AI and the organisation of workNorway
KI-avtalen
AI agreement
Letter of intent signed in October 2026 by Norway’s government and 8 labour-market organisations to cooperate on responsible use of AI at work.
- Type
- Movement or network
- Origin
- Government of Norway with LO, NHO, Unio, KS, Akademikerne, Virke, YS and Spekter
- First documented
- 2026
- Last reviewed
- 17 November 2026
In short
KI-avtalen is a letter of intent signed on 2 October 2026 by Norway’s government and 8 main labour-market organisations. It commits them to cooperate on responsible introduction and use of artificial intelligence in working life until 2029. A coordination group led by the digitalisation minister will first agree on common principles. Concrete measures remain to be written.
Where it comes from

Unio announced on 25 September 2026 that government and social partners would sign an agreement on artificial intelligence (Unio, 2026a). Prime Minister Jonas Gahr Støre and the leaders of 8 main organisations signed it on 2 October 2026 (Spekter, 2026).
Signing took place at the Ministry of Digitalisation and Public Administration, in Oslo’s new government quarter (Akademikerne, 2026). Unio links the agreement to an expert committee report, “De ti KI-bud”, published the previous year (Unio, 2026b).
Norway already regulates technology change at company level through collective agreements. Lasse Maugesten, of consultancy TRY, ties the new agreement to a 1975 framework agreement on technology at each workplace (Maugesten, 2026).
- 2025SINTEF finds weak union involvement in AI processes, in a report for Akademikerne
- 2025LO and NHO revise their Basic Agreement for 2026 to 2029 without an AI clause
- 2026Unio announces on 25 September that an AI agreement will be signed
- 2026Government and 8 main organisations sign KI-avtalen on 2 October
- 2026Coordination group set to hold its first meeting on 27 October
How it works
- Legal form. It is an “intensjonsavtale”, a letter of intent, between the state and the main organisations (YS, 2026).
- Parties. Employers are represented by NHO, KS, Virke and Spekter. Workers are represented by LO, Unio, Akademikerne and YS (YS, 2026). Prime Minister and digitalisation minister sign for government (Regjeringen, 2026).
- Duration. The agreement runs from 2 October 2026 to 31 December 2029 (KI-avtalen, 2026).
- Coordination group. Minister Torgeir Micaelsen leads a coordination group of parties, state agencies and research communities (Spekter, 2026). It meets up to 6 times in its first year (YS, 2026).
- Themes. Parties cooperate on productivity, staff skills, preventing exclusion and harm to working environment, and sharing knowledge on responsible use (Regjeringen, 2026).
- Legal frame. AI is to be introduced in line with Working Environment Act and established codetermination arrangements (Regjeringen, 2026).
- Group members. Members include KI Norge, Datatilsynet, Arbeidstilsynet, Statens arbeidsmiljøinstitutt and research centres TRUST and AI LEARN (Regjeringen, 2026).
- First task. The group’s first task is to agree on principles for responsible use (Spekter, 2026). Its first meeting was set for 27 October 2026 (YS, 2026).
- Open agenda. Unio lists 5 questions for discussion, including which decisions must stay with humans and how gains can reach society as a whole (Unio, 2026b).
Who uses it
The signatories are the Government of Norway and 8 organisations: NHO, KS, Virke, Spekter, LO, Unio, Akademikerne and YS (YS, 2026; Akademikerne, 2026). Each organisation has published its own statement on the agreement.
No individual employer is named as applying the agreement in sources read for this entry. Spekter reports AI adoption among its members without naming one that applies the agreement (Spekter, 2026).
What the evidence says
Peer-reviewed research
No peer-reviewed study of KI-avtalen was found; the agreement was signed in October 2026. Research on its institutional base exists.
Lloyd and Payne (2026) compared food and drink processing in Norway and the United Kingdom through interviews with union officers and shop stewards. They found that workplace union strength had a larger effect on unions’ role in digitalisation in Norway.
Surveys and reports
Kamsvåg, Andersen and Landmark (2025), at research institute SINTEF, studied AI in knowledge work for Akademikerne. In a panel of 1,603 respondents, 6 in 10 did not know whether union representatives took part in AI processes.
The same SINTEF report found that on average 20% of union representatives were involved in procurement, development or implementation of AI. It found no statistically significant difference between private, state, municipal and hospital sectors (Kamsvåg, Andersen and Landmark, 2025).
Ødegård, Andersen and Aspøy (2025), at research foundation Fafo, surveyed LO union representatives for LO. Of 1,071 respondents, 58% said new technology had led to new ways of working and 7% reported job cuts linked to it.
In the same Fafo survey, 30% of union representatives rated their competence to discuss new technology as high or fairly high. Another 35% rated it low or fairly low (Ødegård, Andersen and Aspøy, 2025).
Spekter, an employer organisation, reports its own member survey. The share of members that had adopted AI rose from 55% in 2025 to 78% in 2026 (Spekter, 2026).
Official data and company documents
LO and NHO revised their Basic Agreement for 2026 to 2029 on 28 November 2025. The negotiation protocol contains no new provision on artificial intelligence (LO and NHO, 2025).
Spread level 4 of 6
Independent coverage. Reported by business press or business-school cases.
Limits
- Non-binding. The agreement is a letter of intent; Unio states it contains no finished solutions (Unio, 2026b).
- Content still open. Principles and measures are left to the coordination group; LO’s deputy leader Henriette Jevnaker said it was too early to name concrete measures (Håbrekke, 2026).
- No link to collective agreements yet. The LO and NHO Basic Agreement for 2026 to 2029 added no AI clause (LO and NHO, 2025).
- Weak starting point. Before signing, 20% of union representatives on average were involved in procuring, developing or implementing AI (Kamsvåg, Andersen and Landmark, 2025).
- No evaluation. No measurement method or evaluation of the agreement was found in sources read for this entry.
Criticism
- Lasse Maugesten, TRY. He argues the follow-up should oblige parties to keep a shared picture of which tasks AI takes over in Norwegian businesses. He notes that national statistics show who uses AI, not which tasks disappear (Maugesten, 2026).
- Spekter. Strategy director Tonje Sagstuen says detailed political steering, unpredictable funding and unclear rules slow organisations down. Spekter will ask the coordination group for less detailed regulation (Spekter, 2026).
- Unio. Leader Steffen Handal warns that AI could increase inequality if gains concentrate among a few owners and technology companies (Unio, 2026b).
- Akademikerne. Leader Lise Lyngsnes Randeberg calls current union involvement in AI “too poor”. An Akademikerne panel in June 2026 found 1 in 4 respondents reported no union involvement (Akademikerne, 2026).
Would it work where you are? Free account
Free with an account
- Would it work where you are? On every practice, 4 conditions (law, culture, economics and organisation), each with a verdict
- Spread alerts when a practice moves up the spread scale
- Trendemic Weekly, the free email that recaps the practices of the week
The culture this practice needs
Nine cultural conditions decide whether KI-avtalen can take root in an organisation. This profile shows the level each condition needs and which ones are critical.
Adoption toolkit Decide whether it can live in your organisation
- The cultural profile of this practice: the nine conditions it needs and the ones that are critical
- The assessment “Can this leadership practice live in your organisation?”
- A one-page decision note, a 90-day test plan and the objections to expect
- Comparisons between practices and a workshop deck
Adoption toolkit, €249 a year, includes Field Notes. One practice can also be unlocked for €49. Opening soon.
Frameworks. The Leading, Deciding, Evaluating and Disagreeing scales come from Erin Meyer, The Culture Map (PublicAffairs, 2014). The trust, risk-taking and openness elements come from Steve Gruenert and Todd Whitaker, School Culture Rewired (ASCD, 2015). The autonomy and cross-team cooperation conditions are by Benjamin Chaminade. The profile of this practice is an editorial assessment by Trendemic, based on the sources of this entry.
Sources
- Regjeringen (2026). Regjeringen og partene i arbeidslivet inngår KI-avtale. Press release, 2 October 2026. regjeringen.no
- KI-avtalen (2026). Intensjonsavtale om ansvarlig innføring og bruk av kunstig intelligens i arbeidslivet. Government of Norway and 8 main labour-market organisations. regjeringen.no
- Unio (2026a). Partene i arbeidslivet og regjeringen inngår KI-avtale. Unio. unio.no
- Unio (2026b). Den norske modellen kan bli Norges KI-fortrinn. Unio press release via NTB Kommunikasjon. kommunikasjon.ntb.no
- YS (2026). Har inngått avtale om kunstig intelligens i arbeidslivet. Yrkesorganisasjonenes Sentralforbund. ys.no
- Akademikerne (2026). Signerte KI-avtale. Akademikerne. akademikerne.no
- Spekter (2026). Spekters medlemmer er godt i gang allerede. Spekter. spekter.no
- LO and NHO (2025). Protokoll fra forhandlingene om revisjon av Hovedavtalen LO-NHO. LO. lo.no
- Kamsvåg, P. F., Andersen, T. K. and Landmark, A. D. (2025). Kunstig intelligens i kunnskapsarbeid og samspillet med den norske modellen. SINTEF Digital, report 2025:00627, for Akademikerne. akademikerne.s3.amazonaws.com
- Ødegård, A. M., Andersen, R. K. and Aspøy, T. M. (2025). Tillitsvalgtpanelet. Oppsummering av resultater fra 2024. Fafo-rapport 2025:25. fafo.no
- Lloyd, C. and Payne, J. (2026). Digitalisation, unions and ‘country-effect’: does union strength at the workplace matter? Journal of Industrial Relations, 68(3), 357 to 382. journals.sagepub.com
- Håbrekke, J. (2026). Spillereglene i norsk arbeidsliv må gjelde for tek-gigantene òg. FriFagbevegelse (LO Media). frifagbevegelse.no
- Maugesten, L. (2026). Kina har en plan for KI og arbeidstakerne. Vil vi få det samme i Norge? Computerworld Norge. cw.no