What changed, and what did not
The question for this publication's readers is whether AI is showing up in corporate earnings as a measurable productivity lever. The Q3 2026 sources in our dossier give a split answer. The measurable productivity gains come from conventional programs: headcount reductions, turnaround spending, procurement and volume. The agentic-AI case rests on a software vendor's own announcement. We lay out both so you can see the gap.
Dow: the biggest self-help program in the dossier has no AI in it
On Dow's Q3 call, finance chief Jeffrey Tate said: "we remain focused on our commitment to implement approximately $1.3 billion in net benefits this year as well as $1.3 billion in 2027."1 Sequential self-help improvement in the quarter exceeded $150 million.1 The call summary also records the CEO, named there as Karen Carter, saying the Transform to Outperform program is on track for "approximately $700 million in growth and productivity improvements this year." The program includes the reduction of approximately 4,500 Dow roles, about 70% of which had been implemented.1 Tate also said selling, general and administrative costs were down: "our SARD was actually down $130 million."1
The program is a headcount-and-cost program. The excerpt we hold does not credit automation or AI for any of it, and we do not claim otherwise.
Why Dow needs it is visible in the SEC-checked numbers. Revenue fell from $44.622 billion in fiscal 2023 to $42.964 billion in 2024 and $39.968 billion in 2025.2 Operating cash flow fell from $5.196 billion to $2.914 billion and then $1.032 billion over the same three years.2 Earnings per share swung from $1.57 in 2024 to a loss of $3.70 in 2025.2 Cost of revenue dropped to $37.435 billion in 2025, but revenue fell faster.2 Tate described the environment this way: "Persistently high energy costs continue to challenge industry economics."1
The recovery signals are modest and specific. Net sales were up 13% year over year, with growth in every operating segment, and operating EBITDA was above $1.5 billion.1 Q3 free cash flow was $1.1 billion.1 For the fourth quarter, Tate expects seasonality "largely offset by more than $100 million of lower sequential turnaround spending," and polyethylene integrated margins up about $0.02 per pound.1
On the balance sheet, Tate cited "approximately $14 billion of liquidity and no substantive debt maturities until 2029."1 That is roughly 35% of Dow's entire 2025 revenue of $39.968 billion (our arithmetic on the checked figure).2 Reported cash was $4.609 billion in Q3 2025 and $4.11 billion in Q1 2026.2 The remainder of the $14 billion is not itemised in our dossier. We do not know how much of it is committed credit lines.
Union Pacific: fuel is the cost, merger cash is the plan
Union Pacific reported net income of $2.1 billion and adjusted earnings per share up 15% to $3.53 on 6% volume growth. Its CEO, named in the summary as Vincenzo Vena, noted: "Higher fuel prices were a big cost driver."3 CFO Jennifer Hamann put numbers on it: "Fuel expense grew 69% on a 64% increase in average fuel price," which she placed at $2.56 to $4.21 per gallon.3 The company lifted 2026 capital spending to $3.5 billion and affirmed high-single-digit reported EPS growth.3 Hamann also said: "We will continue growing our cash reserves in anticipation of closing the Norfolk Southern merger in the second half of 2027."3
Rio Tinto: a productivity number worth reading closely
Rio Tinto reported $870 million of productivity benefits banked by the end of June, against a $1.8 billion year-end target.4 Peter Cunningham said the breakdown includes about $530 million in cost savings, with the rest coming from volume improvements.4 So the headline figure mixes cost-cutting with simply producing more. That is a useful reminder when any company announces a single productivity number.
The AI evidence: a vendor announcement, not a result
The one AI-specific source in our dossier is SAP's October 6, 2026 announcement from SAP Connect. SAP said it "showcased the Autonomous Enterprise in action," with Joule Work, Joule Assistants and the SAP Business AI Platform. CEO Christian Klein said: "This is a breakthrough moment for business AI," and "The Autonomous Enterprise has arrived." The release names Novartis, Morgan Foods and Nestlé among customers.5
It is a press release, so it is a statement of intent and positioning. The excerpt we hold contains no measured savings, no deployment sizes and no cost figures. Nothing in the Dow, Union Pacific or Rio Tinto material links agentic software to the savings those companies reported. The narrative behind this piece describes early AI and agentic adoption as a longer-term productivity lever. On this evidence that describes expectation, not booked results.
How much to trust each source
Our measured reliability tracking shows that, for the October 22 earnings-call summaries covering Dow and Union Pacific, 60% of 1,084 checked claims held up.1,3 For the SAP and Rio Tinto items, 52% of 6,050 checked claims held up.4,5 Neither rate is high. That is why we lean on the SEC-checked Dow figures and on direct quotes attributed to named speakers. The Rio Tinto and SAP items should be read as company claims we have not independently confirmed.
What to watch
- Whether Dow's 2027 target of $1.3 billion in net benefits is delivered, and whether any AI or automation is credited for it. So far none is.1
- Rio Tinto's progress from $870 million toward $1.8 billion, and whether the volume-versus-cost split of about $530 million in cost savings shifts.4
- Whether SAP's customers publish cost or time figures. Until they do, the AI case stays unquantified.5
- Union Pacific's cash build ahead of the Norfolk Southern close in the second half of 2027.3
- The FDA's target decision date of November 14, 2026 for Summit Therapeutics' ivonescimab application.6