AI organisatorisk förändring måste behandlas som social negotiation process, inte technical integration, visar CHI 2026-studie med 15 UX designers (Georgia Tech). Adoption unfolds across three scales där efficiency discourse carries hidden social dimensions: Individual level (efficiency vs self-worth conflict), Team level (reconfiguring responsibilities, trust erosion), Organizational level (culture mirror, decision-making power imbalance). För dig som projektledare betyder detta: treating AI som bara productivity tool = guaranteed failure. 86% ledare säger “efficiency gains” men workers experience deskilling, role threats, autonomy loss. AI adoption är site där roles/relationships/power reconfigured – requires values negotiation, not tool deployment.
Tre skalor: Individual, Team, Organization – conflicting värden
CHI 2026 design workshops + follow-up interviews:
- 15 UX designers (startups to large tech companies)
- USA, Canada, South Korea
- Sectors: finance, healthcare, education, IT
Methodology: Scenario-building exercise (hypothetical organizational situations)
- Follow-up reflections
- Five-stage adoption framework (Discuss, Explore, Test, Select, Adapt)
Core finding: “The central issue is not whether AI adds value, but what does value actually mean“
Individual level: Efficiency vs Self-Worth paradox
Primary motivation: Efficiency & productivity
- P8: AI made them “10 times better”
- P12: “mechanism for improving efficiency”
- Automate repetitive work (resizing wireframes, documentation, presentations)
Men hidden tensions emerge:
P14 (honest admission): “Sometimes makes me think, do I rely on AI tools too much? I’m afraid of losing my value within organization… automating workflow… seems I might be easily replaceable by leadership. I feel losing my core skills—taking a gamble here.”
P9 (junior designer worry): “Even though faster, more efficient… got to think what muscles are not being used as result of using these tools.”
Hidden labor overlooked:
- Repeated prompting, trial-and-error
- Constant checking, adjustment
- Correcting errors, hallucinations
- P11: “I struggle getting very relevant outcomes straight away. Have to do lot of hidden trials more often.”
För projektledare: Efficiency metrics ignore invisible work. Designers spend MORE time troubleshooting AI than saved through automation. Plus professional anxiety about skill atrophy, replaceability.
Team level: Reshaping roles, eroding trust
Informal diffusion patterns:
- Tools spread via “word of mouth”, “lunch & learns”
- Early adopters = “power users” demonstrating workflows
- P13: “Market approved tools” (ChatGPT) vs “false advertising” (LinkedIn promotions)
Role boundary conflicts:
P2 & P3 (concern för content designers): “There’s sentiment around their job security… since LLM deals with human language, lot of potential for content generation” “Emotional burden particularly for those whose responsibilities overlapped most directly with AI”
P14 (radical prediction): “Less than year, junior engineer replaced easily. I submitted PR to GitHub even though don’t know how to code. If design, product vision, coding aligned by one person, workflow much faster. Sooner or later word ‘developer’ will be gone.”
Trust erosion via transparency failure:
P4 & P6 warning: “Magic eight-ball thinking” – accepting AI outputs uncritically
- Colleagues submitted strategy documents containing fabricated links
- UX researcher fed large datasets without constraints → plausible but inaccurate outputs
P9 (transparency essential): “Important communicate whether you used AI for specific task before pass on to another member… important have transparency fact you used AI tool”
P2 (productive conflict lost): “Tendency considering results from AI as easy answer… less political. People not eager debate or discuss or come up with own ideas… get lazy, consider AI answer be best solution, not think further. Less passionate, less thoughts included in design.”
För projektledare: AI doesn’t just redistribute tasks – it restructures relationships, accountability, professional identity. Without transparency protocols + structured oversight, teams develop “convenience bubble” där quick answers prioritized över collaboration/learning/rigor.
Organizational level: Culture mirror, not neutral tool
Leadership vs practitioners values conflict:
Leadership framing: Efficiency, productivity, cost savings Workers experiencing: Deskilling, craft loss, role threats
P10: “Business just wants use AI to generate UI, and it’s currently trash. Should I just keep hammering it until magic happens?”
Adoption decision-making power:
- Formal adoption = management-driven
- Practitioners positioned as “end-users” not decision-makers
- P4 & P5: Adoption “something managers like P6 decided”
Compliance vs innovation tension:
- Large firms: Cumbersome approval processes outlasting project timelines
- P13: “Sometimes get compliance approval in week 7 of 8-week project”
- Healthcare (P15): HIPAA slows adoption significantly
- Client-facing industries: Client bans override internal desires
Organizational culture reflected:
P9 (culture questions): “How would affect company culture, how workers actually work together… Will team members have differing opinions about AI being integrated? What systems/practices will be put in place resolve these conflicts?”
P12 (overwhelmed): “Emotion I have is I need wrangle this thing… figure out how grow as using it, but not be overcome by fact it has power do multiple different things. We all have sense this going wipe out so many roles, but how do we find ourselves shape shifts in mix of all it?”
För projektledare: AI adoption mirrors existing power dynamics. Top-down implementations reinforce managerial control över worker autonomy. “Efficiency” becomes code för standardization, surveillance, de-skilling – not genuine productivity gains.
Values vs Value: Graeber’s distinction critical
Graeber’s framework (anthropological):
- Value (economic): Exchange, productivity, efficiency
- Values (social): Ideals of what is good/desirable
CHI study shows these are ENTANGLED: Even “efficiency” discourse carries social dimensions:
- Responsibility (who accountable för AI outputs?)
- Trust (can we rely on AI-generated insights?)
- Autonomy (do workers control AI integration?)
P15: “Going one level deeper to ask what concrete goals new tool meant achieve”
Efficiency är contested site där:
- Economic value claims (faster outputs)
- Collide with social values (professional worth, collaboration, rigor)
För projektledare: Treating efficiency as neutral metric = fundamental mistake. Every “productivity gain” claim involves hidden trade-offs: whose labor becomes invisible, whose expertise devalued, whose autonomy constrained.
Fem praktiska recommendations
1. Treat adoption as organizational change, not tool deployment:
- Design workshops surfacing value conflicts
- Multi-stakeholder deliberation (not just management decision)
- Explicit negotiation: what are we optimizing? Whose values prioritized?
2. Implement transparency protocols:
- Citation-like practices distinguishing human/AI contributions
- Disclosure requirements før AI-generated outputs shared
- P9’s principle: “Communicate whether used AI before pass on”
3. Measure hidden labor:
- Track prompting time, error correction, verification effort
- Don’t just count outputs (designs created, documents written)
- Include “invisible trials” (P11) i productivity calculations
4. Protect professional development:
- Junior developers/designers need “muscle-building” opportunities
- Balance AI efficiency gains mot skill atrophy risks
- P9: “What muscles not being used as result of using these tools?”
5. Worker agency i adoption decisions:
- Not just “can workers use AI?” but “can workers decline without penalty?”
- P15: Leaders foster safe dialogue space, clearly explain benefits, engage teams
- Inspiration från Writers Guild of America (WGA) negotiations: union-backed boundaries på AI scriptwriting use
Bottom line
AI organisatorisk förändring requires treating adoption som values negotiation, not technical integration. CHI 2026 studie (15 UX designers) visar efficiency discourse carries hidden social dimensions across tre scales. Individual: efficiency vs self-worth paradox, hidden labor overlooked (prompting, troubleshooting, verification). Team: role boundaries blur, trust erodes via transparency failures, productive conflict replaced by “convenience bubble”. Organization: leadership frames efficiency, workers experience deskilling/role threats, adoption mirrors existing power dynamics/culture, decision-making power concentrated i management. Graeber’s distinction: value (economic) entangled med values (social) – efficiency är contested site. Practical pathways: treat som organizational change, transparency protocols, measure hidden labor, protect skill development, worker agency i decisions. Future research: examine how AI redistributes responsibility, reshapes relational labor, strengthens/undermines worker power.
Källa: “The Values of Value in AI Adoption: Rethinking Efficiency in UX Designers’ Workplaces” av Inha Cha, Catherine Wieczorek & Richmond Y. Wong, Georgia Institute of Technology, publicerad april 2026 (CHI ’26 Conference, Barcelona).