Independent AI advisory · Hamburg
Most AI programmes have a strategy. Few make it to production.
They don’t fail on the technology – they fail in the middle, between strategy and delivery. That’s where I work: with leadership and delivery teams in knowledge-intensive, regulated, and EU-facing organisations, where speed and rigour need to go together.
01 / Services
How I can help
Three engagement shapes, depending on where you currently stand. I’m at my best when you need clarity, decisions you can act on, and momentum – not more slides.
Strategy
AI strategy and operating model
Clarity on priorities, approach, and the operating model behind them. Bringing ambition and reality into a roadmap that teams can align on.
Typical outcomesPrioritised use cases, a workable operating model with roles and guardrails, success metrics teams can align on, and a roadmap with a clear sequencing logic.
Architecture and delivery
From pilot to production
Turning ‘we should build something’ into a real plan – and the plan into a production system that holds up in regulated contexts.
Typical outcomesReference architectures fitted to your system landscape, build-vs-buy assessments, production-grade implementations with quality and safety checks, and documentation that stands up to internal audit.
Sparring and oversight
A reliable sparring partner
For AI governance, vendor selection, architecture reviews, and critical delivery decisions – on a continuing basis or in clearly bounded reviews.
Common topicsTranslating governance into lived practice, technical and commercial vendor due diligence, delivery risk reduction, and coaching for product and engineering leads.
02 / Approach
How I work
Four steps I keep across almost every engagement – whether it’s strategy, architecture, or a specific production effort.
Understand the reality
What’s already in place, which constraints are non-negotiable, where the real bottlenecks are. Before any recommendation comes an honest read of the starting point – including the topics nobody internally likes to name.
Make decisions visible
Trade-offs, risks, dependencies, and ownership, framed so leadership teams can actually decide them – not keep deferring them.
Get something working
Even in strategy engagements, I back assumptions with tangible outputs – a prototype, an architecture sketch, an evaluation. That cuts uncertainty faster than another round of workshops.
Build capability, not dependency
Frameworks, decision templates, and a clear scaling path – I hand over so your teams can carry on without me. That’s an explicit part of the mandate, not a friendly side-effect.
03 / Selected engagements
What I’ve been working on lately
Three anonymised excerpts from current and completed mandates. Happy to share more in conversation.
A suite of specialised AI applications, each built around a specific team workflow
Conception and build of an application family of eight specialised AI web apps – each developed in close collaboration with the relevant team. The approach: first deeply understand what is actually being done, then build an application that supports that workflow as well as possible, instead of a generic tool that has to somehow accommodate the workflow afterwards. Tech stack: Next.js, PostgreSQL, S3, Microsoft Entra ID, Azure Key Vault. In production with field-level confidence scoring and auditable documentation.
Group-wide AI policy and three-tier governance model
A risk-based four-tier usage model, a three-tier decision-rights model with named accountability, an AI system register, and a formal acknowledgement mechanism for all staff. Aligned with the EU AI Act and ISO 42001, with a compliance roadmap to the August 2026 enforcement deadline and a strategic briefing for the executive team that frames AI as a business-model question, not just a technology one.
Research platform with a six-stage Intelligence Engine
Strategic conception, product architecture, and go-to-market planning for an AI-powered platform that systematises the leap from raw AI output to decision-ready market analyses. Six stages with defined human control points, evidence-based individual claims with confidence scoring, and a four-layer quality assurance process. Target market: institutional investors and strategy consultancies in the healthcare sector. Each deliverable combines a condensed core analysis, transparently sourced individual evidence, a clear separation of facts, estimates and judgements, plus a structured risk and red-flag overview.
04 / Tools
Frameworks and formats I use regularly
Built up over years, tested across multiple engagements, adapted each time to the context. Not a methods catalogue for its own sake.
Five AI skills
A tool-agnostic competence model with five transferable skills for AI literacy. Each skill comes with a guiding question, core dimensions, and concrete development steps – an anchor for curricula, self-assessments, and role profiles that holds up even as the tools change.
Risk-based AI usage model
A four-tier model for AI use in organisations – Approved Tools, Permitted Low-Risk Use, Approval Required, Prohibited – with an AI system register and a formal acknowledgement mechanism. Translates the EU AI Act into daily operational decisions, not just compliance documents.
AI Council for strategic questions
A structured multi-perspective methodology: five AI advisors with distinct roles analyse a question independently, then anonymously peer-review one another, with a chairman synthesis that lands on a clear verdict and one concrete next step. Used for market analyses, portfolio decisions, and positioning questions.
QA workflow for AI in deliverables
A six-step process for handling AI use in client deliverables: pre-use classification, tool selection, drafting, human review, disclosure decision, final approval – with a standardised disclosure template. Answers the increasingly common requirement to indicate where and how AI has been used in a deliverable.
05 / Industries
Where I add the most value
Deliberately narrow. In other contexts there are better-suited advisors – and I’ll happily name them when I know any.
EU institutions and institutional donors
Multilingual, stakeholder-intensive contexts with high transparency requirements. Innovation and compliance need to be thought through together, not sequentially. Experience with procurement logic, donor rules, and translating strategy into lived practice.
International advisory and knowledge organisations
Multi-location consulting groups with distributed teams, in-house knowledge production, and multiple business units. Introducing AI where it actually changes workflows – not as a tool selection question, but as a question of operating model, governance, and capability.
Healthcare and Life Sciences
Manufacturers, providers, and investor-facing market intelligence in the highly regulated healthcare sector. AI applications where evidence, traceability, and auditable documentation determine market acceptance and approval – at the intersection of MDR, GDPR-Health, and the EU AI Act.
06 / Fit check
When I’m not the right fit
At least as important as the list above. If anything below applies, I’ll say so early and, where I can, suggest alternatives.
I take on a limited number of engagements at any given time. This list does the pre-filtering that would otherwise cost two meetings:
- Pure body-leasing or hour contingents without a strategic frame.
- Writing tender or grant submissions end-to-end without clear input from your side.
- ML research questions at the model or algorithm level – there are better addresses for that.
- Initiatives without a client-side decision owner. Advisory alone won’t carry it.
- Contexts where speed clearly matters more than rigour – that rarely fits how I work.
07 / About
About
For over 20 years I have worked with organisations on complex technology initiatives – from enterprise architecture and digital transformation to AI strategy and delivery in regulated environments. What interests me most is the messy middle: where strategy meets delivery and ‘we should’ becomes ‘we shipped’.
What makes that middle so hard is, almost always, a translation problem: leadership, governance, architecture, and delivery speak different languages. That’s where my work sits.
I currently lead AI initiatives in a communication-intensive, EU-facing environment, and work in parallel with several organisations as an independent adviser across Europe. My approach is straightforward: understand the context, identify what matters, deliver pragmatically.
Alongside this, I build and run Weekado, an AI-native weekly planner for individuals. Taking my own product all the way into live operation keeps me close to the part I advise on: getting things into production.
08 / Contact
Tell me what you’re trying to achieve.
A short note with two or three sentences is enough: what you’re working on, what’s getting in the way. I’ll respond promptly and suggest a sensible next step – even if I’m not the right person in the end.
martin@breitsprecher.de