Weeknote 19 – 23 January

I write these weeknotes mainly for myself, as a tool for reflection and an aide memoire, and share them openly for anyone who finds them useful.

This week was bitty, with LOTS of meetings and chats that made me reflect on the role of government and public sector, and how the services we deliver are different from commercial offerings. That distinction matters when we decide what services are valuable for us to provide, where the market can meet needs instead, and how we approach buy versus build decisions, market enablement, or regulation. It also helps articulate the unique value (USP) of our products and services.

A bit of context

I studied Politics and International Relations at uni, and later on did a Masters in Public Administration, so I enjoy geeking out and reflecting on how the theory of government and public services fits with reality.

Below, I explore some of the main considerations I find most useful when thinking about whether we should develop services in house. This is not exhaustive or definitive. These distinctions are subjective, and shaped by different worldviews and political ideologies. I’ve used the AI Health Coach product, which we’re jointly working on in our portfolio with the Department of Health and Social Care (DHSC), to exemplify how these dynamics play out in practice. 

Designing for equity and universal access

Unlike most private sector offerings, which often use price as a core exclusion mechanism, public services need to be universally accessible. They must provide a meaningful alternative for people who can otherwise not access any or the right support.

There is a growing market of AI-enabled apps and tools that help people build and stick to healthy habits. Many rely on subscription models or expensive add-ons such as wearables. And only in the last month both ChatGPT and Microsoft Copilot have launched their dedicated Healthcare offerinngs.

A key benefit of developing an AI Health Coach in the public sector is the ability to provide it free at the point of use, and to design it deliberately so that it does not exacerbate existing health inequalities.

Equity is not an accidental by-product for us. It has to be a core design constraint.

Appetite to risk

Government cannot launch products that are clinically unsafe. We have a duty of care that goes beyond what many private companies hold themselves accountable for. This means operating to a very high bar for clinical safety and regulation.

For an AI Health Coach, the risks are real – from recommending someone who’s experiencing symptoms of a stroke to go for a run, to not noticing someone is experiencing a mental health crisis and escalating out appropriately. Meeting these risks responsibly requires rigorous clinical safety testing and regulatory compliance.

Many private sector tools operate as black boxes and focus narrowly on constrained behaviours. That can be commercially viable. It is not sufficient when you are designing for real people in complex circumstances¹.

Accountability and value for money

Public services must provide value for money. We pay taxes and expect that money to be used well. At the same time, demand is high and resources are limited.

This creates a tension. Evaluating benefits in complex systems is difficult, particularly when outcomes are long-term or preventative. For the AI Health Coach, this means being disciplined about using data and in-depth user research to guide decisions, rather than relying on hunches or unproven experimentation.

Uncertainty does not remove the obligation to be accountable. It raises the bar on how we make decisions.

Efficiencies and scale

There are services that government is uniquely placed to provide at scale. In physical infrastructure, such as roads, this is widely accepted. In digital infrastructure, the case is often less obvious, but still compelling.

Shared data, common platforms, and national entry points and curation surfaces such as the NHS App act as digital plumbing. These are foundations that are usually better built once, centrally, and to a high standard.

For the AI Health Coach, there is long-term value in designing it to integrate with other NHS channels and data. Being delivered by government makes that integration easier and more sustainable.

A note on delivery

These considerations help explain why we might choose to work on a service. They do not dictate how that service should be delivered.

There is still significant scope for government to improve how it works. Being appropriately cautious about clinical risk or health inequalities does not mean delivery should be slow, closed, or overly bureaucratic. We should still work in the open, be bold, adopt iterative test-and-learn approaches, and thrive on taking calculated risks in small, controlled ways to avoid larger failures later.

What I am reading this week

This week I did not finish a single book and instead have three on the go, which I hope to complete next week (otherwise there may be mild anxiety!).

But I did read something fun.

Prompted by a discussion on clinical reference sets, I enjoyed a blog post on the ‘weird and wonderful world of SNOMED codes’, followed by conversations with colleagues about their own discoveries of obscure and unexpected entries.

I am increasingly convinced these codes are ripe for turning into a game of Cards Against Humanity – SNOMED edition. This may yet happen soon.

---

1 – Sarah Fisher wrote a great blog post on this topic about why we can’t just be designing for those who wear smart watches