Month: September 2025

I did:

Magic number

Three day week for me, which really messed with my sense of time. I thought Wednesday was Friday and next week was the week after. Anyway, this happened…

• Chatted about Domain Driven Design and Opportunity Space Mapping, which seem to work really well together. They give us a way to organise, understand and assess messy, overlapping parts of the organisation.
• Talked about ways product managers can expand their responsibilities and find learning opportunities; across a domain, complementary products and adjacent roles. Product is a generalist role and there are lots ways to get useful skills.
• Started thinking about how to write a business case for some new work. I’ll probably use my usual FOPP metrics (financial, outcomes, performance, and progress) along with our north star metric.
• Joined up three problems to one solution. I like it when that happens. It feels like we’re delivering more value than tackling each problem separately.
• Thought through the things I intend to work on for the rest of the year. I’m reviewing my annual objectives soon so I want to have formed a position on whether I should be measured on outputs, outcomes, behaviours, growth, etc.

I read/listened to:

Product in service

Read Scott Colfer’s new book about what product management looks like in large, often non-commercial, organisations. There’s a lot to agree with. I’m a longstanding believer in the product management role being highly contextual and that applying product thinking is more important than “doing product management.”

Structure and social skills

Interesting podcast challenging Tuckman’s model for forming teams and concluding that it doesn’t really work. Instead, to be successful, teams need the right structure and the social skills to work together.

From Feature Factory to Finance Fluency: How I Learned to Speak CFO

“Executives aren’t against our ideas; they just can’t hear them when we speak in features instead of finance.” I completely agree product managers should be thinking and taking about business objectives such as income and cost, but I can’t help thinking that the other side of the problem is the attitude and approach of leadership towards product managers. When leaders believe their job is to do all the thinking (including understand the financials) and just give product managers features to build, then it’s not surprising that’s the language product managers speak back to them.

Software as signs

“The software your organisation produces is a product of and a reflection of how your organisation operates, as well as its beliefs and values.” I guess that makes software a lagging indicator of culture (just as conversation is leading indicator of culture). I wonder how you’d go about assessing an organisation’s software and what conclusions you’d reach about their culture.

Delivering successful AI after it breaks your existing operational model

I love Sarah’s perspective. She suggests our current product delivery model is outdated as it’s designed for deterministic software not nondeterminisitic, probabilistic systems like GenAI. I’m not sure more controls is the answer but it poses an interesting question to ponder… what would a nondeterminisitic product development process look like?

I thought:

Layers

An architect told me about how they have the idea of Layers. Made me wonder if we can think of products in a similar way, with an interface layer, an action layer, an outcome layer, etc. And that made me think about James Reason’s idea of Swiss cheese layers where failure is because the holes in layers line up.

Do as AI say not as AI do

There are some interesting hypocrisies in organisation’s position on AI. They want to use it to their advantage but don’t want others to use it in ways that disadvantage them.

A few months ago, Shopify CEO Tobi Lütke stated that staff had to prove a job couldn’t be done by AI before it can be filled by a person. This week, Spotify published their position on AI music on their platform and launched tools to stop it because people are using AI in ways that undermine their business model.

And in hiring, organisations don’t want candidates to use AI to apply for jobs, but they’re happy to use AI to screen CV’s and they want their staff to use AI in their day-to-day work.

I did:

• Picked up my new laptop.
• Talked about workload management and how we think about what is worth doing.
• Wrote up my thinking on using test and control groups to improve the robustness of financial contribution reporting. I love this kind of critical thinking.
• Chatted about Domain Driven Design, architecture, mapping tech stacks and change implications. Although I’m a firm believer that product management isn’t about the technology, it’s also not much without the technology.
• Got involved in ‘a rock and a hard place’ problem. No viable solutions have presented themselves yet.

Success state roadmap

There is no single roadmap template that covers every part of the product development lifecycle. How weird is that? Now/Next/Later roadmaps are for discovery, and Gannt-type charts are usually used for delivery, and there’s no format for aligning around success.

So I wrote about success state roadmaps and how they show us whether our strategy is leading to a successful product. And I also got a mention in one of Paul Brown’s articles.

I saw:

Look up

I saw the International Space Station and the President of the United States. I’ll let you decide which one was more impressive.

I read/watched:

AI and Product Management: Becoming More Evidence-Guided

“AI’s biggest promise is not in accelerating the way we work today, but in helping us switch to evidence-guided, discovery-driven work.” This is a step forward in the AI as a tool for product managers thinking as it suggests AI can make it easier to do some of the things that are currently hard to do.

The future of product management

Notes on Notes from Ray Wang’s talk at INDUSTRY 2025:

• Companies that build AI-first business models will outperform the market 10 to 1
• The job as a product manager involves collecting data and monetizing that data.
• Realization that there is not enough data to get to a level for stakeholders to trust… and the need to pool data by industry value chains and build partnerships for data.

Digital-era healthcare

I read James Plunkett’s piece on digital-era healthcare and James Higgott’s responses, particularly his point about hierarchical power structures and the ‘do as I say’ culture at the NHS. I’d take what James Higgott says a level deeper and say the big problem with the NHS is that the wrong people have the power.

If digital is about anything, it’s about disrupting power structures. That’s what being user-centred is about. It’s how many of the early Internet products and companies succeeded. So, if the NHS wants to become a digital-era organisation it needs to drastically disrupt its own power structures.

Doctors know they have all the power. They know they are protected from complaint by the patient advice and liaison service. As a carer, I’ve seen it first hand many times. NHS staff making decisions that aren’t in the best interests of their patient and knowing there is nothing patients can do about it.

For the NHS to really change, it needs to fundamentally change the relationship between NHS staff and patients. It needs to put patient power first.

Intentionality FTW

Watched John Cutler describing the journey of how Dotwork figured out mapping product operating systems. Interesting as always. It’s really hard to know how much company operating systems, product operating models, etc., make a difference to the success of the products and organisations. One thing’s for sure, there’s no perfect answer. But I always go back to the question, “What are we optimising for?”

I thought:

Conveyor belts and mazes

Fundamentally conceptually, most products are conveyor belts. Their primary purpose is to take users from one state to another. It’s like I’ve said before, products don’t achieve outcomes, products change user’s behaviours and it’s that change that achieves an outcome (it’s why outcomes are unpredictable). And moving users along the conveyor belt is how those products change behaviour.

The problem with conveyor belt products is they are follow an industrial-age, mechanistic worldview that doesn’t fit our understanding of the world as a complex system.

So far, the only alternative to conveyor belts is mazes. These are products that are designed to keep you in the same state for as long as possible. Their definition of outcome is that user behaviour is changed in more predictable ways and then maintained for as long as possible.

The problem with maze products is they rely on algorithms, and we haven’t yet learned (or don’t want to because capitalism) how to make algorithms that don’t create negative spirals of doom (yeah, I’m talking about social media).

When I see thought-leaders complaining about LinkedIn’s algorithm not showing their super insightful posts to enough people, I think they are confusing a maze product with a conveyor belt product. They expect LinkedIn to help them get somewhere when really LinkedIn just wants to keep them posting.

Fixed and variable value

If you’ve ever done any business cost analysis you’ll know that variable costs change with the amount a business you do and fixed costs stay the same not matter how much business you do. I wonder if the same thinking applies to value. So, legal compliance, security, etc., are fixed value. An organisation needs them whether they have a thousand or a million users. The value (however you might define it) stays the same. Product, on the other hand, is variable value. The more users behaving in ways that drive business results, the more value is created.

Cost calculator for automation

I’ve been wondering how to quantify the question of whether its worth automating a process?

Maybe the calculation looks something like this: (Upfront cost of setting up automation + ongoing cost, e.g., licenses, monitoring, etc.) <> (Time cost of doing it manually x number of times action undertaken – value of what could be done instead) = Yes or no.

The evolution of the digital product

At first, ‘digital product’ meant a product that used technology (software). Then, the Internet came about and a ‘digital product’ became one that used technology and data to create and exchange value. Now, a ‘digital product’ uses technology, data, and artificial intelligence. This gives us a way to think about AI as using data in value-generating ways that weren’t previously possible. Soon enough, technology, data and AI will be the usual for digital products and whatever is next will be at the frontier.