Examples

Decks and diagrams AI agents built in CoPage.

Every document on this page is a real CoPage document, built by an AI agent through CoPage’s MCP server or the free CoPage CLI, checked, and drawn here by CoPage’s own renderer. The consulting firm, Ostrell Vane, and its clients are fictional.

Client decks

Haldenmark 2030: AI value programme · AI programme proposal

A 24-month AI adoption programme for an industrial group: value case, use-case portfolio, operating model, AWS platform, plan and fees. 14 slides, with 2 linked architecture diagrams.

OSTRELL VANE

PROPOSAL · GROUP EXECUTIVE COMMITTEE · HAMBURG, 14 OCTOBER 2026

Haldenmark 2030: putting AI value into the P&L

An enterprise AI adoption programme for the Haldenmark 2030 margin ambition

Illustrative engagement. Ostrell Vane and Haldenmark are fictional; client figures are invented, market data is real and cited.

€95m

run-rate EBIT benefit from AI by FY2029,
before run cost

Strictly confidential · Prepared for Haldenmark Group

1
SUMMARY · UNDERSTANDING · POINT OF VIEW · VALUE · APPROACH · TEAM & TERMS

A 24-month programme can add €95m run-rate EBIT
by FY2029, about half the margin gap.

Source: Haldenmark Group data FY2025 (illustrative); Ostrell Vane analysis; Council of the EU (29 Jun 2026), AI Act simplification

Haldenmark Group · AI Value Programme · Strictly confidential

Ostrell Vane | 2

PRELIMINARY

SITUATION

Haldenmark 2030 targets an 11.5% EBIT margin by FY2029: +290bp, about €186m at today's €6.4bn revenue. The board expects AI to contribute.

COMPLICATION

126 GenAI pilots since 2023, 7 in production; 34% of 9,000 chat seats used weekly; no AI benefit tracked. AI and cloud spend (€11.8m) grows 2.6× a year, 40% without an owner. 28% of technicians retire within 8 years. Under the EU AI Act, AI literacy already applies to all 23,800 staff, and AI-generated content must be marked by 2 Dec 2026.

RESOLUTION

Focus on 10 use cases in three value levers

€95m run-rate EBIT by FY2029 (about 150bp); €85m after €10m run cost

Go deep in three lighthouse domains

Service (COO Service), quote-to-order (CCO) and procurement (CPO) hold €74m; each owner redesigns the workflow before build

One AWS platform and gateway

Every model call tagged to an owner: owned AI spend 60% → 100%; agents governed by autonomy level

Adoption, not seats

AI Value Office and AI Academy lift weekly active users from 34% to 70%

Self-funding from FY2028

€64m investment and run cost, €170m benefits to FY2029, net €106m

Decisions by 6 Nov 2026: the Group Executive Committee approves Phase 1 (10 weeks, €2.4m fixed); the CEO names a full-time programme director and chairs the AI Council; COO Service, CCO and CPO each own one lighthouse.

2
SUMMARY · UNDERSTANDING · POINT OF VIEW · VALUE · APPROACH · TEAM & TERMS

126 GenAI pilots since 2023 put 7 into production; none tracks a benefit.

Haldenmark Group, FY2025

Source: Haldenmark annual report FY2025; management interviews Sep 2026 (illustrative); Ostrell Vane analysis

Haldenmark Group · AI Value Programme · Strictly confidential

Ostrell Vane | 3

€6.4bn

revenue; 8.6% EBIT margin, target 11.5% by FY2029

126 → 7

GenAI pilots since 2023 → in production; no benefit tracked

34%

weekly active users of 9,000 chat seats

€11.8m

AI & cloud spend FY2025, growing 2.6× a year; 40% without an owner

28%

of 4,100 technicians retire within 8 years

9

data silos: three SAP ERPs and six field‑service systems; no installed‑base view

What we heard in 14 leadership interviews

Eleven business units choose pilots bottom-up; none has a business case or exit criteria.

Technicians search 1.2m documents in nine languages; senior experts are the real knowledge base.

IT holds the AI budget; no business owner signs off benefits.

SO WHAT

The constraint is focus, ownership and data, not technology.

3
SUMMARY · UNDERSTANDING · POINT OF VIEW · VALUE · APPROACH · TEAM & TERMS

Data and adoption, not models, hold Haldenmark back: both sit at level 1 of 5.

AI readiness by capability, level 1–5; target FY2028

Source: Ostrell Vane pre-proposal readiness assessment, 14 interviews, Sep 2026 (illustrative); Council of the EU (29 Jun 2026), AI Act simplification

Haldenmark Group · AI Value Programme · Strictly confidential

Ostrell Vane | 4

PRELIMINARY

Capability

1

2

3

4

5

What we found

What it takes

Value tracking & portfolio

No benefit ledger; 126 pilots without stage gates

AI Value Office; ledger verified by Finance
Owner: finance value controller

Data foundation

Three SAP ERPs, six field-service systems, no installed-base view

Installed-base data product; knowledge index
Owner: head of data

AI platform & LLMOps

AWS landing zone and chat assistant live; no gateway, no evals

AI gateway, evals, agent runtime
Owner: platform owner

Talent & AI literacy

About 40 data scientists; no literacy programme, though the AI Act requires it for all 23,800 staff

AI Academy: literacy for all, 600 champions
Owner: HR and academy lead

Responsible AI & compliance

No AI inventory; marking of AI-generated content due 2 Dec 2026 (Art. 50)

Inventory, autonomy levels, EU AI Act controls
Owner: Haldenmark Legal

Adoption & workflow redesign

34% weekly active; tools added to unchanged work

Redesign before build; adoption KPIs per squad
Owner: programme director

Today

To close

Target FY2028

Beyond target

SO WHAT

Fix data and adoption first. Both must climb three levels by FY2028, the widest gaps of the six; the platform needs one.

4
SUMMARY · UNDERSTANDING · POINT OF VIEW · VALUE · APPROACH · TEAM & TERMS

Only 37% of respondents see EBIT impact from AI; three hypotheses explain Haldenmark’s gap.

Share of respondents, %, 2026

1. 5%+ of EBIT from AI plus significant value; n=1,719, fielded May–Jun 2026.
Source: McKinsey (Aug 2026), The state of AI in 2026: On the road to ROI; BCG (Sep 2025), The Widening AI Value Gap: Build for the Future 2025; Gartner (Jun 2025, May 2026)

Haldenmark Group · AI Value Programme · Strictly confidential

Ostrell Vane | 5

Use AI in 3+ functions

56%

Scale AI across the enterprise

44%

See any EBIT impact from AI

37%

AI high performers¹

6%

Use is not value: only 6% are high performers.

H1 Value sits in core operations, not chat seats

70% of AI’s potential value is in core functions such as sales and marketing, manufacturing, supply chain and pricing (BCG)

At Haldenmark: Service, quote-to-order and procurement hold €74m

H2 Redesign the workflow before adding AI

About 3 in 4 high performers fundamentally redesigned workflows, against about 1 in 4 of others (McKinsey)

At Haldenmark: No pilot changed a process

H3 Govern agents by autonomy level

Gartner expects over 40% of agentic AI projects to be cancelled by end-2027; it recommends governance by four autonomy levels

At Haldenmark: No agent inventory; one HR tool becomes high-risk

SO WHAT

5
SUMMARY · UNDERSTANDING · POINT OF VIEW · VALUE · APPROACH · TEAM & TERMS

€95m comes from three levers:
cost to serve €39m, commercial €38m and spend €18m.

Run-rate EBIT impact by FY2029, €m

Note: run-rate at full adoption, before run cost.
Source: Haldenmark Group data FY2025 (illustrative); Ostrell Vane value model

Haldenmark Group · AI Value Programme · Strictly confidential

Ostrell Vane | 6

INDICATIVE

Technician productivity +5%
Technician copilot, 25 min a day

€14m

Truck rolls −10%
Remote diagnostics and dispatch

€11m

First-time-fix 74% → 81%
Parts prediction

€7m

Shared-services cost −4%
Finance and IT service desk agents; owner: CFO

€7m

Cost to serve
€39m
Owner: COO Service

Tender effort −30%
Tender analysis agent

€10m

Price realisation +0.35pp
Configure-price-quote (CPQ) agent, on €4.1bn new equipment

€14m

Aftermarket revenue +2%
Installed-base offers, on €2.3bn

€14m

Commercial
€38m
Owner: CCO

Should-cost and spend analytics
0.3% of €3.6bn addressable spend

€11m

Contract compliance and renewals
0.2% of €3.6bn addressable spend

€7m

Third-party spend
€18m
Owner: CPO

€95m
run-rate EBIT
by FY2029 (≈150bp),
before €10m run cost

SO WHAT

No single bet. No driver exceeds €14m (15% of the €95m), so one late use case cannot sink the programme.

6

All 14 slide titles

OSTRELL VANE
A 24-month programme can add €95m run-rate EBIT by FY2029, about half the margin gap.
126 GenAI pilots since 2023 put 7 into production; none tracks a benefit.
Data and adoption, not models, hold Haldenmark back: both sit at level 1 of 5.
Only 37% of respondents see EBIT impact from AI; three hypotheses explain Haldenmark’s gap.
€95m comes from three levers: cost to serve €39m, commercial €38m and spend €18m.
Ten of twelve candidate use cases carry the €95m; three lighthouse domains hold €74m.
€64m of investment and run cost returns €170m by FY2029, a net €106m.
A 25-person AI Value Office tracks value; business owners run the squads.
Ten use cases run on one AWS platform, so every model call has an owner.
The technician copilot answers from 1.2m documents in nine languages and acts only with approval.
Three phases over 24 months: diagnose in 10 weeks, prove three lighthouses, then scale.
A joint team of 22; €3.2m of our €18.5m fees depends on verified EBIT.
Three decisions by 6 Nov 2026 let the diagnostic start on 16 Nov 2026.

Northmere Mutual: AI FinOps · AI FinOps diagnostic

An insurer’s AI spend tripled in nine months, a fifth of it unowned: root causes, savings levers, a multi-cloud control plane and a 12-week plan. 14 slides, with 2 linked architecture diagrams.

AI FinOps diagnostic · AI Steering Committee · Dublin, 21 October 2026

Northmere Mutual: every AI dollar attributable, governed and worth it

Findings from a four-week diagnostic and a proposed 12-week programme

Illustrative engagement. Ostrell Vane and Northmere Mutual are fictional; client figures are invented, market data is real and cited.

OSTRELL VANE

Strictly confidential · Prepared for Northmere Mutual Group

MONTHLY AI RUN COST, $K, JAN–SEP 2026

420

1,380

3.3× since January; 45% of it can go without stopping a use case

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

1
SUMMARY · DIAGNOSIS · ROOT CAUSES · SOLUTION · PLAN & CASE

At September volume, $621k a month (45%) can be cut; no use case stops.

Note: at September volume and full delivery; risk-adjusted, the run-rate is $914k a month.
Source: Northmere billing data Jan–Sep 2026 (illustrative); Ostrell Vane analysis

Northmere Mutual Group · AI FinOps diagnostic · Confidential

Ostrell Vane | 2

Situation

AI is in daily use: the copilot handles 1.1m conversations and the claims agent 85k claims a month; 1,200 engineers use coding agents. Monthly run cost grew from $420k in January to $1,380k in September.

Complication

FY2026 heads for $13.0m against a $9.0m budget (+44%); the budget runs out at the end of October. One dollar in five is unattributed, and nobody can say what a claim or a conversation costs.

Resolution
$621k a month (45%) is addressable at September volume across eight causes; new volume then needs a budget approved by the FinOps lead.
A control plane in front of three providers and the GPU cluster, run by the AI platform team: owner tags, budgets, caching, routing and one cost ledger in the FOCUS standard.
Unit costs fall 63–75%: a claim from $2.46 to $0.69, protected by quality gates.
12 weeks, $0.9m fixed fee: $5.6m a year risk-adjusted, payback in month 4.
21 Oct 2026

Decisions today: the Steering Committee mandates the gateway (untagged calls refused from 30 Nov) and approves the programme: 12 weeks, $0.9m fixed fee; the CFO owns FinOps and names its lead and the use-case owners by 28 Oct.

INDICATIVE
2
SUMMARY · DIAGNOSIS · ROOT CAUSES · SOLUTION · PLAN & CASE

Inference got 280× cheaper in 2022–24, yet firms’ GenAI spend tripled in 2025.

1. Price per million tokens for a model scoring at GPT-3.5 level on the MMLU (Massive Multitask Language Understanding) benchmark.
Source: Stanford HAI (Apr 2025), AI Index 2025; Menlo Ventures (Dec 2025); Anthropic (Jun 2025); FinOps Foundation, State of FinOps 2026 (Feb 2026, n=1,192); Tokenomics Foundation, State of Tokenomics (Sep 2026, n=472)

Northmere Mutual Group · AI FinOps diagnostic · Confidential

Ostrell Vane | 3

Inference price¹ Stanford HAI

280×

cheaper GPT-3.5-level inference, Nov 2022 → Oct 2024

Nov 2022

Oct 2024

Enterprise GenAI spend, $bn Menlo Ventures

11.5

37

2024

2025

SO WHAT

Volume beats price. Agents use about 4× the tokens of chat, multi-agent systems about 15× (Anthropic).

FinOps teams now own it. 98% of FinOps practitioners manage AI spend, up from 31% in 2024 (FinOps Foundation).

Value is the gap. 43% name proving value or ROI as a top AI cost challenge, the most cited; 12% have no owner for AI spend (State of Tokenomics).

For Northmere: cheaper tokens will not cap the bill; volume grows 16% a month (page 4).

$20.00

$0.07

3
SUMMARY · DIAGNOSIS · ROOT CAUSES · SOLUTION · PLAN & CASE

Run cost rose 3.3× since January; FY2026 heads for $13.0m, 44% over budget.

Monthly AI run cost, $k, 2026

Source: Northmere provider invoices and internal GPU (graphics processing unit) costs, Jan–Sep 2026 (illustrative); forecast at +16% a month, the Jan–Sep average

Northmere Mutual Group · AI FinOps diagnostic · Confidential

Ostrell Vane | 4

Budget
run-rate
$750k

Actual

Forecast at +16%/month

420

485

563

653

758

885

1,026

1,190

1,380

1,601

1,857

2,154

Jan

Feb

Mar

Apr

May

Jun

Jul

Aug

Sep

Oct

Nov

Dec

3.3×

INDICATIVE

So what

The $9.0m FY2026 budget runs out at the end of October; at this pace December alone ($2,154k) is 2.9× the budget run-rate.

4

SUMMARY · DIAGNOSIS · ROOT CAUSES · SOLUTION · PLAN & CASE

One dollar in five has no owner: about 140 API keys carry $277k a month.

AI run cost by use case, September 2026, $k

Source: Northmere billing and gateway logs, Sep 2026 (illustrative)

Northmere Mutual Group · AI FinOps diagnostic · Confidential

Ostrell Vane | 5

Use case

Run cost, September

Unit cost or price

Volume / utilisation

Coding agents

310

Contact-centre copilot

277

Unattributed (about 140 keys)

277

Claims-intake agent

209

Own GPU cluster

144

Reserved capacity (PTU)

Document intelligence

$258 per engineer a month	1,200 engineers
$25.20 per 100 conversations	1.1m conversations
not measurable	about 140 API keys, 20% of spend
$2.46 per claim	85k claims
$4.10 per GPU-hour	48 GPUs, 22% utilised
paid whether used or not	PTU pilot, 31% used
$26.00 per 1,000 pages	2.6m pages

Total

$1,380k

With an owner today: 80% of spend (target >95%)

So what

Of the $277k, $61k is forgotten pilots that can stop; the AI platform team tags the other $216k to an owner by 30 November.

SUMMARY · DIAGNOSIS · ROOT CAUSES · SOLUTION · PLAN & CASE

Eight causes in four families drive the overspend; none is the token price.

Note: levers applied one after another; no double counting.
Source: Ostrell Vane diagnostic (illustrative)

Northmere Mutual Group · AI FinOps diagnostic · Confidential

Ostrell Vane | 6

INDICATIVE

Addressable run cost by family and cause, $k per month, September 2026

Which part of September’s $1,380k can go?

$621k
a month addressable

Visibility

61

Forgotten pilots among about 140 unowned keys

61

Model & prompt design

368

Context resent at full price

221

Frontier model for routine turns

96

Documents run synchronously

51

Capacity

127

GPUs 22% utilised

84

Reservation 31% used

43

Usage behaviour

65

Top 5% of coding users = 38% of coding-agent spend

56

9% of claims make >40 calls

9

SO WHAT

Design is the big lever. Model and prompt design drives $368k of the $621k: context resent at full price and frontier models on routine turns.

Paid for, not used. GPUs run at 22% and the reservation at 31%; right-sizing both saves $127k a month.

One owner per family. AI platform team: visibility. Business-unit (BU) product owners: model and prompt design. FinOps lead: capacity and usage budgets.

6

All 14 slide titles

Northmere Mutual: every AI dollar attributable, governed and worth it
At September volume, $621k a month (45%) can be cut; no use case stops.
Inference got 280× cheaper in 2022–24, yet firms’ GenAI spend tripled in 2025.
Run cost rose 3.3× since January; FY2026 heads for $13.0m, 44% over budget.
One dollar in five has no owner: about 140 API keys carry $277k a month.
Eight causes in four families drive the overspend; none is the token price.
45% of September’s spend, $621k a month, can be cut without stopping any use case.
One gateway across three providers and the GPU cluster makes every dollar attributable and capped.
Four tags on every call give showback by owner and a unit cost per claim.
Unit costs fall 63–75%: a claim drops from $2.46 to $0.69 with quality gates.
Six owned KPIs track delivery; the plan already assumes only 75% of savings land.
Product owners answer for unit costs; chargeback starts in April 2027.
Quick wins fund the structural levers; the programme pays back in month 4.
Three decisions on 21 October start the programme on 2 November.

Eskerholm Bank: Cloud operating model · Cloud operating model

A bank leaves two data centres: maturity assessment, operating-model options, team topology, landing zone, migration waves and KPIs. 14 slides, with 2 linked architecture diagrams.

OSTRELL VANE

EXECUTIVE COMMITTEE · HELSINKI, 28 OCTOBER 2026

Eskerholm Cloud 2029: a target operating model to exit two data centres and run cloud as a product

Shared management, a policy-as-code landing zone and a migration factory to exit DC2 and DC3 by Dec 2028, with platform teams serving product teams as an internal service

Illustrative engagement. Ostrell Vane and Eskerholm Bank Group are fictional; client figures are invented, market data is real and cited.

900

applications to move to AWS and Azure

2

data centres to exit: DC2 and DC3

Dec 2028

when the data-centre leases end

Strictly confidential · Prepared for Eskerholm Bank Group

1
SUMMARY · CASE FOR CHANGE · TARGET MODEL · MIGRATION · VALUE & NEXT STEPS

Shared management and a migration factory exit DC2 and DC3 by Dec 2028 for €74m.

Source: Eskerholm IT data (illustrative); Ostrell Vane analysis

Eskerholm Bank Group · Cloud 2029 operating model · Confidential

Ostrell Vane | 2

Situation

Complication

Resolution

Two data-centre leases (DC2 and DC3) end in Dec 2028. 18% of workloads already run on AWS and Azure (€21m a year).

At 35 applications a quarter the move takes 6.4 years. Maturity is 1.9 of 5, a new environment takes 6 weeks, and 34% of cloud spend has no owner. EU DORA (regulation) requires exit strategies for information and communication technology (ICT) services that support critical or important functions.

Adopt shared management

One owner per layer: the Cloud Business Office owns the guardrails, four platform teams the platform, and each of 28 product teams its workloads.

Build the factory

Landing zone v1 live by 5 Feb 2027; the migration factory (40 FTE) ramps from 35 to 170 applications a quarter in 2028.

Move 900 applications in four waves

The Cloud Business Office confirms each application's path; the first pass retires 137 and retains 103.

€74m over 2027–29

Net +€18m by 2030; peak funding €37m at end 2028; payback during 2030; €31m a year from 2030.

Decide today: (1) endorse shared management, final decision 15 Jan 2027 · (2) appoint the Head of the Cloud Business Office by 30 Nov 2026 · (3) pre-release €8m of the 2027 investment for landing zone v1 and the factory.

INDICATIVE

2
SUMMARY · CASE FOR CHANGE · TARGET MODEL · MIGRATION · VALUE & NEXT STEPS

At 35 a quarter, the exit takes 6.4 years; the leases end in 2.2 years.

Cumulative applications migrated, Q1 2027 – Q4 2028

Source: Regulation (EU) 2022/2554 (EU DORA); Eskerholm data (illustrative); Ostrell Vane analysis

Eskerholm Bank Group · Cloud 2029 operating model · Confidential

Ostrell Vane | 3

Lease end, Dec 2028

Q1'27

Q2'27

Q3'27

Q4'27

Q1'28

Q2'28

Q3'28

Q4'28

W0 · 20 pilots

W1 · 200

W2 · 340

W3 · 340

120

220

390

560

730

900 wave plan

Gap: 620
applications

280 at today's pace
of 35 a quarter

SO WHAT

620 applications stranded. At today's 35 a quarter, only 280 of the 900 move before the leases end in Dec 2028.

4.9× today's pace at peak. The waves need 170 applications a quarter through 2028, 113 on average.

Regulation raises the bar. EU DORA (regulation) has applied to banks since 17 Jan 2025: exit strategies and a register of ICT third parties.

H1 2027: 20 pilots while the factory is built (see p. 11)

3

SUMMARY · CASE FOR CHANGE · TARGET MODEL · MIGRATION · VALUE & NEXT STEPS

Eskerholm scores 1.9 of 5; FinOps and portfolio are weakest of eight dimensions.

1. Ostrell Vane scale: 1 = ad hoc, 3 = standard across teams, 5 = optimised; informed by the CNCF platform engineering maturity model (4 levels) and FinOps maturity stages.

Source: Ostrell Vane maturity assessment (illustrative); CNCF (Nov 2023), Platform engineering maturity model; FinOps Foundation, FinOps Maturity Model

Eskerholm Bank Group · Cloud 2029 operating model · Confidential

Ostrell Vane | 4

PRELIMINARY

Strategy & governance

Landing zones

SecOps

SRE/Ops

FinOps

Developer experience

Skills

Portfolio

Today, average 1.9

Target 2029, average 3.6

Inner ring = 2, outer ring = 4

Dimension	Today	Target 2029	Gap
FinOps	1.5	3.5	2.0
Portfolio	1.6	3.3	1.7
Developer experience	1.7	3.5	1.8
Landing zones	1.8	4.0	2.2
SRE/Ops	1.9	3.5	1.6
Skills	2.0	3.4	1.4
Strategy & governance	2.1	3.8	1.7
SecOps	2.4	3.8	1.4
Average	1.9	3.6	1.7

SO WHAT

One owner for the weakest links. FinOps (1.5) and portfolio (1.6) move to the Cloud Business Office.

Landing zones has the largest gap (+2.2) and gates every wave: v1 by 5 Feb 2027.

Cloud maturity by dimension, score 1–5¹; table sorted by today's score, weakest first; SRE = site reliability engineering

SUMMARY · CASE FOR CHANGE · TARGET MODEL · MIGRATION · VALUE & NEXT STEPS

Shared management is the only option that scores 3 or better on all five criteria.

Source: Microsoft (Jul 2025), Cloud Adoption Framework: Prepare your organization for the cloud (models: Centralized, Shared management, Decentralized); Ostrell Vane assessment

Eskerholm Bank Group · Cloud 2029 operating model · Confidential

Ostrell Vane | 5

Three cloud operating models scored on five criteria, 0–4 (4 = best); Eskerholm today: centralised, ticket-based infrastructure

RECOMMENDED

Centralised

One central team runs governance, security and operations for all workloads

Shared management

A platform hub runs shared services; workload teams work within its guardrails

Decentralised

Each team owns its landing zone, governance and operations

Criterion

Speed for product teams

Control & EU DORA (regulation) compliance

Cost transparency

Scales to 900 applications

Change effort for Eskerholm (4 = easiest)

Lowest score

1

3

1

SO WHAT

No weak spot. Shared management scores 3 or 4 on every criterion; each alternative scores 1 on two of them. It also leads on total (17 vs 13 and 10).

Built for scale. It is the only model that scores 4 on reaching 900 applications, and it leads on cost transparency.

Control stays central. The Cloud Business Office is accountable for guardrails and EU DORA (regulation) exit plans; each product team runs its own workloads.

Ask today: the Executive Committee endorses shared management and takes the final decision on 15 Jan 2027.

Total (of 20)

13

17

10

5
SUMMARY · CASE FOR CHANGE · TARGET MODEL · MIGRATION · VALUE & NEXT STEPS

Five of 14 capabilities are new to Eskerholm; guardrails, landing zones and FinOps lag most.

1. Capabilities break down the eight dimensions on slide 4; policy-as-code guardrails (1.4) sits within Landing zones (1.8).
Source: AWS, Cloud Adoption Framework (six perspectives); Ostrell Vane assessment (illustrative)

Eskerholm Bank Group · Cloud 2029 operating model · Confidential

Ostrell Vane | 6

14 capabilities by Amazon Web Services (AWS) Cloud Adoption Framework perspective: maturity today → target 2029 (scale 1–5), then the gap

Business

Cloud value & business case

2.1 → 3.8 +1.7

Portfolio & 7R

1.6 → 3.3 +1.7

People

Cloud Academy & skills

2.0 → 3.4 +1.4

Organisation & roles

2.0 → 3.6 +1.6

Governance

FinOps & cost allocation

1.5 → 3.5 +2.0

Policy-as-code guardrails¹

1.4 → 3.8 +2.4

ICT third-party & exit risk

2.3 → 3.8 +1.5

Platform

Landing zones & account vending

1.8 → 4.0 +2.2

Internal developer platform

1.7 → 3.5 +1.8

Data platform

2.0 → 3.5 +1.5

Security

Identity & access

2.6 → 4.0 +1.4

Security operations

2.4 → 3.8 +1.4

Operations

SRE & observability

1.9 → 3.5 +1.6

Incident & change

2.2 → 3.6 +1.4

New capability for Eskerholm (5 of 14)

SO WHAT

Foundations first. Guardrails (+2.4), landing zones (+2.2) and FinOps (+2.0) start in the first 100 days.

Five are new. From 1.4–1.9 today to 3.5–4.0 by 2029 needs dedicated teams, not training alone.

Owner: CBO

Owner: CBO

Owner: Cloud Academy

Owner: CBO

Owner: CBO

Owner: CBO

Owner: CBO

Owner: Landing zone team

Owner: DevEx team

Owner: Data platform team

Owner: CISO office

Owner: CISO office

Owner: Compute team

Owner: Compute team

CBO: Cloud Business Office

PRELIMINARY

6

All 14 slide titles

OSTRELL VANE
Shared management and a migration factory exit DC2 and DC3 by Dec 2028 for €74m.
At 35 a quarter, the exit takes 6.4 years; the leases end in 2.2 years.
Eskerholm scores 1.9 of 5; FinOps and portfolio are weakest of eight dimensions.
Shared management is the only option that scores 3 or better on all five criteria.
Five of 14 capabilities are new to Eskerholm; guardrails, landing zones and FinOps lag most.
A 110-FTE central organisation serves 28 product teams, which own and run their workloads.
Single owners and self-service cut the environment wait from 6 weeks to under a day.
A policy-as-code landing zone vends a compliant AWS account in under a day.
Over half the estate moves with little change; only 10% justifies refactoring.
Four waves move 900 applications by Dec 2028; 20 pilots prove the factory first.
€74m invested in 2027–29 pays back during 2030, then saves €31m a year.
Ten monthly KPIs track delivery speed, cost ownership, the 900-application exit and savings.
Three decisions today start the 100-day plan on 2 Nov 2026.