← Back to Archive · Homepage - Newsletter

Web archive · this issue of MCC AI Weekly was sent by email on Thursday, September 3, 2026.

MCC AI Weekly · Thursday, 3 September 2026

Adoption is high, impact is not — this week's gap

MCC AI Weekly — issue cover illustration

Multiple reports this week land on the same uncomfortable number: usage is nearly universal, but the share of firms seeing real profit or productivity impact remains small. The common diagnosis is sequencing — tools bolted onto unchanged workflows underperform, while redesign, coordination, and governance separate leaders from the rest. Below: the research behind that gap, plus the week's news and a practical tip.

Insights

AI attacks hit significant impact at 35% of firms, BCG finds

BCG's 2026 CISO Survey finds that 35% of organizations experienced significant impact from AI-enabled attacks in the past year. The average organization reported three significant breaches and 25 sensitive data incidents. Respondents cite data leakage as the greatest AI-related risk, ahead of other attack types. BCG argues attacks and breaches should now be treated as a baseline expectation, not an exception.

So what

BCG puts this with the CEO, who should initiate cross-functional coordination across the teams involved in cyber defense. This is not a task to leave with the CISO alone. The report frames breaches as a baseline expectation, so containment and resilience planning matter as much as prevention.

BCG →

AI fitness gap: 7.2x performance edge, and it's not linear

PwC surveyed 1,217 executives worldwide, including 81 German firms, and finds companies in the top AI Fitness quintile achieve 7.2 times higher industry-adjusted AI performance than the rest. The effect is non-linear: small fitness gains produce disproportionate performance gains. German firms score above the global median on foundations (5.8 vs. leaders' 6.9) but lag most on cross-industry use (4.6 vs. 7.1).

So what

PwC recommends aiming AI at growth and cross-industry business models, not just efficiency, since only 25% of German firms target revenue growth versus 31% of leaders. Executives should pursue strategic alliances and data-sharing consortiums over vendor-only partnerships, where German firms lag leaders most (21% vs. 41%, and 9% vs. 22%).

PwC →

Copilots alone aren't moving software productivity

McKinsey surveyed 334 product and engineering leaders and found only 25 percent of director-and-above respondents report meaningful AI acceleration, while 30 percent say productivity fell. Ninety-three percent of top-accelerating organizations had embedded AI into redesigned workflows, not layered it onto old ones. Firms that redesigned first were twice as likely to reach 20 percent-plus gains, a difference McKinsey ties to redefined roles and new verification mechanisms.

So what

Leaders should redesign workflows, roles, and verification mechanisms before scaling AI tools, not after. The article finds this sequencing, not tool choice, separates organizations reaching twofold productivity gains from those seeing output fall.

McKinsey →

Pew: a third of US adults now use chatbots for health information

Pew Research surveyed 3,488 U.S. adults and finds 34% ever use AI chatbots for health reasons, from diagnosing symptoms (25%) to understanding lab results (20%) and treatments (22%). Nearly all of these users, 95%, rate the information as at least somewhat helpful. But on sharing personal health information with chatbots, users split: 29% are very comfortable, 26% are not.

So what

Health-tech leaders building or evaluating chatbot tools should treat data-handling transparency as a design priority, not an afterthought. Pew notes some platforms already face scrutiny for storing or selling health data, or using it to train models.

Pew Research Center →

AI use hits 88%, but only 6% see it move profit

McKinsey's Global Survey, covering 1,993 respondents across 105 countries, finds regular AI use reached 88%, up from 78% a year earlier. Yet only 39% report any enterprise-level EBIT impact, and just 6% qualify as high performers where AI drives 5% or more of EBIT. McKinsey finds these high performers set growth and innovation goals alongside efficiency, and most are redesigning workflows rather than layering AI onto existing ones.

So what

Leaders should set growth or innovation objectives for AI initiatives, not efficiency alone, since McKinsey finds high performers pair the two. Prioritize redesigning workflows around AI rather than adding it to processes unchanged, the practice McKinsey ties to scaled impact.

McKinsey →

Why generative AI still isn't platformed

Kevin Boudreau of Northeastern University notes that generative AI usage has scaled fast, with roughly 2.4 billion monthly users and platforms like Cursor and Agentforce posting rapid revenue growth. But he argues this resembles an early platform ecosystem, not the settled architecture that let past general-purpose technologies like electricity reshape entire economies. That deeper transformation, he finds, has barely begun.

So what

Boudreau recommends learning faster than committing, and building assets that survive architectural change rather than betting on today's models. Invest in complements and organizational capabilities no shared AI model can replace, rather than racing to deploy the technology itself.

MIT Sloan Management Review →

Stanford HAI: classify AI agent makers as fiduciaries

Stanford HAI reports that AI agents booking travel, managing finances, or handling health data can be steered to favor developers' interests over users', with no standardized mechanism to disclose the conflict. Broad cross-domain data access, needed for agents to act effectively, also lets them infer sensitive facts users never disclosed. The brief argues developers and deployers in high-stakes domains should be classified as fiduciaries, obligated to act in users' best interest within the delegated task.

So what

Stanford HAI calls for coordinated action across technical standards bodies, federal regulators, and Congress to establish this fiduciary duty. Leaders deploying agents in financial services or healthcare should treat undisclosed conflicts of interest as a live liability, not a hypothetical one.

Stanford HAI →

Weekly Short News

Nvidia adds $440 billion in value after earnings beat

Nvidia shares rose nearly 9% on Thursday after CFO Colette Kress guided for 70% revenue growth in fiscal 2028, adding about $440 billion to its market cap. CEO Jensen Huang said demand exceeds that figure but supply is constrained. Nvidia's ACIE customer segment, covering AI clouds, industrial and enterprise clients, brought in $40.3 billion, up 138% year over year.

CNBC →

TIME names its 2026 list of 100 most influential AI figures

TIME released its 2026 TIME100 AI list, grouping 100 honorees into four categories: Leaders, Innovators, Shapers, and Thinkers. The Leaders category includes OpenAI's Sam Altman, Anthropic's Dario and Daniela Amodei, IBM's Arvind Krishna, and Amazon's Jeff Bezos. Innovators include Cursor's Michael Truell and Cerebras's Andrew Feldman; Shapers include the UN's Doreen Bogdan-Martin and Sequoia's Pat Grady and Alfred Lin.

TIME →

Judge voids Pentagon's Anthropic supply-chain risk designation

U.S. District Judge Rita Lin ruled on Thursday that the Department of Defense acted illegally when it designated Anthropic a national-security supply-chain risk. The 59-page order found the designation followed Anthropic's refusal to lift its bans on fully autonomous weapons and domestic surveillance uses of Claude. Defense Secretary Pete Hegseth had made the designation after failed talks over unrestricted military use of Anthropic's models.

Reuters →

German group files complaint seeking ban on Ray-Ban Meta glasses

German digital rights group HateAid filed a criminal complaint against Meta, Ray-Ban, Oakley and several retailers with Germany's Central Office for Combating Internet Crime. HateAid seeks a ban on sales of the Ray-Ban Meta Wayfarer, arguing the glasses violate Germany's Telecommunications, Digital Services and Data Protection Act because their camera is indistinguishable from ordinary eyewear.

Euronews →

Aschenbrenner's Situational Awareness fund liquidates equities portfolio

Situational Awareness, the roughly $20B AI-focused hedge fund founded by Leopold Aschenbrenner, sold its entire public equities portfolio to Ken Griffin's Citadel. The sale came days after reports the fund was seeking fresh capital following heavy losses in the AI selloff.

The Sequence →

AI Player of the Week

Zapier

Zapier, the workflow-automation platform, now frames itself as the governed connection point between AI assistants like Claude and ChatGPT and over 9,000 business apps. It manages authentication centrally, logs every AI action in one audit trail, and lets IT set app- and action-level access rules across teams. Enterprise clients cite hours saved and error reductions from these workflows.

Zapier →

Tip of the Week

Ask for three ranked options, not one answer

When you request a single answer, the model hides the tradeoffs it weighed to reach it. You either accept the output or start an unstructured back-and-forth to find out what else it considered. Ask instead for two or three ranked options, each with its own reasoning attached. You then compare visible alternatives, and decide, instead of approving a black box.

Found this useful? Forward it to a colleague — they can subscribe here.

You are receiving this email because you confirmed your subscription to MCC AI Weekly. Questions? Just reply to this email — it reaches us directly.

Imprint · Privacy Policy