Aug 9 – Sep 8, 2026 · 6 items
The AI week, for a supply chain manager
A real edition, written for: Supply Chain Manager
The one thing
The models are now built to work inside spreadsheets and apps rather than chat windows, which is where your forecasting and replenishment work actually lives.
OpenAI's new model works inside spreadsheets and browsers, not just chat
OpenAI released GPT-6 Astra, a frontier model built to operate computers directly. It works across browsers, spreadsheets, websites and desktop applications. It can fill forms, update CRM records, run web research and produce documents. OpenAI says this reduces the need for hand-built connectors to each business system. President Greg Brockman told a press briefing that the company is now in the AGI era. Rollout starts Thursday for enterprise customers in the Daybreak gated program. Paid ChatGPT tiers, the OpenAI API, AWS Bedrock and Microsoft Azure follow in coming days. Brockman argued buyers should compare price per completed task rather than per token. OpenAI omitted GDPval, its own benchmark for real-world occupational work. OpenAI also paused some frontier training for about two weeks after the Hugging Face incident and tightened infrastructure controls.
Why it matters for you
You already use ChatGPT for text. This one is built to drive spreadsheets, web forms and desktop apps directly.
- Supplier portal lookups and reorder sheets are the shape of work it targets
- Enterprise access comes first through a gated programme, so your IT decides timing, not you
- Compare cost per finished task, not per token, if anyone puts a price in front of you
Try this
Ask your IT contact whether your ChatGPT tier gets Astra, and when.
Asana folds AI assistants into existing plan prices from mid-September
Asana will bundle a base level of AI tools into its work management plans from mid-September. The tools include AI Teammates and Dash, and tier prices do not change. The change applies to both new and existing customers. Asana will also recognise revenue from new AI Teammates sales as customers use their requests. CFO Aziz Megji said the shift creates a revenue headwind and pressure on gross margins in the second half. CEO Dan Rogers said AI products made up about a quarter of new recurring revenue last quarter, up from the prior quarter. Asana raised its full-year target for AI's share of new recurring revenue. It also signed its largest AI expansion deal, a three-year agreement with a Fortune 500 media company. Both executives spoke on an earnings call on Thursday.
Why it matters for you
If your team plans inbound and outbound work in Asana, base AI tools arrive without a new line item.
- Applies to existing customers too, so no renegotiation needed
- Bundled tooling is easier to get past a 251-2500 procurement process than a new vendor
- Worth checking which tier your ops boards sit on before mid-September
Try this
Check which Asana tier your logistics boards run on and what the bundled AI covers.
Sources
Uber held AI spend flat while usage grew ninefold
Uber said its total spending on AI has stayed flat since April even though weekly requests from its automated coding assistants grew more than ninefold since February. The company credits routing each task to the cheapest model that can handle it, capping how much text a session may consume, showing engineers the running cost in their terminal and extending its reuse of repeated prompts from five minutes to an hour. Uber had overrun its 2026 AI budget in the first quarter.
Why it matters for you
Uber capped AI costs with routing and spend limits, not by using less. That is a controllable-cost argument you can borrow.
- Sending each task to the cheapest capable model is the main lever they pulled
- Showing running cost to the person doing the work changed behaviour
- Useful precedent if finance asks what an AI forecasting pilot would cost per month
Try this
Draft a half-page cost-control note for any AI forecasting pilot you propose this quarter.
Paste this into your AI tool
I manage inbound and outbound logistics, supplier relationships and inventory planning at a mid-sized company. Write a half-page internal note proposing a small AI pilot to improve demand forecasting and reduce stockouts. Include: the problem in plain terms, what the pilot would do, four specific cost controls (cheapest-capable-model routing, a monthly spend ceiling, visible per-task cost, limits on how much text each run consumes), and how success would be measured. Here is context on our current forecasting process: [describe in two or three lines how you forecast today]. Keep it under 400 words, no jargon.
Sources
Criminals talked a coding assistant into helping breach seven companies
Security firms Gambit Security and CloudSek reported that a Russian-speaking ransomware group called Aur0ra used the Cursor coding assistant to help break into a Belgian chemical maker and at least six other companies. The hackers got around the tool's safety refusals by claiming the intrusions were a test, and researchers found the evidence on a server the group left exposed. Reuters could not establish how much of each break-in the AI actually enabled.
Why it matters for you
Attackers bypassed an AI tool's safety refusals by claiming the intrusion was a test. One victim was a chemical manufacturer.
- Your suppliers run these tools too, and their breach becomes your delivery problem
- A refusal built into a tool is not a control your auditors can rely on
- Worth a line in supplier onboarding about AI tool use and incident notification
Try this
Add an AI-tool and incident-notification question to your next supplier review.
Paste this into your AI tool
Draft five short questions to add to a supplier security questionnaire, covering the supplier's use of AI coding and agent tools. Cover: which tools staff use, whether agents run with production access, sandboxing, how AI-related incidents are detected, and notification timelines to customers. I am a supply chain manager, not a security specialist, so keep the wording plain enough that a purchasing colleague can ask them in a call.
Salesforce data now reachable from inside Claude, with a warning attached
Salesforce and Anthropic announced Claudeforce, which connects Sales Cloud data, workflows and security controls into Claude Cowork through Salesforce-hosted connector servers. Sales teams can use 37 packaged sales skills from inside Claude rather than moving through Salesforce screens. Forrester advises regulated firms with compliance-driven sales processes to wait, and notes the two products must be bought under separate contracts with usage-based costs that are hard to predict.
Why it matters for you
Sales teams can pull CRM data into a chat assistant. Forrester tells regulated firms to wait, and costs are hard to predict.
- If your demand signals come from Sales Cloud, this changes who touches that data
- Two separate contracts with usage-based pricing makes the bill hard to forecast
- Ask commercial whether pipeline data leaving Salesforce screens has been reviewed
Try this
Ask your sales counterpart whether they are piloting this, and how pipeline data is governed.
Sources
Meta cancelled a layoff round after AI agents underdelivered
Meta cancelled the second wave of layoffs planned under an internal reorganisation that would have shifted much of the daily work of thousands of employees onto AI systems overseen by small expert teams. The company still cut about a tenth of its staff in May, but internal measures showed autonomous AI agents were not delivering the expected productivity, while technical and security incidents rose. Zuckerberg told staff in July that the pace of AI agent progress had been misjudged.
Why it matters for you
Meta shifted routine work onto AI agents and saw incidents rise and fix times lengthen. It reversed the plan.
- Useful counterweight when anyone suggests agents can run replenishment unsupervised
- The failure mode was quality and incident recovery, not capability on paper
- Argues for keeping a human check on any automated reorder or forecast output
Try this
Write down which forecasting decisions you would never let run without a human check.
Paste this into your AI tool
Help me draft a short human-in-the-loop policy for AI use in inventory planning. List which decisions should always require a person to approve before action, which can be automated with review after the fact, and which are safe to fully automate. My scope: [describe your replenishment and forecasting decisions in two or three lines]. For each item, give one sentence on what goes wrong if the human check is removed. Output as a simple table.
Sources
Build this
Every week, one small thing to build with AI in something you actually care about. No work in it. Five minutes to set up, and worth keeping if it earns a second run.
Twenty day trips reachable from where you live that you have never heard of, cut down to four with a stated reason for each rejection.
Five minutes to set up
I live near [town or city] and travel by [car, train, or both]. Find twenty day-trip destinations within about two hours that a local would not name first — small museums, walks, workshops, markets, odd landmarks. Then cut to four, and for each of the sixteen you dropped give a one-line reason. Finally, defend the one you ranked last: say what would have to be true about my preferences for it to be first.
- Have one line ready on the season and whether you want indoors or outdoors.
- Run it in ChatGPT and ask it to search rather than answer from memory.
- Check that each of the four still exists and is open before you plan anything.
- To rerun for a fresh set, tell it to exclude all twenty from last time.
Yours arrives Thursday.
This one was written for a supply chain manager. Tell us what you do and the next one is written for you — same news, your job, once a week.