil Gjerding Nielson

"Want to automate feedback and other skills"

Christian Jessen

"help me think outside the box or nuance my thoughts"

Anne Fuglsang Borg

"learn how to use AI to gather information even better"

Mads Buur Bach

"better verify news stories... don't have to wait for larger news organizations"

10 / 20

AI Skills & Prompting

Half the group asks for help steering AI — from prompting tips to understanding what is possible and what the pitfalls are. In the training they will learn that the core skill is not "write better prompts" but letting AI do the prompt work — the Prewash method: "Give me a prompt to analyse (subject)" instead of formulating it yourself.

David Buch

"Lear skills to prompt better and train my critical sense towards AI"

Mikkel Fruerboel Secher

"Better prompting in general."

Lars Apel

Wants a clearer picture of what's possible right now and in the near future.

Anne Fuglsang Borg

"I want to learn how to implement AI in my everyday life without compromising my ethical compass."

Lasse Bergkvist Jessen

"gain a larger understanding of the possibilities of AI and the pitfalls"

Emil Gjerding Nielson

"Insight into tools, prompts etc."

Nanna

"Teach me things I didn't know I needed"

Bo Bergstedt

"Seeing journalists getting more comfortable with AI"

Pelle Lykkebo Mørk

Wants sharper methodology, aware of editorial judgment risk.

Mads Oxlund Petersen

"learn things like scraping information"

8 / 20

Tool Orientation & Automation

Eight journalists want to know which tools exist and how to automate recurring tasks. Not learning one tool — but understanding the landscape and making things work.

Sanne Lau Pedersen

"setting up search mashines on certain websites that notifies me when certain words are used"

Lars Apel

"helping me scan news across the world" — has tried agents with limited success.

Lasse Bergkvist Jessen

Tried CoPilot Agent for Power BI data — hit challenges.

Mads Buur Bach

"Edit videos e.g. applying blurred faces etc. easier"

Bo Bergstedt

"Building tools to help journalists"

Emil Gjerding Nielson

"Want to automate feedback"

Mathias Overgaard

"learn about more smart and free AI tools"

Sebastian

"Just that I get some tools, that I can use in my daily work"

4 / 20

Data Security & Ethics

Four participants explicitly raise concerns about data security, editorial independence, and the reliability of AI-generated knowledge. These are not beginner concerns — they are institutional policy questions that the training must address head-on.

Pelle Lykkebo Mørk

"navigating the use of commercial LLMs in a newsroom context — specifically around data security, copyright"

Anne Fuglsang Borg

"implement AI in my everyday life without compromising my ethical compass"

Christian Jessen

"I am very skeptical with AI as a tool for generating knowledge"

David Buch

"train my critical sense towards AI... pros AND cons of AI"

Current Tools in Use

Which AI tools the newsroom already uses — and how broad the landscape is.

ChatGPT

75%

15/20

Copilot

25%

5/20

Image gen

15%

3/20

Gemini

10%

2/20

PimEyes

10%

2/20

Transcription

10%

2/20

6

different tools

1.8

tools per person (avg)

75%

use ChatGPT as base

Success Criteria

When is the training a success? In the newsroom's own words.

9

Bo Bergstedt

"Seeing journalists getting more comfortable with AI and the many new options"

9

Christian Jessen

"If I find completely new ways to use AI to solve complex issues. And most importantly, if AI can help with video / images verification"

8

Pelle Lykkebo Mørk

"Success is leaving with concrete, immediately usable tools and methods"

8

Emil Gjerding Nielson

"Insight into tools, prompts etc. that I can use in my daily work. Reflections on possibilities."

7

Peter Møller

"If I learn new tools and techniques for research and verification." (translated from Danish)

7

Lars Apel

"Getting a clearer picture of what is possible right now and in the near future"

7

David Buch

"To master the skills i described before"

6

Anne Fuglsang Borg

"I would love to gather more knowledge about AI in general."

6

Sebastian

"Just that I get some tools, that I can use in my daily work"

6

Mads Buur Bach

"Knowledge about and ready to use tools so that I can better verify news stories"

6

Lasse Bergkvist Jessen

"It would be a success for me if I am able to gain a larger understanding of the possibilities of AI and the pitfalls"

5

Mathias Overgaard

"If I receive concrete tools that can improve my research and make my work more efficient"

5

Joachim Saxtorph

"If I learn something new that is easily integrated into my everyday worklife"

5

Franziska Weiss Lauritzen

"That I get lots of knowledge, ideas and tools for how to integrate it in my work" (translated from Danish)

5

Mikkel Fruerboel Secher

"Learning new techniques and technologies to use in my work. New ways of thinking."

5

Sanne Lau Pedersen

"That I get new concrete research tools that I can use in my day to day life"

4

Nanna

"Teach me things I didn't know I needed"

3

Mads Oxlund Petersen

"If I get at least one skill og Roll that I Can use afterworths."

3

Jakob Hohlmann Villumsen

"If I in the future can have smarter ways to connect the dots and find out crucial information about people on Facebook, Instagram etc"

—

Marie Møller Munksgaard

"If I go home with one or two new ways to use AI in my everyday work life that I didn't know of."

Frustrations & Attention Points

Signals from the intake form that the training must address.

Deep skepticism about AI accuracy

Lars: "I still find googling stuff myself faster than using AI." Mathias: spends "considerable time double-checking." Peter: AI "invents answers and invents sources." This is not resistance — it is earned journalistic skepticism. The training must demonstrate reliability, not assert it.

Failed automation attempts

Sanne tried Python+Task Scheduler for website monitoring — "it never worked." Lars tried agents for news scanning — "with little luck." Lasse tried CoPilot Agent for Power BI — hit challenges. Three journalists independently attempted technical automation and all failed. The training must show what actually works.

Data security & editorial independence concerns

Pelle explicitly raises "data security, copyright, and ensuring sensitive or unpublished material isn't fed into models that may train on it." He also warns against "offloading independent editorial judgement to AI." These are not beginner concerns — they are institutional policy questions.

The Bo Bergstedt factor

Bo (score 9, deepfakes since 2018) explicitly writes: "don't customize the course for me — I am an outlier at TV2." He wants to build tools for colleagues. Use him as co-trainer, not as student. His goal aligns with the training itself: helping journalists become comfortable.

Danish-language participants

Peter, Franziska, and Nanna responded in Danish. This signals they may be more comfortable working in Danish. Some AI tools perform differently in Danish vs English. The training should address multilingual prompting.

Training Recommendations

Six concrete recommendations based on the intake analysis.

1

Start with OSINT and verification — not general AI

70% of participants want verification skills. Begin with a live demonstration: take a current news image and show how AI verification works in 90 seconds. Use Pelle's E-3 Sentry case as real material. This is not a warm-up — this is the core demand.

2

Teach the Prewash Method — not prompting

Instead of teaching "write a good prompt," teach the newsroom to let AI do the prompt work: "Give me a prompt to analyse (subject)". This addresses both David's desire for better prompting and Lars's frustration with unreliable results.

3

Differentiate with Bo as pivot point

The group has three levels: beginners (3-4: Jakob, Mads O., Nanna), middle (5-6: Mathias, Joachim, Franziska, Mikkel, Sanne, Anne, Sebastian, Mads B., Lasse), and advanced (7-9: Peter, Lars, David, Pelle, Emil, Bo, Christian). Use Bo and Pelle as peer-trainers for the advanced group. Jakob's ambition to "connect the dots" on social media is the compass.

4

Address the data security question directly

Pelle's concerns about feeding sensitive material into commercial LLMs reflect a TV 2 institutional issue, not a personal one. Include a concrete block on which tools train on your data and which don't. This is not optional — it is a prerequisite for newsroom adoption.

5

Demonstrate the difference between analyzing and summarizing

Multiple journalists use AI "as a search engine." Lars "still finds googling faster." Show what goes wrong with "summarize this" versus what happens with "analyze this document for unsupported claims and methodological gaps". The difference is the difference between a parrot and a research assistant.

6

End with a personal action card

Each participant leaves with three concrete actions tied to their own work. Jakob: OSINT patterns on social media. Lars: news monitoring that actually works. Pelle: systematic verification methodology. Sanne: monitoring tools that replace her failed Python setup. Christian: image/video verification pipeline. No abstract tips — only actionable steps.