# covington-mckinsey-agentic-transformation-office-2026-09-01

## Veille

Article by **Chase Covington**, **Tomer Slaney** (partners, New York) and **David Pralong** (senior partner, Charlotte), published by the *McKinsey Transformation Practice* in **September 2026** (exact day unspecified, dated the 1st by convention); 8 pages, 2 exhibits whose content is not reproduced in the text. **(1)** Finding: the transformation office (TO) is slow to modernize its own tools; in a turnaround involving **3,000** employees, **more than 15** people produced a **300-page** report every four weeks. **(2)** Three pillars entrusted to agents: *performance intelligence*, *single source of truth*, *conversation facilitation*. **(3)** Reported gains: **35 to 40%** of time on core tasks, sometimes **70%**; a central team reduced from **ten to three** people. **(4)** Guardrails: cascading errors between agents, false alerts, traceability of each step. **(5)** Sequence: reporting first, action automation next, change management within the TO. The authors present the text as "a first look at the opportunity for TOs". The cases are not named, no measurement method is described, and two of the tools cited (**Wave**, **Lilli**) belong to McKinsey. The text follows on from [[sternfels-mckinsey-60000-people-20000-agents-officechai-2026-01-14]], where McKinsey already counts its agents among its workforce.

## Titre Article

The agentic transformation office: Redefining the economics of change

## Date

2026-09-01

## URL

https://www.mckinsey.com/capabilities/transformation/our-insights/the-agentic-transformation-office-redefining-the-economics-of-change

## Keywords

transformation office, transformation office, IA agentique, autonomous agents, McKinsey, performance intelligence, single source of truth, conversation facilitation, steering committee, SteerCo, reporting, Wave, Lilli, Cursor, guardrails, traceability, workflow orchestrator, change management, time savings

## Authors

Chase Covington et Tomer Slaney (partners, McKinsey, New York), David Pralong (senior partner, McKinsey, Charlotte).

## Ton

A consulting article addressed to executives and transformation program leaders, signed by three partners from the practice in question. Classic genre construction: an opening line ("The transformation office (TO) has a dirty secret"), a diagnosis of the traditional model, a distinction between generative AI (a personal tool) and agentic AI (an operational system), three pillars each illustrated with a quantified case, followed by guardrails, a new division of roles, and a deployment sequence. Assertive and forward-looking register ("The potential is generational"), consulting-firm vocabulary (*infostructure*, *SteerCo*, *management rhythm*). The examples are anonymized ("one recent transformation", "a recent deployment") and specify neither sector, measurement period, nor comparison point. The authors themselves describe the text as a first look and a playbook sketch. The risks section acknowledges that mitigation remains behind awareness.

## Pense-betes

- **Thesis**: the transformation office is the function where agentic AI carries the most weight, because its outputs touch nearly every workstream; automating tracking, reporting, and coordination frees up human capacity, presented as its main constraint.
- **Diagnosis of the traditional model**: rigid roles, exhausting meeting rituals, heavy reporting; fragmented data (more than 100 non-integrated documents for financial tracking in one case); parallel tools improvised by workstreams.
- **Distinction drawn**: generative AI assists an individual on a task; the agent plans, acts, and loops multi-step workflows without waiting for a prompt.
- **Pillar 1, performance intelligence**: continuous synthesis of initiative data, flagging of negative deviations, surfacing of exceptions only. Case: 30,000 metrics and 900,000 data points, agentic framework built "largely" with **Cursor** and integrated into **Lilli** Teamspace; a shift from "what happened" to "what is likely to happen".
- **Pillar 2, single source of truth**: a live view of value delivered, risks, and pending decisions; dynamic linking of datasets (Wave, ERP, HR, production) rather than a centralized data lake, a workstream the authors say took more than a year; simulation of additional funding in 2026.
- **Pillar 3, conversation facilitation**: a 5-hour SteerCo briefing cut to under 10 minutes; reports down from 12 to 3 days per month, and around 200 additional person-hours saved each month in one case.
- **Day to day**: checking invitations (the agent invites the forgotten initiative owner), preparing SteerCo decks broken down by workstream, follow-ups and data-entry requests sent without human intervention.
- **Guardrails**: a data error can propagate between connected agents; too many false alerts at first ("two steps forward and one step back"); every step an agent takes must be transparent and traceable.
- **Division of labor**: humans keep difficult escalations, stakeholder relations, and culture; ten processes automated, team reduced from one manager and nine staff to one manager and two, more than 25 days freed up.
- **Deployment sequence**: map workflows, set governance, build capabilities; reporting and insight first, cross-system actions next; reposition staff as workflow orchestrators.
- ⚠️ **Scope**: cases not named, no measurement method or comparison point; Wave and Lilli are McKinsey tools.

## RésuméDe400mots

In this September 2026 article, three partners at the McKinsey Transformation Practice argue that the transformation office, the function responsible for steering change within the company, is often one of the slowest to modernize its own tools. Hand-assembled presentations and scattered data produce late information, diverging narratives between workstreams, and parallel tools.

Two cases quantify this ceiling. In a two-year turnaround affecting 3,000 employees, more than fifteen full-time people spent half their time validating data for a report of over 300 pages produced every four weeks. In another transformation, more than 100 non-integrated documents were used for financial tracking.

The authors distinguish generative AI, which assists an individual on a task, from agentic AI, which plans, acts, and loops multi-step workflows. They describe three pillars. Performance intelligence continuously synthesizes initiative data and surfaces only exceptions; in a case involving 30,000 metrics and 900,000 data points, an agentic framework built largely with Cursor and integrated into Lilli Teamspace replaces hours of pivot tables. The single source of truth dynamically links Wave, the ERP, HR, and production tools, without a centralized data lake, and makes it possible to simulate the effect of additional funding. Conversation facilitation prepares steering committees: a five-hour analyst briefing is produced in under ten minutes, and reporting goes from twelve to three days per month in one case.

The reported gains reach 35 to 40% of time on core tasks, sometimes 70% or more. In one deployment, automating ten processes reduced the central team from one manager and nine staff to one manager and two, freeing up more than 25 days of capacity.

The article flags risks: a data error can propagate between connected agents, false alerts are numerous at first, and mitigation remains behind awareness. Every step an agent takes must be traceable. Escalations, stakeholder relations, and culture remain human.

Deployment requires mapping workflows, setting governance rules, and repositioning staff as workflow orchestrators. The early movers start with reporting, then automate cross-system actions. Without the office team's trust, manual habits return. The cases are not named.

## GrapheDeConnaissance

- McKinsey —publie→ article The agentic transformation office (DOCUMENT, 0.97)
- Chase Covington —travaille_chez→ McKinsey (ORGANISATION, 0.97)
- David Pralong —travaille_chez→ McKinsey (ORGANISATION, 0.97)
- Tomer Slaney —travaille_chez→ McKinsey (ORGANISATION, 0.97)
- Bureau de transformation agentique —est_variante_de→ Bureau de transformation (CONCEPT, 0.9)
- Bureau de transformation agentique —utilise→ IA agentique (TECHNOLOGIE, 0.95)
- Performance intelligence —fait_partie_de→ Bureau de transformation agentique (CONCEPT, 0.95)
- Source unique de vérité —fait_partie_de→ Bureau de transformation agentique (CONCEPT, 0.95)
- Facilitation des conversations —fait_partie_de→ Bureau de transformation agentique (CONCEPT, 0.95)
- Performance intelligence —utilise→ Cursor (TECHNOLOGIE, 0.88)
- Performance intelligence —utilise→ Lilli (TECHNOLOGIE, 0.88)
- Source unique de vérité —utilise→ Wave (TECHNOLOGIE, 0.9)
- McKinsey —a_créé→ Wave (TECHNOLOGIE, 0.92)
- McKinsey —a_créé→ Lilli (TECHNOLOGIE, 0.92)
- McKinsey —mesure→ 35 à 40 % de temps gagné sur les tâches cœur du bureau de transformation, 70 % ou plus dans certains cas (MESURE, 0.85)
- McKinsey —mesure→ équipe centrale ramenée d'un manager et neuf collaborateurs à un manager et deux après automatisation de dix processus (MESURE, 0.85)
- McKinsey —affirme_que→ une erreur de donnée peut déclencher une réaction en chaîne entre agents connectés (AFFIRMATION, 0.9)
- McKinsey —recommande→ commencer par le reporting et l'insight, puis étendre à l'automatisation d'actions entre systèmes (AFFIRMATION, 0.92)
- McKinsey —affirme_que→ le succès du déploiement dépend de la conduite du changement au sein du bureau de transformation (AFFIRMATION, 0.9)

---
Canonical: https://www.thekb.eu/en/fiches/covington-mckinsey-agentic-transformation-office-2026-09-01/