Systems and Usage Modes

This chapter is the register of the AI estate of Nordvik Region: every system and use case that the fourteen volumes give the Region, with its usage mode (Mode 0 Ambient AI to Mode 5 Autonomous production agents), risk class, owner, provider, decision record and the volume in which it first appears. It then sorts the same estate by usage mode, follows how the estate changes from AI Governance (Book 2) to Vibe to Dev (Book 13), and lists what the estate deliberately does not contain. The chapter draws on every volume from Book 2 onwards. Architecture as Code (Book 1) has no example organisations and supplies foundations only.

Two kinds of fact appear. The profile of the Region as it stands in November 2026 (the benefits-triage system, the staff assistant, the citizen-service assistant, the correspondence classifier, the case-note summariser, the Mode 3 tools of the platform team) is established. Everything dated 2027 or later is illustrative fiction written in November 2026, and each volume labels it so. Where a row below carries a later date it is illustrative, and the table says which volume holds it.

The estate as one register

The register has one row per system or use case. "Owner" is the role the volume names; the volumes use slightly different titles for the same post, a point returned to at the end of the chapter. "Records" lists the decision records and registers that the volume attaches to the system. The first-appearance column names the volume in which the system is first established, not the first volume that mentions it.

System or use case Mode Risk class Owner Provider and hosting Records First in
Benefits-triage system (claims ranker; ranks applications for financial assistance, about 40,000 a month; a caseworker decides every case) 4 High-risk, Annex III point 5(a); Region is deployer Head of Benefits Unit TriageSoft, tender NR-2026-117; runs in the Region's regional data centre; TriageSoft retrains quarterly inside the Region's platform AIDR-0007 (v3.2.0), AIDR-0011 (oversight design), AIDR-0018 (v4.0.0 without postcode features); INV-0042; FRIA-2026-002; EUDB-DEP-2026-004213; NR-ALG-2026-003 Book 2
Staff assistant (about 2,400, then about 3,200 users) 4 Limited risk Chief Digital Officer (Books 2, 3); Head of IT and Digitalisation (Books 10, 11) Open-weight model at a European cloud provider, EEA only, customer-held keys, contract NR-2026-031 AIDR-0211; UC-2026-061; UCD-2026-022; NR-ALG-2026-007; SVC-NR-006 Book 2
Citizen-service assistant 4 Transparency duty, Article 50 disclosure Head of communications (Book 13 profile of the Landscape model); head of citizen service (Book 5) Open-weight models on accelerators reserved at a European hosting provider in a Nordic data centre; prompt citizen-service/v7 AIDR-0014; SVC-NR-005 Book 2
Correspondence classifier (fine-tuned open-weight 7B) and correspondence summarisation 4 Not stated Head of the registry Region-hosted; ML-BOM v2.1.0 with SLSA provenance AIDR-0042; UC-2026-017; LM-2026-009 Book 2
Case-note summariser (180 social-services caseworkers; special categories of data) 4 Not stated Head of social services On premises, two nodes in separate data centres, all egress denied Usage-mode register usage-modes/nordvik-region.yaml Book 2
Living-support triage (public register entry for the claims ranker) 4 High-risk Director of Social Services as senior responsible owner See Chapter 27 on whether this is the same system as the benefits-triage system NR-ALG-2026-003; AIDR-0188 (threshold at least 0.70); FRIA summary v2 Book 2
Travel-allowance scorer (retired) 4 Not stated Head of Payroll Services Not stated AIDR-0203; NR-ALG-2022-011 replaced by NR-ALG-2026-009, planned retirement 2026-11-30 Book 2
Ambient AI in licensed software (NR-AMB) 0 Not classified None until 2027-06-30; then AI governance lead (illustrative) Software vendors, non-EU, various Landscape model; supplier notice clauses requested Book 2, Landscape
Conversational assistants (NR-CA; approved enterprise tool, consumer endpoints blocked in the proxy) 1 Not stated Regional director signs the acceptable-use policy Assistant vendor, non-EU Acceptable-use policy; NR-CA Book 4, Landscape
Embedded workplace assistant pilot (NR-EWA) 2 Class accept-and-monitor (Book 5) Head of digitalisation Model chosen by the suite vendor, non-EU jurisdiction Pilot; reach small, then medium on 2027-09-01 (illustrative) Book 2, Book 5
Platform-team code agent (NR-UC-017) 3 Not stated Head of Platform Region-hosted open-weight model Usage-mode register; sandbox_fail_closed: missing (open gap) Book 2, Book 3
Agentic work tools of the platform team and the security architect (NR-AWT) 3 Not stated CISO as mandate owner Open-weight on reserved capacity, Nordic hosting MND-NR-01 (AIDR-0511, AIDR-0503); MND-NR-DEV-01 (AIDR-1108) Book 6
SOC triage assistant (NR-SEC-001) 4 Not stated CISO; a named person in the security unit Self-hosted on reserved capacity AIDR-0301; MND-NR-SOC-01 (AIDR-0714); svc-nr-sec-001 Book 4
Exposure prioritisation (NR-SEC-003) 4 Not stated CISO Self-hosted Level 0 Advisory Book 4
AI red teaming of the assistants (NR-SEC-002) Not stated Not stated Head of AI platform Self-hosted, open-weight only AIDR-0302; RoE-NR-2027-003 Book 4
Restore-drill and chaos programme (NR-SEC-004) Not stated Not stated Security unit Planner at Level 0 in test Programme record Book 4
Rostering and route system ShiftWise 4 High-risk, Annex III point 4(b); Region is deployer only HR director; each unit manager approves every roster ShiftWise, tender NR-2025-089, EU supplier FRIA-2026-007; DLG-NR-ROS-01 (AIDR-0602); DRF-NR-01 (AIDR-0607) Book 7
Recruitment screening pilot (applicant tracking module) 4 High-risk, Annex III point 4(a) HR director Supplier non-EU in the Landscape model AIDR-0601 (paused 2026-09-18); retired 2027-06-30 in the Landscape model (illustrative) Book 7
Home-care dashboards (planned against actual time per visit) Not stated Treated as high-risk by decision AIDR-0609 (illustrative), which makes the Region a provider Head of home care Built in house DLG-NR-HCD-01 (Level 0); AIDR-0609 Book 7
Meeting-summary and action-item assistants 4 (staff assistant) and 2 (pilot) See above Not stated See above DLG-NR-SA-01 (Level 0); AIDR-0620 Book 7
Hunting-query pack and detection-drafting assistant 4 Not stated Detection owner Self-hosted Level 1; AUS-NR-SOC-01 Book 8
Operations assistant NR-OPS-001 4 Not stated Head of platform Region Cloud AIDR-0901; Level 0 Advisory Book 10
Scribe NR-OPS-002 Not stated Not stated Service desk lead Region Cloud Level 1 Assisted Book 10
Runbook runner NR-OPS-003 4 Not stated Head of platform Region Cloud MND-NR-OPS-003 (AIDR-0916); Level 2 Supervised Book 10
Council-paper summariser 4 Not stated Head of IT and Digitalisation Region MND-NR-LDR-01; Level 1 Assisted Book 11
IDE assistant pilot NR-DEV-001 (12 developers) 2 Not stated Head of IT and Digitalisation Self-hosted open-weight code model DLG-NR-DEV-01 (AIDR-1101); AIDR-1102 Book 12
Builder workspace (editor assistant and delegated tasks in a sandbox) 2 and 3 Build is not high-risk Platform lead Region platform AIDR-1200; MND-NR-BLD-2031-01 (AIDR-1201); LANE-NR-2031-01 Book 13
service-point-check (the build; reads three sources, writes a weekly report) Calls no model at run time Tier B, then Tier C; never Tier D Builder (product owner) Region Git service and scheduled-job template BLD-NR-2031-0001; AIDR-1203 Book 13

Table 5.1: The AI estate of Nordvik Region as the series gives it. Rows dated 2027 or later are illustrative. A mode of "not stated" means that the volume does not give one. (2, Nordqvist, 2026) (3, Nordqvist, 2026) (4, Nordqvist, 2026) (6, Nordqvist, 2026) (7, Nordqvist, 2026) (8, Nordqvist, 2026) (10, Nordqvist, 2026) (11, Nordqvist, 2026) (12, Nordqvist, 2026) (13, Nordqvist, 2026) (15, Nordqvist, 2026)

The first inventory of the Region, reported in AI Governance, Chapter 6, found eleven AI services in use and three of them without review. The origin and hosting policy was adopted in 2026 and the approval matrix is published with the algorithm register (2, Nordqvist, 2026). The Landscape model of the Region, which is a scenario for the year 2027, counts eleven units on 1 January 2027 (the model file holds thirteen entries, two of them planned and not counted on that day) and thirteen on 31 December 2027, the last figure including the retired recruitment pilot (15, Nordqvist, 2026).

The AI estate of Nordvik Region sorted by usage mode

Figure 5.1: The estate by usage mode. Mode 4 holds most of the systems; Mode 5 is empty by policy.

The estate in September 2032: footprint view

AI Environment adds no system to the register. It counts six uses of the existing estate, gives each a footprint record FPR-NR-000n and a benefit record BEN-NR-000n, and states the volumes of September 2032 (all illustrative). Where a use is run is part of the picture: the Region owns no accelerators and buys capacity, with 24 reserved accelerators and up to 8 on demand at the hosting provider under NR-2026-031 (region eu-north-1; the capacity plan of 2027 had 16 and 8), an average pool utilisation of 0.58 over the third quarter of 2032, a hosted large-provider model reached through the gateway for one route, and the Region Cloud in nordvik-dc (about 640 kW installed IT power, PUE about 1.55, no waste-heat reuse at baseline) (14, Nordqvist, 2026).

Use Service and records Mode Where it runs Volume, September 2032 Use owner Footprint known at baseline
FPR-NR-0001 Citizen-service assistant, open-weight route SVC-NR-005; AIDR-0014 4 Reserved accelerators; embeddings in nordvik-dc About 146,000 resolved enquiries a month (71,000 in October 2026, a factor of 2.4 for both routes together) Sara Ekdahl The aged SCI figure of October 2026
FPR-NR-0002 Citizen-service assistant, eligibility-explanation route SVC-NR-005 4 Hosted EU model through the gateway; site, grid and energy per call not disclosed About 22,000 explanations a month (13 per cent of enquiries) Sara Ekdahl None
FPR-NR-0003 Staff assistant SVC-NR-006; AIDR-0211 4 Open-weight model on the reserved accelerators About 410,000 requests a month; about 3,200 users Gunilla Roos None
FPR-NR-0004 Benefits-triage ranking SVC-NR-001; AIDR-0007, AIDR-0011, AIDR-0018 4 TriageSoft in the Region Cloud, quarterly retraining on the Region's platform About 40,000 claims ranked a month; four retrainings a year Head of the Benefits Unit (unnamed) None; the AI share of nordvik-dc is not metered
FPR-NR-0005 Correspondence classifier and case-note summaries AIDR-0042; UC-2026-017 4 On premises, a 110-million-parameter encoder About 120,000 items routed and about 9,000 summaries a month Maja Lindqvist None until measured (October 2032)
FPR-NR-0006 Builder workspace assistant AIDR-1200 2 and 3 Platform team's workspace About 40 builders; about 9,000 sessions a month Maja Lindqvist None

Table 5.1a: The six uses that AI Environment counts. Ambient AI and the embedded-assistant pilot (Modes 0 and 2, SVC-NR-008) are not itemised; they enter the environmental register as supplier-reported rows. (14, Nordqvist, 2026)

The register of the uses is itself a record of the Region: register/environment-nr.yaml, adopted by AIDR-1300 on 2032-09-20, one row per use with its owner, footprint record, benefit record, budget and evidence age (14, Nordqvist, 2026).

The estate by usage mode

The usage modes are the series' own scale, set out in AI Governance, Chapter 2. The Region's usage-mode register (usage-modes/nordvik-region.yaml, version 2026.11.1, reviewed every 90 days) places three systems in Mode 4, limits Mode 3 to the small platform team and lists Mode 5 under modes_not_permitted (2, Nordqvist, 2026). The table sorts the estate by mode and gives the date or figure where a volume does.

Mode What the Region has Where the volumes place it Control that bounds it
Mode 0 Ambient AI AI features in licensed software (NR-AMB, since 2024-01-01 in the Landscape model, Level 4 in the model); the phishing-report classifier in the email gateway; the vendor AI of the managed security provider that closes low-severity alerts out of hours; ambient classification in the service-management tool, switched on only for the service desk; ambient completion in the development environment, on by default after a vendor update in March 2030; computer-aided detection in a gastroenterology department, switched on by a vendor update in 2028 and found in 2029 Books 3, 4, 6, 8, 10, 11, 12 Declarations and notice clauses (NR-SUP-SECAI-001, annex MS-2); weekly sample of closures; an owner for NR-AMB from 2027-06-30 (illustrative)
Mode 1 Conversational Approved enterprise assistant, public material only for the security architect; consumer endpoints blocked in the proxy; a self-hosted assistant for alert data beside NR-SEC-001; prototyping by a product owner (Book 12, Book 13) Books 4, 6, 8, 12, 13 Acceptable-use policy signed by the regional director; data standard STD-AI-002 v1.3
Mode 2 Embedded Pilot of an embedded workplace assistant (extended in 2029 to the 18 members of the regional management group and their assistants); IDE pilot NR-DEV-001; the builder workspace editor assistant Books 2, 5, 11, 12, 13 Tenant policies and audit logs (from 2027-09-01, illustrative); usage profile valid from 2030-10-07
Mode 3 Agentic tools Code agent of the platform team (NR-UC-017); work tools under MND-NR-01; four of seven platform engineers hold the permission; delegated tasks in the builder sandbox Books 2, 3, 6, 10, 12, 13 Mandate manifests; protected paths; no commit or push; sandbox without production access
Mode 4 AI applications Benefits triage, staff assistant, citizen-service assistant, correspondence classifier, case-note summariser, SOC triage assistant, exposure prioritisation, ShiftWise, operations assistant and runbook runner, council-paper summariser Books 2 to 11 AI gateway; AI Decision Records; FRIA where high-risk; delegation profiles
Mode 5 Autonomous production agents None Every volume Rule: no autonomous production agents; Mode 5 not started without containment, a stop mandate, evaluation evidence and a committee decision (AI Governance, Chapter 23)

Table 5.2: The estate by usage mode. (2, Nordqvist, 2026) (3, Nordqvist, 2026) (4, Nordqvist, 2026) (6, Nordqvist, 2026) (8, Nordqvist, 2026) (10, Nordqvist, 2026) (11, Nordqvist, 2026) (12, Nordqvist, 2026) (13, Nordqvist, 2026) (15, Nordqvist, 2026)

Across the usage modes

The mode distribution on 1 January 2027 in the Landscape model is one unit in Mode 0, one in Mode 1, one in Mode 2, one in Mode 3 and seven in Mode 4. On 31 December 2027 Mode 4 holds nine marks, the retired recruitment pilot among them, and there is still no Mode 5 mark (15, Nordqvist, 2026). The weight of the estate is therefore in Mode 4, where the Region buys or hosts an application and a person is accountable for what it does. The two modes that the Region understands least are the outer ones. Mode 0 grows without a decision, which is why AI Red Team found that the vendor's Mode 0 functions in the endpoint and email products acted at what amounted to Level 3 without anyone having decided it (4, Nordqvist, 2026). Mode 3 is narrow because the platform team has seven engineers and the Region has no agent operator.

Degrees of delegation across the estate

The series' autonomy levels (Level 0 Advisory to Level 4 Autonomous) are set per system and action class. The ceilings the volumes record for the Region are collected below. A ceiling is not a licence: AI Developer and AI Leader both say that the practice before the ceiling was lower than the ceiling.

System or class Ceiling or level Source
Benefits triage, rank the queue Level 0 in the Landscape model; a caseworker decides every case Books 2, 5 and Landscape
Security AI generally Programme charter NR-SEC-PROG-2027 (approved 2027-01-18): maximum usage mode 4, offensive work maximum Level 2, irreversible actions Level 1, security data never leaves the EU and EEA Book 4
NR-SEC-001 Level 1 Assisted, read-only service identity Book 4
NR-SEC-003 Level 0 Advisory Book 4
ShiftWise rosters and routes Level 2 Supervised; offering open shifts to staff who opted in at Level 3 Bounded from 2027-03-11 (illustrative); management consequence class capped at Level 1 Assisted by DRF-NR-01 Book 7
Executive-level use (DLG-NR-EXE-01, AIDR-1002, 2029-10-01) Level 1 Assisted in every class; the people class is not delegated Book 11
Operations (DLG-NR-OPS-001) Level 2 on every class that changes production; assistant Level 0; runbook runner Level 2; rollback and failover Level 1 Book 10
Development (DLG-NR-DEV-01, AIDR-1101, 2030-09-23) Specify 0, design 0, generate 1, refactor 1, test 1, review 0, deploy 0, operate 0, document 2, migrate 0 Book 12
Builder assistant (permit-by-tier.yaml, 2031-12-17) Ceiling Level 3 Bounded, never Level 4 Book 13
Environmental budgets (DLG-NR-ENV-0001) and the platform team's Mode 3 tool (MND-NR-ENV-0001, AIDR-1315, 2032-11-02, proposed) Alert and defer automatically; stop automatically at 1.10 of the monthly quota; per session at most 80 model calls, 150 tool calls and 45 minutes, and 800 sessions a month; the level number is not given Book 14

Table 5.3: Ceilings on delegation across the estate. All rows from 2027 onwards are illustrative. (4, Nordqvist, 2026) (7, Nordqvist, 2026) (10, Nordqvist, 2026) (11, Nordqvist, 2026) (12, Nordqvist, 2026) (13, Nordqvist, 2026)

The ceilings do not agree everywhere. The Landscape model gives the agentic work tools a ceiling that falls from Level 2 to Level 1 on 2027-05-01, when the architect's mandate is aligned with the role charter. AI Developer later places the platform team's agentic drafting at Level 2 from AIDR-1116 (2030-11-18) while the generate ceiling in DLG-NR-DEV-01 is Level 1, and Vibe to Dev lets the builder assistant reach Level 3 Bounded while AI Developer keeps Mode 3 to the platform team. These are listed as open threads in Chapter 27.

How the estate changes from Book 2 to Book 14

The estate grows by layers, and each volume adds the layer that its subject needs.

Volume Years What is added to the estate
Book 2, AI Governance 2026 The core: benefits triage, staff assistant, citizen-service assistant, correspondence classifier, case-note summariser, Mode 3 code agent; the usage-mode register; the first inventory (eleven services, three unreviewed)
Book 3, AI Platform and Security 2026 The reduced reference platform: gateway, registry, evaluation suites, control plane; overlay architecture/overlays/nordvik-region.yaml on enterprise-ai-platform@2.0.0 (AIDR-0002); reserved accelerators at the hosting provider
Book 4, AI Red Team 2027 Security AI: NR-SEC-001 to NR-SEC-004; programme charter; no AI-enabled red team of its own
Book 5, AI Threat 2026 to 2027 No new system. A register of nine threats and the first removal of a feature (postcode) as the reference case of a harm path closed before harm
Book 6, The Transformed Security Architect 2026 to 2027 Mandate MND-NR-01 for the architect's Mode 3 tools
Book 7, AI Boss 2026 to 2027 ShiftWise, the home-care dashboards, assistants that propose action owners; the recruitment pilot paused
Book 8, AI Blue 2027 Hunting-query pack, detection-drafting assistant, MND-NR-SOC-01; vendor AI at the managed security provider made visible
Book 9, AI Purple 2027 to 2028 A bought loop; a correlator at Level 1 and a critic at Level 0 under MND-NR-EMU-01 and -02; no emulation agent
Book 10, AI Operations 2028 to 2029 NR-OPS-001 to NR-OPS-003; the service catalogue as the unit of operation
Book 11, AI Leader 2029 to 2030 The council-paper summariser; the embedded assistant pilot reaches the management group
Book 12, AI Developer 2030 to 2031 NR-DEV-001; the development assistants tender NR-2030-041 (outcome not given)
Book 13, Vibe to Dev 2031 The builder workspace; service-point-check
Book 14, AI Environment 2032 to 2033 No new system; a footprint and benefit record for each of six uses, the environmental register, environmental budgets with a stop path, a placement policy, and a Mode 3 mandate MND-NR-ENV-0001 for the platform team's tool (proposed)

Table 5.4: The growth of the estate by volume. (2, Nordqvist, 2026) (3, Nordqvist, 2026) (4, Nordqvist, 2026) (5, Nordqvist, 2026) (6, Nordqvist, 2026) (7, Nordqvist, 2026) (8, Nordqvist, 2026) (9, Nordqvist, 2026) (10, Nordqvist, 2026) (11, Nordqvist, 2026) (12, Nordqvist, 2026) (13, Nordqvist, 2026) (14, Nordqvist, 2026)

Two features of the change stand out. First, three systems carry the Region's whole high-risk exposure: the benefits-triage system (Annex III point 5(a)), the rostering system (point 4(b)) and, until it was paused, the recruitment pilot (point 4(a)). The AI Act duties of a deployer under Articles 26 and 27 apply to the first two from 2027-12-02, and AI Boss adds that the Region carries provider duties for the home-care dashboards if it keeps them as high-risk (7, Nordqvist, 2026). Second, the most-changed record is the benefits-triage model itself: version 3.2.0 under AIDR-0007, replaced by version 4.0.0 under AIDR-0018 after a parity drift of 0.06 lasted 61 days against a retirement criterion of 0.05 for 60 days. AI Platform and Security dates AIDR-0007 as proposed 2026-03-02 and accepted 2026-04-15, with the model 4.0.0 in shadow on 2027-01-29 and promoted on 2027-02-10 (illustrative) (3, Nordqvist, 2026).

The estate shrinks in two places. The recruitment pilot is paused by AIDR-0601 on 2026-09-18, after a union for care staff asked how the ranking was made and the supplier could not show it, and is retired on 2027-06-30 in the Landscape model without being resumed (7, Nordqvist, 2026) (15, Nordqvist, 2026). The travel-allowance scorer NR-ALG-2022-011 is replaced by NR-ALG-2026-009 with planned retirement on 2026-11-30 (2, Nordqvist, 2026).

What the estate deliberately does not contain

The volumes agree on a list of things that the Region has chosen not to have. Each is a decision or a rule and is recorded as such.

  • No Mode 5 agent. The policy lists Mode 5 as not permitted. AI Platform and Security and AI Red Team state the conditions for a first one: containment, an independent stop path and evaluation evidence (3, Nordqvist, 2026) (4, Nordqvist, 2026). AI Red Team names the earliest candidate: automatic session revocation after confirmed credential theft, with a decision recorded whatever the outcome.
  • No agent team for the security architect. AIDR-0503 (2026-11-04) defers it: the Region would need a second reviewer, and capacity, not technology, decides. The triggers for reopening are a second security architect post, agent operators in the platform team, or a Mode 5 agent in the environment (6, Nordqvist, 2026).
  • No emulation agent. AI Purple has no agent that emulates an adversary, no continuous emulation pipeline and no scenario-count target. The contracted tester declares AI use in Modes 1 to 3 and uses no Mode 5 emulation agent without a signed mandate (9, Nordqvist, 2026).
  • No consequence-class AI about people. AI Boss leaves the M3 Consequence function empty. The recruitment screening is paused, no system is delegated for pay review, and DRF-NR-01 caps the class at Level 1 Assisted. AIDR-0610 forbids line managers from directing agents until there is oversight capacity. Shift swapping and the agent-to-agent negotiation component of the rostering system are switched off (AIDR-0621) (7, Nordqvist, 2026).
  • No AI in the benefits decision. A caseworker decides every case. The triage rank is not a basis or a reason for a decision (AIDR-1018, February 2030, illustrative), and the proposed change to the delegation order is that no automated decision function exists (the Local Government Act, Chapter 6, section 37) (11, Nordqvist, 2026).
  • No Level 4 for the delegated classes. No volume gives any consequence-bearing class Level 4. The Landscape model marks the citizen-service assistant and ambient AI at Level 4 for particular action classes (answering a resident; vendor features), which is a point where the model and AI Red Team differ on the label (15, Nordqvist, 2026).
  • No sensitive data in assistants without approval. The data standard STD-AI-002 v1.3 admits public data in any approved tool, internal data only in the Region's tenant (nordvik-eu), confidential data only in the staff assistant with pseudonymised logs, and sensitive data in no system without a data protection impact assessment and a system-specific approval (2, Nordqvist, 2026).

Open threads in the register

The following are carried to Chapter 27. The working value in this chapter is stated for each.

  1. Same system or two? AI Governance describes living-support triage as a system of its own (lsb-triage v2.3.0, NR-ALG-2026-003, a FRIA summary dated 2026-02-11) while the profile says the same claims ranker orders the living-support queue. The FRIA in Chapter 10 of that volume is FRIA-2026-002, completed 2026-06-20. The dossier lists them as one system with two records.
  2. EU database identifier. EUDB-DEP-2026-004213 (Chapter 9) against EUDB-2026-NR-00412 (Chapter 17), both in AI Governance.
  3. Model version. The inventory gives 3.2.1 while AIDR-0007 gives 3.2.0.
  4. AI Act role of the staff assistant. NR-UC-020 records provider; the use-case descriptor UCD-2026-022 records deployer.
  5. Unit count on 2027-01-01. The Landscape model file holds thirteen entries, two of them planned, and the picture of the day has eleven marks. Landscape has corrected its Chapter 5 from thirteen to eleven, so its chapters now agree; what remains is a nuance, that its register of the estate lists thirteen entries and that thirteen marks are counted on 2027-12-31 including the retired pilot (Chapter 27).
  6. Titles of the owner of the staff assistant. Chief Digital Officer in Books 2 and 3; Head of IT and Digitalisation in Books 10 and 11.
  7. Embedded workplace assistant pilot. The vendor's model sits under a non-EU jurisdiction (Book 5, class accept-and-monitor) and the profile places the pilot in Mode 2.
  8. Six uses, two numberings. AI Environment numbers its six uses FPR-NR-0001 to FPR-NR-0006 and reserves FPR-NR-0007 first as the next use and then, in its Chapter 5, as the model footprint sheet requested from the hosted-model supplier; the platform team's Mode 3 tool is FPR-NR-0008. Chapter 27.
  9. Plan for Mode 3 code agent. NR-UC-017 has sandbox_fail_closed: missing, an open gap in AI Platform and Security; no later volume records its closure.

Summary

The estate of Nordvik Region is a Mode 4 estate with narrow edges. It has one system that carries most of the legal weight (the benefits-triage system), a small set of assistants, security and operations tools capped at Levels 0 to 2, a rostering system in the management function, and a builder workspace at the end of the series. It has no Mode 5 agent and chooses not to have one. The register shows that the Region's estate grows by decision record: nearly every system carries an AIDR, a mandate or a delegation profile, and most of the estate's risk sits in the few systems that the Region buys rather than builds. The weakest places are the outer modes, where the Region has the least control: ambient AI in licensed software, where an owner was only named in 2027 in the model, and agentic tools, where the mandate can be tighter than the practice or the reverse. The register also shows that the volumes use different ceilings and titles for the same things, and that the dossier cannot resolve them without the author.

Sources

  1. Nordqvist, G. AI Governance: Governing Artificial Intelligence in European Enterprises and Public Authorities. (2, Nordqvist, 2026)
  2. Nordqvist, G. AI Platform and Security: Building and Securing AI Platforms in European Enterprises and Public Authorities. (3, Nordqvist, 2026)
  3. Nordqvist, G. AI Red Team: AI as an Active Part of Cyber Defence in European Enterprises and Public Authorities. (4, Nordqvist, 2026)
  4. Nordqvist, G. AI Threat: Are the Dystopias Coming True? (5, Nordqvist, 2026)
  5. Nordqvist, G. The Transformed Security Architect. (6, Nordqvist, 2026)
  6. Nordqvist, G. AI Boss: When Artificial Intelligence Manages People. (7, Nordqvist, 2026)
  7. Nordqvist, G. AI Blue: How Artificial Intelligence Transforms the Blue Team. (8, Nordqvist, 2026)
  8. Nordqvist, G. AI Purple: How Red and Blue Learn Together When AI Joins the Loop. (9, Nordqvist, 2026)
  9. Nordqvist, G. AI Operations: How Artificial Intelligence Transforms IT Operations. (10, Nordqvist, 2026)
  10. Nordqvist, G. AI Leader: How Leadership Roles Change When AI Is Used Broadly. (11, Nordqvist, 2026)
  11. Nordqvist, G. AI Developer: How the Developer Role Changes When AI Is Used Broadly. (12, Nordqvist, 2026)
  12. Nordqvist, G. Vibe to Dev: From Vibe Coding to Professional Software, Step by Step. (13, Nordqvist, 2026)
  13. Nordqvist, G. AI Environment: Designing AI for the Greatest Benefit and the Least Environmental Harm in European Enterprises and Public Authorities. (14, Nordqvist, 2026)
  14. Nordqvist, G. Landscape: A Fact-Based Picture of Architecture and AI at Work. (15, Nordqvist, 2026)