transformationplaybook.aiSqwyz
AI Wire · 12 OCT 2026

McKinsey: 80% Say AI Lifts Their Productivity, 37% Report a Positive EBIT Impact

Modern data centre with rows of white server cabinets

Photo: Tony Marinescu on Unsplash

5 Oct - 12 Oct 2026
37%

of respondents report a positive EBIT impact from AI, while 80% say it has improved their individual productivity, according to McKinsey's 2026 state of AI survey.

Six per cent qualify as AI high performers. McKinsey says nearly three-quarters of those high performers substantially redesigned workflows, compared with about a quarter of other organisations. In a separate readiness survey, just 27% of leaders say their organisation is ready for an agentic future.

The week in three lines

  1. McKinsey's 2026 survey shows the gap between personal and enterprise AI value: 80% say AI has improved their individual productivity, 37% report positive EBIT impact and 6% qualify as high performers. Nearly three-quarters of those high performers substantially redesigned workflows, against about a quarter of other organisations.
  2. Three Forrester analyses point to the same limit from different sides. Product data built for filters limits what shopping agents can answer, SAP process context the customer does not control limits what agents can safely do, and AI suppliers that share dependencies limit resilience.
  3. Forrester recommends mining site search, customer service, reviews, returns and store staff for "will it", "can it" and "does it" questions, defining which details matter by category and testing assistants with those questions. Scoring product data by the share of those questions it can answer is an inference from that advice, not Forrester's metric.

Productivity, profit and readiness (McKinsey surveys, 2026)

Say AI has improved their individual productivity80%
Report positive EBIT impact from AI37%
Leaders who say their organisation is ready for an agentic future27%
AI high performers6%

Redesign and the value gap (McKinsey outcome shares, plus one Accenture plan figure)

Early-stage adopters that redesigned workflows and report enterprise-level value (McKinsey)32%
Early-stage adopters that did not redesign and report enterprise-level value (McKinsey)6%
Leaders planning end-to-end redesign with AI at the core within 12 months (Accenture survey)15%

The papers

McKinsey Quarterly5 Oct 2026Operating model

McKinsey: the AI value gap is an organisation design problem

Drawing on its 2026 state of AI survey, McKinsey reports that 80% of respondents say AI has improved their individual productivity, 37% see positive EBIT impact and 6% qualify as high performers. It argues the missing step is redesigning work and teams around outcomes, with humans and agents mapped together.

Why it matters

McKinsey's mission factory takes one business outcome at a time, and mapping which handoffs should disappear is a practical first step. Our reading, not McKinsey's: without a sponsor who can move funding and scarce talent, the programme stays a tool rollout.

Read the source
The playbookFree toolkit
McKinsey7 Oct 2026 (first on Fortune.com, 1 Oct)Value gap

McKinsey: faster steps in contact centres and coding are no template for AI value

McKinsey says AI adoption is flat at 89% of organisations while high performers hold at 6%. Among early-stage adopters, those that redesigned workflows were 5.3 times as likely to report enterprise-level value (32% versus 6%).

Why it matters

McKinsey notes the headline contact centre and coding studies came from settings that already had queues, tracking and testing in place, so a faster step elsewhere may not carry over. Before approving a tool, ask who owns the outcome end to end and who can remove a handoff.

Read the source
The playbookFree toolkit
Forrester6 Oct 2026Playbook

Product data built for filters cannot answer the questions AI shopping agents ask

Forrester argues that product data built for filters cannot answer the questions shoppers and AI agents now ask, such as whether it will fit under an airline seat when the listing lacks packed depth. It cites 89% of consumers who have used answer engines for retail using them to research products or services.

Why it matters

Returns reasons are one place Forrester says to look for the missing details. Scoring each category by the share of real customer questions its data can answer is our suggestion, not Forrester's.

Read the source
The playbookFree toolkit
Forrester7 Oct 2026Governance

Four decisions to make before SAP's agents start executing work in your systems

Forrester sets out four decisions for CIOs before SAP's agents execute work, including who bears the failure risk once pricing is tied partly to agent actions. It treats lock-in through learned process knowledge as an architectural risk, not a proven problem.

Why it matters

Agree how a completed outcome is counted before consumption pricing starts, because retries and failed runs fall on the customer. Test whether agent memory can be exported before relying on it.

Read the source
The playbookFree toolkit
Forrester7 Oct 2026AI risk

Several AI suppliers add little resilience if they share what sits underneath

Forrester argues that several AI suppliers give little resilience if they share infrastructure, testing or data. It cites cyber evaluations in which, by the testing firm Irregular's account, a scenario unintentionally allowed internet access and some agents attacked real systems.

Why it matters

For each AI use case touching stock, pricing or fulfilment, map the model, cloud, data and testing dependencies. A second supplier adds resilience only if it shares none of them.

Read the source
The playbookFree toolkit

Also published

What nobody is saying

McKinsey puts the gap between 89% adoption and 6% high performance down to redesign, and acknowledges that consulting firms often earn fees by optimising what already exists. In the same week, Accenture's own survey says only 15% of leaders plan end-to-end redesign with AI at the core within 12 months, in an announcement launching a service group of more than 2,000 specialists. The two firms agree on the diagnosis, yet neither offers outcome data separating redesign work from optimisation work, and McKinsey's 32% versus 6% comes from survey respondents and measures no intervention. Anyone buying redesign help should ask for outcome data from past redesigns.