Conway's Law in Talent Architecture: Designing Teams to Match the Software You Want
Melvin Conway proved teams ship their own communication structure as software. The Inverse Conway Maneuver flips this — HR designs reporting lines to mirror the ideal microservice architecture, not the legacy org chart.
- Conway's Law (1968): your software architecture will mirror your team's communication structure — whether you like it or not.
- If 3 teams own a compiler, you will ship a 3-pass compiler. If 6 teams own a payment flow, you will ship 6 layers of payment middleware.
- Inverse Conway Maneuver: design teams to mirror the architecture you WANT, then let the system follow.
- Used at scale by Amazon ('two-pizza teams'), Spotify ('squads'), and the UK Government Digital Service (GDS).
- HR sits at the centre of this — team design is now an architectural decision, not just a people decision.
If you walk into a tech company and read their codebase before meeting anyone, you can sketch the org chart with surprising accuracy. Conway proved this in 1968 and the pattern has not weakened in 58 years. The org IS the architecture. Which means HR is, quietly, an architect.
Conway's Law in one sentence
“Organizations which design systems are constrained to produce designs which are copies of the communication structures of these organizations.”
Translation: every interface between two teams becomes an interface in the code. Every approval chain becomes a coupling. Every Slack channel that does not exist becomes a missing API.
The Inverse Conway Maneuver
If team structure dictates architecture, then to change the architecture, change the teams first. This is the Inverse Conway Maneuver — popularised by Skelton & Pais in Team Topologies (2019).
- Reorgs follow business unit politics
- Architecture is whatever the orgs produce
- Microservices end up as 'distributed monolith'
- Cross-team handoffs everywhere = latency, bugs, finger-pointing
- Architect the target system first (e.g. 6 bounded contexts)
- Then design 6 stream-aligned teams to own them
- Each team has full ownership: build, deploy, on-call
- Interfaces are intentional — APIs and contracts, not Slack DMs
Team Topologies — four team types
| Team type | Purpose | Example |
|---|---|---|
| Stream-aligned | Owns a slice of the product end-to-end | Checkout team, Onboarding team |
| Platform | Provides internal services to stream teams | Developer platform, Data platform |
| Enabling | Coaches stream teams on a new capability | AI enablement team, SRE coaches |
| Complicated-subsystem | Owns deeply specialised tech | Search ranking, ML inference |
How HR runs an Inverse Conway redesign
- Sit with the CTO/Chief Architect. Sketch the target architecture on a whiteboard — bounded contexts, services, data flows.
- Draw the IDEAL team map — one stream-aligned team per bounded context, supported by platform teams.
- Compare to today's org chart. The deltas are the reorg.
- Move people in waves, not in one Big Bang. Three quarters is normal.
- Update the operating model: each new team owns hiring, on-call, and roadmap for its slice. No more 'central architecture committee' bottlenecks.
- Measure: DORA metrics (deployment frequency, lead time, change failure rate, MTTR) should improve within 2 quarters.
- Step 1 — Architecture diagramWhat system do we want to operate?
- Step 2 — Team boundariesOne team per bounded context. End-to-end ownership.
- Step 3 — Platform layerInternal services so stream teams move fast
- Step 4 — Enabling coachesTeach new capabilities without becoming a dependency
Amazon's 'two-pizza team' rule is Inverse Conway applied at scale: any team must be small enough to feed with two pizzas. That constraint cascaded into SOA and eventually AWS. The team design produced the architecture, not the other way around.
Takeaways
- Your codebase already looks like your org chart. The only question is whether you chose it.
- HR is an architectural function in any tech company over 50 engineers.
- Reorgs without an Inverse Conway lens regenerate the same problems with different names.
- Conway, M.E. — How Do Committees Invent? — Datamation, 1968
- Skelton & Pais — Team Topologies — IT Revolution Press, 2019
- DORA — Accelerate State of DevOps 2023 — Google Cloud DORA
- MacCormack, Baldwin, Rusnak — Exploring the Duality Between Product and Organizational Architectures — Harvard Business School, 2008
Read next
All playbooksStop talking about 'upskilling'. Start measuring exactly what percentage of each role belongs to the human and what belongs to the AI agent — and design the job around the split.
Anthropology says humans cap out at 150 stable relationships — and 50, 15, and 5 are also magic numbers. Why your team feels broken at 23 people, 52 people, and 153 people, and the refactor pattern that fixes it.
An honest map of where AI is delivering value in HR today, where vendors are overselling, the risk and regulation landscape, and a 30/60/90 pilot framework.