Sign inGet started
Skills / Engineering-leadership

Team Topologies

Team structure is the foundation of software architecture. Get the team structure wrong and no amount of technical investment will produce the software system you want — because Conway's Law will shape the architecture whether you design it or not.

Team Topologies is a prescriptive organizational design framework from Matthew Skelton and Manuel Pais that gives engineering leaders a structured vocabulary and a repeatable process for structuring software teams. Built on four fundamental team types and three interaction modes, it replaces ad-hoc org design with deliberate, flow-optimized structures. By applying Conway's Law as a design tool rather than an obstacle, teams can eliminate hidden dependencies, reduce cognitive overload, and accelerate software delivery.

By Matthew Skelton & Manuel Pais · Free
Specimen 01 · Live diagnosisTeam Topologies
Input

“We have about 120 engineers. There's a Platform team of 15 that handles infrastructure and CI/CD, a Data team of 20 that owns the data warehouse and…”

Diagnosis
Let's map your team topology.
Full transcript ↓
Calibrated referenceagent-skills.ai
The gap

Stop fighting Conway's Law — design team structures that produce the architecture you want

Team Topologies rests on two foundational pillars: four fundamental team types and three core interaction modes. The four types are: Stream-aligned teams (the primary delivery unit, owning an end-to-end flow of business value from a domain segment), Platform teams (providing self-service internal products to reduce cognitive load on stream-aligned teams), Enabling teams (specialists who help stream-aligned teams overcome obstacles without creating permanent dependency), and Complicated-subsystem teams (handling deep-specialist domains such as ML pipelines or video codecs). Teams interact via exactly three modes: Collaboration (close joint work with shared ownership for a bounded discovery period), X-as-a-Service (clean API contract with minimal touch, for stable interfaces), and Facilitating (active coaching where one team helps another adopt new capabilities and then steps away). The framework explicitly measures cognitive load across three dimensions — intrinsic (domain complexity), extraneous (tooling and environment friction), and germane (learning investment) — and uses it to calibrate team size and scope. The 'Reverse Conway Maneuver' is the central design move: deliberately shape team boundaries to produce the desired software architecture, rather than letting accidental communication structures produce accidental systems. Topology patterns are dynamic: explicit sensing triggers signal when to shift interaction modes as products and teams mature. Each team publishes a Team API defining how it interfaces with others — code ownership, documentation, versioning, communication channels, and working agreements.

The problem

Most engineering organizations are designed by accident — teams form around people who happened to work together, inherit legacy ownership, and accumulate responsibilities until they are cognitively overloaded. The result is teams blocked by platform bottlenecks, stream-aligned work interrupted by shared-service queues, and Conway's Law silently encoding organizational dysfunction into the software architecture. Leaders sense something is wrong but lack a precise vocabulary to diagnose the topology or a framework to redesign it.

The solution

Stop diagnosing team problems with vague terms like 'communication breakdown' or 'ownership gaps' — apply the four team types and three interaction modes to produce a precise topology diagnosis, a concrete target design, and a sequenced migration path your organization can execute.

You bring
  • A description of your current team structure — team names, sizes, and what each team owns
  • Observable pain signals: delivery bottlenecks, teams that feel overloaded, unclear ownership boundaries, or excessive cross-team coordination
  • Your target software architecture or system decomposition goals
  • Organizational constraints: headcount limits, existing reporting lines, and technology stack context
You get
  • A current-state topology map classifying each team into one of the four types with anti-patterns flagged
  • A cognitive load assessment per team, identifying which are overloaded and why
  • A target topology design with specific team type assignments and interaction mode specifications for every significant team pair
  • Team API templates for any team, defining its interface contract with the rest of the organization
  • A phased topology evolution roadmap with sensing triggers for when to shift interaction modes
In action

Watch the methodology work.

Three specimens from a single real session: the same situation, unaided and calibrated, the full transcript, and the skill answering live in the channel where the work happens.

Fig.01The same situation, two ways
Specimens A · B
Specimen AVague
Unaided

“Six product squads wait an average of 9 days for every infrastructure change because the Platform team runs a single request queue. The Data team lead is context-switching across four squad projects simultaneously and has missed two warehouse stability milestones. An engineering director suspects the solution is 'more platform engineers,' but can't explain why the current structure produces these outcomes — or what a better structure would look like.”

×Vague — nothing concrete to act on.
Specimen BCalibrated
With Team Topologies

“The Platform team has published a self-service developer portal: squads provision environments, configure CI/CD pipelines, and manage deployments without filing tickets — the interaction mode is X-as-a-Service with a documented Team API. The Data team has separated its work: warehouse and ML infrastructure are protected as Complicated-subsystem work; a rotating Enabling cohort runs time-bounded analytics capability engagements with squads, then exits cleanly. The engineering director has a topology diagram and interaction mode matrix that grounds every resourcing and restructuring conversation in a shared framework.”

Specific, grounded, and ends with one clear next step.
Fig.02A real session, unedited
In-app · assistant
Team TopologiesClaude
Team Topologies
Let's map your team topology. Before I can recommend changes, I need to understand your current structure. Can you walk me through your engineering teams — names, rough sizes, and what each one primarily owns? I'm especially interested in where delivery is slow, where teams feel overwhelmed, or where ownership is unclear.
We have about 120 engineers. There's a Platform team of 15 that handles infrastructure and CI/CD, a Data team of 20 that owns the data warehouse and ML pipelines, and six product squads of 10-15 each. The squads feel constantly blocked waiting on Platform, and the Data team gets pulled into every squad's analytics needs.
Two things stand out immediately in Team Topologies terms. First: if squads are constantly blocked waiting on Platform, the Platform team is operating as a gating dependency rather than a self-service internal product. A Platform team should reduce cognitive load on stream-aligned teams via X-as-a-Service — squads should provision what they need without waiting for a response. The blocking behavior suggests the interaction mode has defaulted to something closer to a shared-services queue when it should be self-service consumption. Second: the Data team being 'pulled into every squad's analytics needs' is a sign it's trying to be both a Complicated-subsystem team (owning the warehouse and ML infrastructure — appropriate) and an on-demand analytics service (not appropriate). Let me ask: when the Data team gets pulled into a squad project, is a data engineer temporarily embedded, or do squads submit requests to a queue?
Both, actually. They have a formal request queue for dashboards and ML models, but sometimes a squad pulls a data engineer in for a sprint when something is urgent.
The 'borrowing a data engineer for a sprint' pattern is improvised Collaboration mode — it resolves the immediate pain but doesn't scale and creates extraneous cognitive load for the Data team through context-switching. The formal request queue is the same problem institutionalized. In both cases, the root issue is that the Data team has accumulated two distinct types of work that have different optimal interaction modes: the warehouse and ML platform work is Complicated-subsystem work requiring deep specialist focus — that should be protected. The squad analytics work is a facilitating or self-service opportunity, not a permanent service dependency. How much of the Data team's capacity would you estimate goes to reactive squad support versus warehouse and ML platform work?
Probably 60% reactive. Our data lead brings this up constantly. We've been debating embedding data engineers into squads, but the data lead worries about losing platform cohesion.
Your data lead is right to resist embedding. In Team Topologies terms, embedding stream-aligned resources from a Complicated-subsystem team is a common but counterproductive response — you get short-term squad unblocking at the cost of the specialist depth that makes the team valuable. The better intervention is to separate the two types of work rather than dissolve the team. Here's the pattern that fits: protect the warehouse core and ML infrastructure as genuine Complicated-subsystem work — this team publishes a self-service data platform API (standard ingestion tooling, documented data contracts, ML model serving endpoints). Squads consume via X-as-a-Service. Free the remaining capacity to operate as a time-bounded Enabling function — squads that need analytics capability get a deliberate engagement where a data specialist helps them build that capability and then steps away, instead of owning the squad's analytics indefinitely. This way the deep platform work stays protected, and the squad relationship becomes an enabling relationship instead of a blocking queue. Want me to sketch out the full target topology including the interaction mode matrix and the Platform team's bottleneck issue?
Yes — this reframes the whole problem. I've been treating this as a headcount issue but it sounds like an interaction mode issue.
Exactly — and that's the central insight of Team Topologies. Adding headcount to a team operating in the wrong interaction mode usually amplifies the bottleneck rather than resolving it. Here's my initial target topology recommendation:
Data Platform sub-team (~12 people, Complicated-subsystem type): Owns warehouse core, ML infrastructure, ingestion pipelines. Publishes a Team API with documented data contracts, self-service tooling, and ML serving endpoints. Interaction mode with squads: X-as-a-Service — squads consume without engaging the team directly.
Data Enablement function (~8 people, Enabling type): Time-bounded engagements (one sprint to one quarter max) helping squads instrument their product, build dashboard capability, or adopt a new ML feature. Interaction mode: Facilitating. The explicit goal is to transfer capability and exit — not to create a permanent analytics dependency.
Platform team: Needs a separate cognitive load audit, but the same principle applies — the target is self-service developer tooling that squads can consume without filing tickets. I'd recommend a 60-day roadmap to publish a golden-path CI/CD pipeline as the first X-as-a-Service offering.
Shall I generate the full Topology Evolution Roadmap with phased migration steps and the Platform team's Team API definition?
Reply to Team Topologies
What the skill is doing
01Topology Assessment & Anti-Pattern Detection
02Cognitive Load Audit
03Team Type Classification
Fig.03The same skill, in the channel you already use
WhatsApp · live
9:41
TTTeam Topologiestyping…
Messages are end-to-end encrypted. No one outside of this chat can read them.
Today
We have about 120 engineers. There's a Platform team of 15 that handles infrastructure and CI/CD, a Data team of 20 that owns the data warehouse and M…9:14✓✓
Before I can recommend changes, I need to understand your current structure.9:14
Can you walk me through your engineering teams — names, rough sizes, and what each one primarily owns?9:15
I'm especially interested in where delivery is slow, where teams feel overwhelmed, or where ownership is unclear.9:15
Message

The same skill, where the work happens.

No new app to learn. The methodology runs over the WhatsApp Business API, so the answer lands as a reply in the thread you’re already in — same rigour, zero context-switch.

Reads the situation, names the pattern, returns one concrete next move.
Delivered in seconds, inside a conversation that already exists.
Specimen · WhatsApp Business API · live
Capabilities

What it does, specifically.

Each capability is a distinct move drawn straight from the source methodology — not a generic assistant guessing.

CapabilityC-01

Topology Assessment & Anti-Pattern Detection

Analyzes your current team structure against known Team Topologies anti-patterns: stream-aligned teams carrying platform responsibilities they shouldn't own, enabling teams that have become permanent bottlenecks, or platform teams operating as gating dependencies rather than self-service products. Surfaces the mismatch between your actual topology and the topology your software architecture requires.

Based on Skelton and Pais's catalogue of observable topology anti-patterns — including 'flow-blocking platform teams,' 'ad-hoc topologies,' and 'unintentional complicated-subsystem teams' — documented across enterprise case studies in the book.
CapabilityC-02

Cognitive Load Audit

Assesses each team's aggregate cognitive load across three dimensions: intrinsic load (inherent domain complexity), extraneous load (tooling friction, process overhead, context-switching), and germane load (learning investment for capability growth). Identifies teams where total load exceeds sustainable capacity and recommends scope adjustments, enabling team interventions, or platform investments to reduce extraneous load.

Applies Skelton and Pais's three-category cognitive load model and their core design principle that team scope and size must be calibrated so cognitive load stays within the team's capacity — the primary constraint used to set team boundaries.
CapabilityC-03

Team Type Classification

Classifies each team in your organization into exactly one of the four Team Topologies types: Stream-aligned, Platform, Enabling, or Complicated-subsystem. For each classification, explains the appropriate behavioral expectations, size range, ownership model, and success metrics — and flags teams currently behaving outside their intended type.

Applies the four fundamental team types with their specific characteristics and decision criteria as defined by Skelton and Pais, including the principle that stream-aligned teams own 'a flow of work from a segment of the business domain' with end-to-end accountability.
CapabilityC-04

Interaction Mode Design

Specifies the correct interaction mode for every significant team pair: Collaboration (joint work with shared ownership, time-bounded), X-as-a-Service (clean API contract, minimal interaction), or Facilitating (active coaching to transfer capability). Identifies pairs currently using the wrong mode and defines sensing triggers for when the current mode should transition.

Applies the three core interaction modes with their specific conditions, duration norms, and transition triggers — including Skelton and Pais's guidance that Collaboration should be time-bounded and transition to X-as-a-Service once interfaces stabilize.
CapabilityC-05

Reverse Conway Mapping

Works backward from your desired software architecture to derive the team structure that will naturally produce it. Given a target system decomposition — desired service boundaries, domain ownership, API contracts — generates the team type assignments and interaction modes that cause Conway's Law to work in your favor rather than against you.

Operationalizes Skelton and Pais's Reverse Conway Maneuver: using Conway's Law as a deliberate design tool by deriving the target topology from the target architecture, rather than letting accidental communication structures produce accidental systems.
CapabilityC-06

Team API Definition

Generates a Team API template for a specific team covering: owned code and services, documentation standards, versioning practices, supported communication channels, meeting availability, and working agreements — making the team's interface to the rest of the organization explicit and manageable.

Based on the Team API concept from Team Topologies, which treats each team's interface as a first-class design artifact analogous to a software API, reducing inter-team coordination overhead by making contracts explicit.
CapabilityC-07

Topology Evolution Roadmapping

Creates a phased migration plan from the current topology to the target state, specifying which changes to make first (low-risk wins versus structural shifts), what sensing mechanisms to install, and what observable triggers indicate readiness to shift interaction modes (e.g., move from Collaboration to X-as-a-Service once an interface has stabilized for two sprints).

Uses the topology evolution and sensing mechanisms from Chapter 8 of Team Topologies, including the characteristic triggers for interaction mode transitions and team type promotions that Skelton and Pais identify across their case studies.
Tested

Graded before it shipped.

Every skill is scored against independent scenarios for methodology fidelity before it goes live — not vibes, a rubric.

What it produces
OutputD-01

Topology Diagnosis Map

A structured representation of your current team structure with each team classified by type, cognitive load level assessed (low/medium/high), and interaction mode labeled for every significant team pair. Anti-patterns are called out explicitly using Team Topologies terminology — not generic labels.

OutputD-02

Target Topology Design

The recommended future-state team structure: team type assignments, scope adjustments, new team formations or splits required, and specified interaction modes for every team pair — with rationale grounded in cognitive load analysis and Conway's Law alignment.

OutputD-03

Team API Document

A Team API template for any specific team covering code ownership boundaries, documentation location and standards, versioning practices, available communication channels, office hours, and working agreements — the interface contract that enables other teams to consume this team's output predictably.

OutputD-04

Topology Evolution Roadmap

A phased migration plan from current to target topology: which structural changes happen in which order, what sensing mechanisms to put in place, and what triggers indicate when to shift interaction modes — for example, when a Collaboration engagement is complete and the teams should transition to X-as-a-Service.

OutputD-05

Interaction Mode Matrix

A grid showing every significant team pair and their recommended interaction mode (Collaboration / X-as-a-Service / Facilitating), with rationale for each assignment, expected duration for time-bounded Collaboration engagements, and the transition trigger for when the mode should change.

The source

Grounded in the original work.

Every answer traces back to a real source and the practitioner who wrote it — not a secondhand summary. Here is the source of record.

Source authorA-01

Matthew Skelton & Manuel Pais

Matthew Skelton and Manuel Pais are the co-creators of the Team Topologies framework and co-founders of TeamTopologies.com, a consultancy and training organization. Their 2019 book 'Team Topologies: Organizing Business and Technology Teams for Fast Flow,' published by IT Revolution Press, has become a standard reference for engineering organization design and is applied by teams at enterprises and high-growth technology companies worldwide. Pais is also the DevOps Lead Editor at InfoQ.

Status · Inspired by Matthew Skelton & Manuel Pais’s work — not yet claimed. Are you Matthew Skelton & Manuel Pais?
Primary sourceS-01

Team Topologies: Organizing Business and Technology Teams for Fast Flow (IT Revolution Press, 2019)

by Matthew Skelton & Manuel Pais

Co-founders of TeamTopologies.com; authors of the 2019 IT Revolution Press book adopted by engineering leaders across global enterprises; Pais is DevOps Lead Editor at InfoQ.

Read the original ↗
Citationteamtopologies.com
In the build queue

Be first to run it.

Team Topologies is being built right now. Leave your email and we’ll tell you the moment it goes live.

Notify meEmail
At launchI lead engineering for a 120-person org and our product squads are constantly blocked waiting on our Platform team. Can you help me assess our current topology and figure out whether we have the wrong team structure, the wrong interaction modes, or both?