AI investment · Product & portfolio leadership

From AI investment
to value in production.

AI activity is not AI impact. The real test is whether the business makes better decisions, creates more value, and can sustain the change at scale.

Many use cases and PoCs → a portfolio worth building
Business transformation, data readiness and organisational capability
Build, recover or stop—with cost accountability

Published September 2026

The portfolio story Financial & operating outcomes ↗

Commercial software portfolios. Production AI. Technology economics.

Investment · Portfolio · Business change · Data · Engineering · Value

Find the Seam.

01 · Business outcomes

Business outcomes.

Value validated. Products in use.
Operating economics made visible.

Financial value
SEK 20M+
≈ €1.9M Annualised Economic Value Delivered
Volvo Cars
DKK 1M+
≈€134K+ Finance-verified annual cloud savings
TDC Brands
Global production LLM25–30%
Validated task-efficiency improvement
Contract Intelligence Copilot · Volvo Cars
450+ users
Flagship programme · Group Management Team visibility
TDC · Customer Issue Intelligence
No systematic categorisationAutomated categorisation in production
≈2 FTE capacity released
≈9 months in production
Cost to serve≈70%
Lower inference cost per interaction
TDC Brands

02 · Investment & portfolio direction

What is worth
building?

Since 2012, I’ve built systems and led teams that improve business decisions—from commercial optimisation software to production AI.

Making employees 5x faster doesn’t automatically make a business 5x faster. The bigger opportunity is to use advances in technology to redesign how work gets done—automating decisions and processes, eliminating manual bottlenecks, and enabling increasingly autonomous, real-time operations.

The starting point

Experiment freely to discover value.
Invest selectively to capture it.

Use cases and PoCs
Stalled initiatives
Imperfect data

My investment judgment

Business value

What changes in the business—and is it worth funding?

Delivery readiness

Can data, integration and engineering support the decision?

Ownership

Who changes the process and operates the capability?

Economics

What does it cost to deliver, run and scale?

The commitment

Invest
Recover
Redirect
Pause
Stop

How I commit capabilityBuy mature capability. Enable teams through shared platforms. Commit specialist engineering where a custom product is worth owning.

03 · Decisions & consequences

Investment choices.
Commercial consequences.

Validate value
Commit and recover
Redirect or stop

Build & embed

Protect warranty expenditure at scale.

Before: Rules-based screening and limited human-review capacityPayment deadlines narrowed the window for assessment
Decision: Originated the product; redesigned claims prioritisationLed the team from business problem to production
After: ML claim ranking and LLM support assisted daily warranty reviewAnalysts retained claim decisions

Flagship programme · Group Management Team visibility

Contract Intelligence Copilot

Global procurement.
Contract intelligence for compliance and efficiency.

Before: Manual search across systems, covering thousands of suppliers and contracts
Decision: Recommended a self-built copilot over vendor solutions that assumed mature data foundations, matching the delivery approach to the data-quality challenge.Led product definition, validation and global productionDesigned a user-feedback loop to surface data-quality issues and guide improvements to the underlying data and the copilot.
After: Streamlined contract access helped buyers follow procurement rules

Recover

Turn years of stalled investment into regular delivery.

Reset direction and delivery through organisational change.

Before: The wrong car mix risked discounting configurations customers did not wantStalled for years; sparse historical data
Decision: Redesigned the operating model around business value and regular deliveryRe-engaged business units and advanced with sparse historical data
After: A production MVP and regular delivery every two weeks

Redirect

Challenge the premise of a C-level strategic initiative.

Before: C-level initiative: predict order response to price changes
Decision: Led the team; backtesting did not support the premiseAgreed a data-led redirection with the Head of Global Pricing
After: Redirected development towards simpler analytics and further data collection

Platforms · Governance · FinOps

Production AI & operating economics

I established GenAI platforms at Volvo Cars and TDC Brands, working with vendors and internal teams.Across these mandates, the work connected legal, cybersecurity, cloud and infrastructure alignment with agentic capability, production guardrails, data security and GDPR requirements. At TDC, I established FinOps and business-unit cost accountability to make economics part of scaling decisions.

FinOps & business accountability

Make technology economics a business responsibility.

FinOps: I established FinOps; rationalised cloud resources to eliminate avoidable costs.
Ownership: I redesigned cost accountability: business units owned their technology expenditure.
Scaling: Unit economics and business value became part of AI investment and scaling decisions. Inference cost per interaction fell by ≈70%.

Regional deployment

Maintenance Copilot for Volvo Cars’ top-performing plant globally.

Constraint: EMEA and China could not share data or AI models.
Decision: Enabled a regional agentic platform to support the plant’s maintenance delivery within China’s separate data and model environment.
Adoption & data quality: The copilot was adopted across maintenance engineering teams. In parallel, we continued to address underlying data-quality issues, with engineers retaining responsibility for maintenance decisions.

04 · Operating model & capability

How I make AI
investment deliver.

My framework became a company asset for new initiatives and turnarounds. It connected product delivery with business-process redesign: how decisions and workflows would change, who owned the outcome, and what evidence justified the next commitment.

We advanced with imperfect data, addressing the gaps that mattered to the business decision and strengthening data modelling, pipelines and integration as products developed and scaled.

Iterative & incremental development

First working increment
Add useful capability
Extend what works

Within every incrementBuild test with users learn & refine

Feedback strengthens the product, workflow and data throughout the lifecycle.

AI product-development frameworkUsed by the three teams I incubated · two workstreams through the lifecycle
Stage	Business-process change	Product & engineering
01Understand	Outcome, baseline and market assumptions	Available data, modelling needs and technical feasibility
02Prove value	Test the proposed workflow with users	Test decision performance; refine data models and pipeline requirements
03Commit	Business case, resources and operating owner	Production pipelines, data quality and remaining uncertainty
04Operate	Adoption, business value and changing conditions	Strengthen data foundations as use scales; monitor quality, cost and maintenance

At each review, the decision could be to continue, change scope, pause or stop.

I made responsibilities explicit across the use-case lead, business-process change lead, product lead and data science lead.