# moon-walsh-mckinsey-rewiring-software-delivery-agentic-era-2026-05-29

## Veille

Short article (**8 pages**, 4 exhibits) from **McKinsey Technology**, authored by **Jared Moon** (senior partner, London), **Rory Walsh** and **Vito Di Leo** (partners, Dublin and Zurich), with **Adam Thelwall**, published in **May 2026**. The authors report productivity gains of **×3 to ×5** with a **60% reduction** in team size at several companies, and attribute them to redesigning the operating model, not to agent deployment alone. Four levers: (1) the **24-hour sprint**, with a night shift of agents (16h) and a human day shift (8h); (2) automating the **handoffs** between requirements, design and code, which they say account for **70% of tech spend** against 30% for deployment; (3) a **knowledge infrastructure** of domain-specific graphs, queried by an agent « bibliothécaire »; (4) resizing teams and the portfolio. On the graph: *« this should not begin with a grand, top-down ontology effort »*. The team figures are presented as illustrative, drawn from the firm's "experience and observation." Extends McKinsey's November 2025 presentation on agent-driven software delivery ([[harrison-maniar-mckinsey-reshaping-software-delivery-agents-2025-11-23]]).

## Titre Article

Rewiring software delivery for the agentic era

## Date

2026-05-29

## URL

https://www.mckinsey.com/capabilities/tech-and-ai/our-insights/rewiring-software-delivery-for-the-agentic-era

## Keywords

McKinsey, agentic SDLC, operating model, 24-hour sprint, night shift, factory of agents, agent squads, handoffs, machine-readable artifacts, graphe de connaissance, agent bibliothécaire, organizational memory, traceability, policy as code, outer loop, team size reduction, pods, reskilling, capacity reallocation, productivity ×3-×5

## Authors

Jared Moon (senior partner), Rory Walsh, Vito Di Leo (partners), avec Adam Thelwall (associate partner) — McKinsey & Company

## Ton

**Profile**: a consulting-firm note aimed at technology leaders (CIOs, CTOs, delivery heads). Short format in four parallel blocks, each closed by a bolded « Main takeaway », and backed by an illustrative exhibit.

**Style**: a narrative opening (a product owner discovering at 9 a.m. the work agents did overnight: *« No one worked late. AI agents did. »*), followed by a prescriptive, imperative register (« Treat your knowledge architecture as strategic infrastructure », « Resize teams »). The vocabulary is that of industrialization: *factory of agents*, *night shift / day shift*, *pipeline*, *review gates*. The figures are highlighted in the margin (×3 to ×5, 60%), with their source given in a one-line note (« McKinsey experience and observation across multiple companies »); only one external source is cited (Boundev AI budget guide for the 70/30 split).

**Authority**: that of the firm and its transformation practice; no named client case, no detailed measurement methodology. **Target audience**: decision-makers weighing team size and portfolio allocation.

## Pense-betes

- **Daily sprint**: at night (16h), the factory of agents enriches requirements, checks architecture, designs, codes, tests and writes a report; during the day (8h), humans run the sprint review, a paired *spec-and-code working session* on critical paths, cross-functional sync (legal, compliance, design), refactoring weak code, refining guardrails (context, skills, prompts, workflows) and planning the next night.
- **Four prerequisites** for the 24-hour sprint: a clear vision of what to build (roadmap, standard); a **standardized and consistent** technical environment (common frameworks, modular architecture); a **requirements → code** path structured identically across projects; the **same stakeholders** across the whole value chain.
- **70/30**: requirements, design, code and testing account for ~70% of tech spend; deployment ~30%. Traditional CI/CD automates only the latter share.
- Exhibit 2: a pipeline of *agent squads* (process flow → enriched requirements → high- and low-level design → validation and transformation microservices → unit-testing squad), with humans **reviewing at the end**. Codified inputs: interface specs, technical guardrails, non-functional requirements, rules catalogue, *YAML rules framework*.
- **Graphe de connaissance par domaine**: links customer feedback, architecture decisions, design docs, tickets, GitHub activity, incidents, summarized compliance rules. Onboarding example: SharePoint (surveys), observability (drop-off points), Jira (planned stories), GitHub (existing services), SMEs' tacit knowledge.
- Claimed gains from the graph: questions that used to require **weeks of interviews** with SMEs handled **in minutes**; every decision traceable back to its source (a deprioritization linked to usage data); faster upskilling.
- **Organic** approach, around priority domains and ongoing programs, not a top-down ontology. Related: [[talisman-modern-data-101-ontology-pipeline-refresh-2026-05-04]].
- Exhibit 4 (illustrative): **~100 FTEs, 200 person-years, 24 months, 10 teams of 8-12** (PO, BA, tech lead, devs, testers) → **~60 FTEs, ~100 person-years, 18 months, 16 teams of 3-4** (PO, tech lead, *AI-enabled engineer*). That is ~50% less effort and ~60% smaller average team size.
- Three priorities for capturing the value: **reskilling** (a central team maintains the agent factories; engineers develop judgment, code review and oversight); integrating **outer loop** roles (risk, legal, testing, procurement), with *policy as code* controls built in by design; **reallocating the freed-up capacity** (roadmaps, modernization, new products).
- Starting point: the few **large programs** that concentrate most of the spend (legacy modernization, brownfield rewrites, launches).

## RésuméDe400mots

Jared Moon, Rory Walsh and Vito Di Leo, with Adam Thelwall, all McKinsey partners, describe in May 2026 how agentic AI is transforming software delivery, which they present as the leading edge of a broader shift in the delivery model. The article opens with a scene: at 9 a.m., a product owner finds that overnight a feature has moved from structured requirements to tested code, with edge cases flagged and architecture dependencies validated. No one worked late; the agents did. According to the authors, several companies are thereby achieving productivity gains of three to five times, with teams reduced by 60%, provided the operating model is redesigned for round-the-clock human-agent collaboration.

First lever: the daily sprint. At night, a factory of agents carries out the structured work; during the day, humans review the outputs, resolve ambiguities, reinforce guardrails and align stakeholders. Their role becomes overseeing the system that produces the artifacts. This presupposes a clear product vision, a standardized technical environment, a uniform requirements-to-code path and stable stakeholders.

Second lever: eliminating human handoffs. CI/CD mainly automates testing and deployment, roughly 30% of spend; the remaining 70%, from requirements to code, stays manual and interpretive. Codifying specs, non-functional requirements, guardrails and sequence diagrams into machine-readable formats lets the pipeline run end to end in a few hours, with humans stepping in only at defined review points.

Third lever: a knowledge infrastructure. Domain-specific graphes de connaissance serve as agents' memory across the entire software lifecycle, linking customer feedback, architecture decisions, tickets, GitHub activity and incidents. An agent « bibliothécaire » answers in minutes what used to require weeks of interviews with experts, and makes every decision traceable. The authors advise against a top-down ontology: the graph should grow organically around priority domains.

Fourth lever: resizing teams and the portfolio. Teams of eight to twelve people give way to pods of three or four skilled profiles supervising the agents; the illustrative example goes from 200 to 100 person-years and from 24 to 18 months. This requires reskilling, integrating control functions (risk, legal, procurement) via policy as code, and deliberately reallocating the freed-up capacity. The authors recommend starting with the large programs that concentrate spend, and expect human roles to concentrate on architecture, product judgment and systems design.

## GrapheDeConnaissance

- Jared Moon —travaille_chez→ McKinsey (ORGANISATION, 0.98)
- Rory Walsh —travaille_chez→ McKinsey (ORGANISATION, 0.98)
- Vito Di Leo —travaille_chez→ McKinsey (ORGANISATION, 0.98)
- McKinsey —publie→ Rewiring software delivery for the agentic era (DOCUMENT, 0.98)
- Rewiring software delivery for the agentic era —mesure→ "×3 à ×5 de productivité, -60 % de taille d'équipe" (MESURE, 0.85)
- Rewiring software delivery for the agentic era —affirme_que→ "Les gains viennent de la refonte du modèle opératoire, pas du seul déploiement d'agents" (AFFIRMATION, 0.92)
- Sprint de 24 heures —utilise→ usine d'agents (CONCEPT, 0.92)
- Sprint de 24 heures —remplace→ sprint de deux semaines (METHODOLOGIE, 0.85)
- Sprint de 24 heures —est_basé_sur→ architecture standardisée et modulaire (CONCEPT, 0.88)
- Rewiring software delivery for the agentic era —mesure→ "Exigences à tests : ~70 % de la dépense tech ; déploiement : ~30 %" (MESURE, 0.85)
- CI/CD —s_applique_à→ tests et déploiement (CONCEPT, 0.9)
- Artefacts lisibles par machine —réduit→ passages de relais humains (CONCEPT, 0.9)
- Graphe de connaissance —permet→ autonomie des agents (CONCEPT, 0.9)
- Graphe de connaissance —fait_partie_de→ SDLC (METHODOLOGIE, 0.85)
- Agent bibliothécaire —utilise→ Graphe de connaissance (TECHNOLOGIE, 0.9)
- Agent bibliothécaire —améliore→ "Questions d'experts traitées en minutes au lieu de semaines" (MESURE, 0.82)
- McKinsey —s_oppose_à→ ontologie descendante (CONCEPT, 0.88)
- Rewiring software delivery for the agentic era —mesure→ "~100 ETP / 200 années-personnes / 24 mois → ~60 ETP / ~100 années-personnes / 18 mois" (MESURE, 0.8)
- Policy as code —s_applique_à→ Outer loop (CONCEPT, 0.87)
- McKinsey —recommande→ "Commencer par les quelques grands programmes qui concentrent la dépense" (AFFIRMATION, 0.9)
- McKinsey —prédit→ "Les rôles humains se concentreront sur l'architecture, le jugement produit et la conception des systèmes" (AFFIRMATION, 0.85)

---
Canonical: https://www.thekb.eu/en/fiches/moon-walsh-mckinsey-rewiring-software-delivery-agentic-era-2026-05-29/