17) Jonas Slørdahl Skjærpe om svinghjuleffekter, liminalfaser og kunsten å ta heis

Truls Jørgensen og Audun Fauchald Strand — Heis Podcast · podcast · 2026-05-18 source ↗ Reader 🖥️

mixed · narrative-argumentative — long-form interview where the guest's career anecdotes (NAV transformation, COVID, going consultant, Silicon Valley trip) carry an argumentative spine about how complex sociotechnical systems change — Hosts Truls Jørgensen and Audun Fauchald Strand interview their former boss Jonas Slørdahl Skjærpe, now partner for technology consulting (Financial Services) at EY, formerly IT-direktør at NAV. The hosts ask short questions; the guest answers at length. Tone is collegial — all three worked together at NAV — so claims are stated with practitioner certainty rather than defended at length. The conversation moves through three loosely chained arcs: (1) the NAV transformation in retrospect, (2) leadership across consultant/leader/professional modes, (3) AI as the next paradigm shift.

Structure

Key terms

• svinghjul / svinghjuleffekt (flywheel / flywheel effect) Technology Diffusion & Transformation Curves
  Slow-building organisational momentum where many capability investments compound silently and then deliver disproportionate returns when an event arrives. Used twice: about NAV's IT capability before COVID, and about transformations more generally — 'energetic processes that happen slowly first, then fast'.
• liminalfase (liminal phase) Technology Diffusion & Transformation Curves
  The transitional phase between two paradigms — after the old rules are abandoned and before the new ones are established. Characterised by friction, higher risk, and the need to let some old structures go before the new can settle. Jonas attributes the term to social anthropologists he worked with at NAV. He applies it both to NAV's transformation and to the current AI moment 'on almost all fronts'.
• kunsten å ta heis (the art of taking the elevator) Technology Leadership (Heis)
  The capability of moving fluently across organisational floors (executive, middle, team) and across silos, translating language and concerns between them. Jonas treats it as both a personal joy and a strategically critical capability for transformation, because organisations leak information vertically and horizontally and need traversal specialists. Requires security in your craft, humility about other perspectives, and metaphor/model fluency. Explicitly trainable.
• S-kurve (S-curve) Technology Diffusion & Transformation Curves
  The adoption/diffusion curve — slow start, steep middle, plateau — invoked to argue that many simultaneous S-curves were rising during the NAV transformation (deploy frequency, share of developers, team formation) and that AI in mid-2026 is still 'very early on the curve'. Distinct from a 'kumulativ' curve.
• fra opposisjon til posisjon (from opposition to position) Technology Diffusion & Transformation Curves
  The transition phase in transformation when the change agents stop fighting and become the institution that invites others in. 'Fighting' stops working past a certain point; one must occupy the structures and meet people where they are.
• fagtom zeitgeist (craft-empty zeitgeist)
  A moment when the field is hungry for substantive learning — books circulated, ledergruppe homework, high book-reading activity — because the discipline is collectively re-skilling for a new paradigm. Used to describe the post-Accelerate NAV moment.
• sosioteknisk arkitektur (sociotechnical architecture) Sociotechnical Architecture
  The integrated design of social/organisational structures together with software architecture. Jonas says discovering he found this kind of system design fun was what made him realise leadership and IT-arkitektur were the same problem in different clothes.
• trykkoker (pressure cooker)
  A successful transformation acts as a pressure cooker that forces organisational floors and silos closer together because tasks now cross hierarchies and silos simultaneously. People riding the elevator become valuable in this state.
• retenking av abstraksjonsnivå (rethinking the abstraction level) Organizational Abstraction Levels
  What happens periodically in software: a new technological wave (high-level languages, OS/DB, internet, cloud) raises the level at which humans express intent, changing what skills are needed and how work is organised. Jonas argues generative AI is doing this again, moving from 'high-level instructions' to 'intentions, with instructions written by machines'.
• AI slop AI Slop
  Low-quality AI-generated output that nobody actually understands or owns. Jonas claims everyone has personal responsibility to ensure what they produce is of quality and that they comprehend it — naming this in a team meeting that morning.
• læringsprosess vs produksjonsprosess (learning process vs production process) Learning Culture & Slack
  Personal heuristic for AI use: at any given moment you are either learning with AI's help or producing with AI's help, and the two cannot be collapsed. You cannot skip the learning that leads up to the production output, because AI presents convincing-looking output that may be substantively poor.
• traktor-analogien (the tractor analogy) Technology Diffusion & Transformation Curves
  Tractors arrived in the US around 1900 but did not displace horses until the 1950s because the product was immature, skills were scarce, and farms were too small to justify capital-intensive operation. Jonas uses this to argue that AI's real impact will unfold over decades of restructuring, not in a couple of years.

Propositions

• explicitcorroboratedLarge organisational transformations follow an S-curve composed of many simultaneous smaller S-curves (deploy frequency, team formation, developer share, etc.), each switching on around the same time.conf: high
  supports ←DORA Community Discussion - Operationalizing DORA within a large organization · p14 — The multi-year timeline (2019→2025) and 'it takes more than a minute to change an organization' instantiate the claim that large organizational transformations follow an S-curve made of many simultaneous smaller S-curves and cannot be shortcut.
• explicitdevelopedBetween paradigms there is a liminal phase characterised by high friction and risk that cannot be shortcut — old structures must be released before new ones settle.conf: high
  extends ←Én enkelt behandling kan gi svært langvarig kolesterolsenkning: – Noe radikalt nytt · p7 — Retterstøl's 'years before this reaches the pharmacy' timeline is the same high-friction liminal-phase pattern (old structures/practices must be worked through before a new paradigm settles), applied to biomedical technology rather than software/AI transformation.
• explicitCapability built years in advance is what makes apparently sudden transformations survivable; NAV's COVID performance was not luck but evidence of correct prior investments.conf: high
• explicitCrisis is the easiest leadership mode because strategic alignment is given; non-crisis leadership requires actively constructing and maintaining a shared narrative.conf: high
  refines5) Vidar Moe: Det eneste Vidar vil er å ha det lett · p5 — Skjaerpe explains why crises drive structural progress — alignment comes for free — and adds the harder condition: non-crisis leadership must actively construct the shared narrative crises provide automatically.
• explicitA change agent must eventually transition from opposition to position — fighting stops working, and you must instead become the institution that invites others in and meets them where they are.conf: high
• explicitLeading at very large scale (1000+) requires giving up on tracking what individuals think of you and instead trusting your own judgment — a freedom not available at smaller scale.conf: high
• explicitThe best partner-consultant role combines owner, leader and practitioner — being 'on the problem' is part of the value, not a confusion of modes.conf: high
• explicitGenerative AI is a fundamental abstraction shift — from high-level instructions to expressing intent — and we are at the start of its S-curve, not somewhere on the plateau.conf: high
• explicitA '10% productivity gain from Claude Code, therefore AI is not transformative' take is a category error: it compares decades of pre-LLM work against a few weeks of AI-assisted work, one on a plateau, the other at curve-start.conf: high
• explicitdevelopedAI's real impact will require decades of organisational and societal restructuring — like the tractor (1900s tech, 1950s adoption) or the car (decades to reshape cities) — not a few years.conf: high
  extends ←Ordinary Engineers, Not Heroic Inventors · p9 — Supplies the concrete historical mechanism (Paul David's electrification account) behind the claim that AI needs decades of organizational restructuring, like the tractor or the car.
• explicitEven more AI productivity will hit a complexity ceiling: faster output means more software, which forces new tools and processes to manage complexity, slowing things back down before rising again.conf: high
• explicitcorroboratedWhen using AI you are either in a learning process or a production process; you cannot skip the learning that leads to the production output, because the model is built to make you feel smart and not question its output.conf: high
  extends9) Nils Brede Moe om AI-paradokset, autonomi, læringskultur og fremtidens team · p7 — Skjaerpe generalizes the AI-threatens-learning claim into a learning-vs-production mode distinction: you cannot skip the learning behind production output because the model is built to make you feel smart.
  supports ←Ordinary Engineers, Not Heroic Inventors · p10 — Both claim the organizational learning process cannot be skipped or shortcut on the way to real productivity gains.
• explicitLarge organisations leak information both vertically (across hierarchical floors) and horizontally (across silos), and the people who can traverse these gaps are 'invaluable' — especially in transformation.conf: high
• explicitRiding the elevator well requires (a) security in your own craft, (b) humility that others hold perspectives you lack, (c) fluency with metaphor, model and analogy; it is trainable, not innate.conf: high
• explicitLeadership credibility comes from being able to ground executive-level advice in something concrete from the lower floors — but this can become a hindrance if every detail must be 'pitch perfect' before speaking up.conf: high
• explicitEven understanding others' views does not require agreeing with them; the leader's duty is to understand the position well enough that disagreement is real, not mistaken.conf: high
• explicitThere is no way the labour market still works if no juniors are hired — a 'staircase' must exist from university into the profession; the question is how juniors learn well, not whether they exist.conf: high
  refinesSMOC.AI's Post — 5 Minutter: AI og utviklere · p5 — Skjaerpe accepts that traditional junior training is breaking down but insists a staircase into the profession must exist, reframing the question from whether juniors exist to how they learn well.
• explicitThe European AI position is structurally weaker than the US's because of investment intensity and concentration — sovereignty considerations matter even within US redistribution debates.conf: medium
• explicitAcross many leader interviews on Heis, the strongest pattern is not technique but epistemic posture — the leaders worth listening to are 'wondering and questioning', rarely doctrinaire.conf: high
• explicitThe podcast itself exists because two leaders who used to translate between organisational floors together wanted to keep doing that work after their jobs no longer required it of them.conf: high
• explicitcorroboratedBig personal-career outcomes have a large luck component — being in the right place at the right time to learn the right things from the best people.conf: high
  supports ←How Anthropic Builds And How Engineering Will Change Soon | Thariq Shihipar · p28 — Both claim large personal career outcomes have a substantial luck/timing component, with visibility and being in the right place as a controllable input to that luck.
• implicitThe healthy posture toward AI is neither maximalism nor refusal but a middle position that takes vendor skepticism seriously while imagining societal consequences rigorously.conf: high
• explicitIntroversion is compatible with senior leadership if you deliberately surround yourself with people who complement your weaker registers (small-talk, large social settings).conf: high
• explicitDoing the actual craft alongside leadership is energy-positive for an introvert leader — substantive discussion gives more than it takes, unlike small-talk.conf: medium

Arguments

Narratives

• First summer job recommendation: Jonas had to recommend either XP or Rational Rose as a process for software development. He picked Rational Rose because 'just wait for the next version, it will be great' — and later realised he was wrong.
  → Frames the speaker as someone who has lived being on the wrong side of a paradigm shift; this licenses the later epistemic humility about AI.
• Accelerate (the book) became homework for the NAV IT leadership team; Jonas wanted to understand what the research actually said and what the implications were. The hosts recall that DevOps practitioners had been claiming these things for years, but the book gave them research legitimacy.
  → Illustrates the 'fagtom zeitgeist' and shows leaders performing the same learning they were asking their organisations to do.
• There were moments during the NAV transformation when Jonas felt 'now we are past the point of no return.' He notes it could have gone wrong — and they could 'have ended up in stortinget's kontroll' — but never felt strong personal pressure.
  → Captures the lived experience of the liminal phase: high risk in retrospect, equanimity in the moment.
• COVID hit NAV with twelve major events at once including a 60x traffic increase ('Norway becoming USA-sized overnight'). Anders Egge drove out laptops alongside scarce toilet paper during the 'dopapir-krise'. Many employees jumped onto Slack on the announcement weekend and worked through.
  → Dramatises both the multi-S-curve nature of the crisis and the cultural readiness that prior investments had built.
• The dagpenger 'forskudd' story: With 400,000 people permitted overnight, the regulations required individual case-by-case judgment, which was impossible. A 'sharp jurist' realised they could create a 'forskudd' (advance) instead of a new benefit, which propagated rapidly through the political top and reached NAV developers; a working solution shipped by Tuesday.
  → Illustrates strategic alignment from political top to delivery team that Jonas argues is otherwise nearly impossible to engineer.
• Silicon Valley visit: Jonas met startup CEOs working on AI, visited Stanford, and was struck by both the investment intensity and the 'quite drøye business in det' — minimal guardrails on biology, industry, tech.
  → Sets up the AI section's combination of awe and unease, and the European-sovereignty concern.
• Tractor analogy: Tractors arrived in the US in the early 1900s with metal wheels (got stuck in soft soil), scarce maintenance skills in rural areas, and farms too small to justify capital expense. Tractors only displaced horses around the 1950s when product, skills and farm sizes had all restructured.
  → Concrete historical case for the implicit premise that general-purpose technologies need decades of structural change before delivering full impact.
• Kyle Kingsbury's car analogy: a long series on how long it took cars to fully transform US cities into car-preferred infrastructure — decades from the first automobile shock to the fully restructured urban environment.
  → A second analogous case to back the tractor argument; also signals shared reading-stack with the hosts (Aphyr/Jepsen).
• Gukesh as the new chess world champion after Magnus Carlsen — reportedly the only young player who was not allowed a computer until age 12, and therefore had to learn chess differently before being accelerated by machine training.
  → Carries the argument that early-stage learners may need protected craft-acquisition before AI assistance — without Jonas having to defend the claim abstractly.
• Law firm policy: some law firms allegedly forbid junior associates from using AI for the first two years until they have basic training. Jonas has 'sympathy' for this approach but acknowledges EY has not chosen it.
  → Adds a parallel sector's experiment with shielded learning, leaving the question open rather than resolved.
• A team meeting that morning where Jonas pointed out that 'AI slop is a thing' and that personal responsibility for output quality is non-negotiable.
  → Grounds the learning-vs-production rule in a concrete recent leadership act.

Questions addressed

• ✅ How does a large organisation actually transform across paradigms?
  Through accumulated flywheel momentum that crosses many simultaneous S-curves, passing through a liminal phase of high friction where old structures must be released before new ones settle, and through the transition from opposition to position once the change is institutionalised.
• ✅ What lessons from the NAV transformation generalise?
  Capability built before the crisis is what makes the crisis survivable; crisis-time alignment is easy because the narrative is given; non-crisis leadership requires constructing that narrative every day; opposition-to-position transitions cannot be skipped.
• ✅ Is leadership a separate discipline from craft?
  No — partner-style work that combines owner/leader/practitioner is energising rather than confusing, provided you explicitly signal which mode you are in so colleagues do not mistake your opinions for decisions.
• ✅ How should we understand AI's place in software engineering right now?
  As an abstraction shift comparable to high-level languages, OS/DB, the internet and cloud, currently at the very start of its S-curve, in a liminal phase whose full impact will take decades of organisational restructuring to deliver.
• ⭕ How should organisations handle juniors in the AI era?
  Open question. Jonas has sympathy for shielded learning (chess, law firms) but admits EY has not chosen that path; the implicit answer is that pedagogy must be redesigned but the shape of the redesign is unclear.
• ✅ What makes someone able to 'take the elevator' across organisational floors?
  Security in your craft, humility about other perspectives, fluency with metaphor and model. Trainable, not innate.
• ✅ Why does a podcast on this topic exist at all?
  Truls says it openly: the hosts wanted to keep doing the work of translating between organisational floors together after their jobs no longer required it of them.
• ⭕ Can the conditions of crisis-time alignment be reproduced in normal times?
  Implicitly no — Truls notes that COVID-time alignment did not become a new normal; things 'became maybe a little better, no more'. The hosts and guest do not propose how to engineer alignment outside crisis.
• ⭕ Is Europe structurally positioned to participate meaningfully in the AI shift?
  Jonas names the gap and the sovereignty concern but offers no path to closing it; the conversation leaves this as an observation, not a plan.
• ⭕ How do you tell whether you are in a real S-curve or a hype curve that will collapse?
  Truls raises blockchain as the cautionary case; Jonas concedes 'it is not always you know when you are in it.' No reliable heuristic is offered.

Gaps

• incompleteThe flywheel/S-curve/liminal frame is used to describe both the NAV transformation (which succeeded) and the AI shift (in progress). The conversation never names a transformation that failed at the liminal phase and asks why — so the framework is at risk of being unfalsifiable on its own terms.
  Read next: Read failed transformations through the same frame — e.g., government-IT projects that died in their liminal phase, large agile adoptions that reverted. Cutler's overload work and the 'people who tried Spotify model' literature would expose where the frame breaks. The Geoffrey Moore 'Crossing the Chasm' tradition has more developed accounts of curves that stall.
• incompleteThe 'learning vs production' rule is stated as a personal heuristic but not connected to any educational theory or empirical evidence. For a leader of consultants asked to scale this practice, there is no model for how the two modes should be partitioned, whether across days, projects, or moods.
  Read next: Donald Schön's reflection-in-action; deliberate-practice literature (Ericsson); Beck's 'tidy first?' framing of preparatory vs delivery work — the tidy/messy split echoes the learning/production split structurally. Worth syntopical pass against Beck's recent 'Starving Genies' work on AI-assisted craft.
• incompleteThe elevator metaphor is generalised from Jonas's own career and from the show's animating premise. The conversation does not seriously consider organisations where vertical traversal is structurally suppressed (military, regulated finance with information walls, classified environments). The implicit premise that elevators are always available may not hold.
  Read next: Read about organisations with designed information barriers — security classifications, MAR-walled investment banking, healthcare compartmentalisation. James C Scott's 'Seeing Like a State' on legibility. Where the elevator is locked, what substitutes?
• incompleteThe introvert-leader claim rests on one practitioner's experience and one rhetorical contrast (scale removes small-talk tax). Jonas notes that some colleagues found it hard to distinguish his opinions from his decisions — a cost that may be specifically introvert-shaped (lower social-cue density) but is not investigated.
  Read next: Susan Cain on introvert leadership; the Cutler thread on 'signals that look like decisions' in leadership communication. The implicit problem of 'low-cue leadership at scale' is worth its own extraction.
• incompleteThe tractor and car analogies argue for slow structural change, but neither is examined for disanalogies. AI compresses learning loops in a way physical machinery cannot. The argument relies on the assumption that adoption is gated by humans, not by capability — which is plausible but not defended.
  Read next: The general-purpose-technology economics literature (Bresnahan & Trajtenberg); David Autor on labour-market restructuring time. Counter-cases: dial-up to broadband transitions, mobile phones — both moved faster than tractors. What predicts speed?
• uninformedThe Gukesh anecdote is reported confidently ('the only one of the young players not allowed a computer before age 12') but is not actually verified in the conversation. Whether true or apocryphal, it is doing argumentative work it has not earned.
  Read next: Quick fact-check; more importantly, treat the underlying claim about protected early craft as a hypothesis and look at the actual evidence on AI use in early-career learning — programming pedagogy research will move on this in 2026-2028.
• incompleteThe claim that NAV's success was driven by capability investments is asserted from inside the case by the very people who made those investments. There is no counterfactual — no comparable agency that did not invest similarly and failed similarly. Some of the 'flywheel' may be storytelling.
  Read next: Compare cross-country COVID welfare-IT responses (Germany, Denmark, UK), and Norwegian agencies of similar size that did not transform similarly. The Skatteetaten and Helsedirektoratet experiences would be a natural comparison.
• illogicalJonas argues both that we are early on the AI S-curve (so 10% gains are not the story) and that complexity will push back on productivity gains as they accumulate. These are compatible but in tension: if the curve climbs steeply because of intent-level abstraction, the complexity ceiling should appear later; if complexity bites soon, the curve is shallower than claimed. The argument is not adjudicated.
  Read next: The Brooks / Mythical Man Month tradition on accidental vs essential complexity; Ousterhout on system design; recent measurements on long-running agent harnesses (Anthropic harness design work) which directly test the complexity-ceiling claim.

Related documents

• 9) Nils Brede Moe om AI-paradokset, autonomi, læringskultur og fremtidens team
• 5) Vidar Moe: Det eneste Vidar vil er å ha det lett
• SMOC.AI's Post — 5 Minutter: AI og utviklere
• 14) Øystein Ulseth pendler mellom å lede og levere
  Organizational Abstraction Levels Sociotechnical Architecture Technology Leadership (Heis)
• Nils Brede Moe's Post
  Learning Culture & Slack Technology Diffusion & Transformation Curves
• Managers Are Struggling to Keep Up with the AI Productivity Boom
  AI Slop Organizational Abstraction Levels
• Are We Entering a New Age of Creativity with the Help of AI?
  Learning Culture & Slack Technology Diffusion & Transformation Curves
• 1) Torbjørn Larsen: Det nærmeste vi kommer Steve Jobs
  Sociotechnical Architecture Technology Leadership (Heis)
• 2) Hans Christian Holte dusjer i informasjon
  Public-Sector Digitalization Technology Leadership (Heis)
• 6) Catherine Janson og Kjerstin Wester: The power of two!
  Public-Sector Digitalization Technology Leadership (Heis)