Agentic AI in Industry: What McKinsey, Accenture, and IBM Public Reports Say

Trends · Core On Dynamics analysis briefing

Agentic AI in Industry: What McKinsey, Accenture, and IBM Public Reports Say

Scope: 2024–2026 public web and PDF

Primary hosts: mckinsey.com · accenture.com · ibm.com

This is a briefing of public reports. It is not original Core On research. No estimated figures.

Slide 1 · Definitions

How each firm names the thing

McKinsey. AI agents: foundation-model systems that act in the real world and plan/execute multi-step workflows. Agentic organization: humans plus virtual and physical agents. Sources: State of AI 2025; The agentic organization (26 Sep 2025); Seizing the agentic AI advantage.

Accenture. Autonomous complex tasks. Goal-directed. Hive / agentic architecture. Source: Hive mind (2 Nov 2024).

IBM. Agents act beyond assist. They orchestrate workflows. The control plane is watsonx Orchestrate. Sources: Agentic AI is here (12 Jun 2026); Control Plane (2 Jul 2026); Blueprint (May 2026).

Slide 2 · McKinsey survey

State of AI 2025

Fieldwork: 25 Jun–29 Jul 2025. n=1,993. 105 countries.

23%scaling agentic AI in at least one function

39%additionally experimenting with agents

62%at least experimenting with agents

≤10%scaling agents in any single business function (stated upper bound)

Regular AI use in at least one function: 88% (was 78%).
EBIT impact reported: 39%. Many of those put AI’s share of EBIT below 5%.
Scaling is often limited to one or two functions.
IT and knowledge management are relatively higher.
Tech/media/telecom and healthcare are relatively higher for scaling.

Source: McKinsey State of AI. Charts not reproduced.

Slide 3 · McKinsey analysis

The gen-AI paradox and the agentic organization

About 8 in 10 use gen AI and still show no meaningful P&L. The article’s name: the gen AI paradox.
The article’s claim, not ours: redesign workflows.
Five pillars of the agentic organization: business / operating / governance / workforce / technology and data.
Organization type: industrial age 89%. Digital 9%. Decentralized 1%.
High performers about 6%: EBIT 5%+ and significant value.
That group: workflow redesign about 2.8×. Agent scaling at least about 3× peers. Strongly agree on senior ownership about 3×.

Advanced-industry potential (Empowering advanced industries, 8 Sep 2025): $450B–$650B additional annual revenue by 2030. A 5–10% uplift. Cost-reduction potential 30–50%.

Sources: Seizing the agentic AI advantage · The agentic organization · Empowering advanced industries.

Slide 4 · Accenture findings

Hive mind and scale claims

Hive mind (Nov 2024):

1/3pivoting innovation with agentic AI

16%fully modernized AI-led processes (9% in 2023)

2.5×revenue growth vs peers, for those orgs

2.4×productivity

Gen-AI use-case scale success: 3.3×.
Newsroom 18 Mar 2025: 3× ROI when scaling industry-fitted solutions on core processes.
Agent Builder. 50+ industry agent solutions in development. Year-end goal 100+.

Read as an Accenture case claim: BMW multi-agent sales productivity 30–40%.

AI Refinery: 36% scaled a meaningful gen-AI solution. 13% report enterprise-grade value.

Sources: Hive mind · Accenture Newsroom 18 Mar 2025 · AI Refinery.

Slide 5 · Accenture platform

Hive architecture

Agentic architecture hive: Utility / Super / Orchestrator.

Trusted Agent Huddle (28 Apr 2025): A2A / MCP. Partner list is in the original.

Distiller (11 Jun 2025): lifecycle frame.

Sources: Trusted Agent Huddle · Distiller.

Slide 6 · IBM findings

Blueprint, May 2026

2,000 senior leaders. 16 countries. 17 industries.

55%developing or deploying an agentic operating model

82%functional silos block autonomy value

25%CEOs: recent AI initiatives met expected ROI

10%say agentic AI primarily drives growth in 2026

Among developers/users, expectations by end of 2027:

Ops efficiency +45%
Revenue +23%
Profitability +37%

Data/governance investment: 77%. Digital-twin / simulation plans: 72%.

Source: IBM IBV blueprint.

Slide 7 · IBM platform

Control plane and workforce

Agentic Control Plane. watsonx Orchestrate. Announced 2 Jul 2026. Generally available Jun 2026 on AWS and IBM Cloud.

Krishna’s point in the announcement: redesign how the business operates. The wording is in the source.

Workforce:

54% of CEOs are hiring AI roles that did not exist a year ago.
About one-third of the workforce to reskill in three years.
New-role examples are in the original.

Sources: Agentic Control Plane · Is your workforce ready.

Slide 8 · Cross-firm

Three columns, one message

Not more pilots. End-to-end workflow redesign, cross-functional orchestration, and governance built in.

	Value / scale reality	Barriers	Governance / architecture
McKinsey	Regular use 88%. Scaling 23%. EBIT impact 39%. About 8 in 10 show weak P&L.	Scale often 1–2 functions. ≤10% in any single function.	Five pillars. Workflow redesign. Senior ownership.
Accenture	Fully modernized 16%. Those orgs 2.5× revenue. 3× ROI when scaling industry-fitted solutions. Enterprise-grade value 13%.	Modernization remains a minority.	Hive: Utility / Super / Orchestrator. Huddle. Distiller.
IBM	55% on an agentic operating model. 10% say it primarily drives 2026 growth.	Functional silos 82%. 25% of AI initiatives met expected ROI.	watsonx Orchestrate control plane. Data/governance investment 77%.

Slide 9 · Industrial operators

Five priorities

Drawn from the briefing. No new statistics.

Do not add pilots. Redesign one workflow end to end.
Orchestrate across functions. That is the blockage IBM names.
Do not bolt on governance later. Control plane, hive, and five pillars all mean structure.
Scale in one or two functions first. Put senior ownership on it.
Design roles and reskilling with the agents. Do not drop agents onto the old org.

Slide 10 · Sources

Primary URLs

Public primary sources only. Core On Dynamics Trends briefing.