PM Status Report - 26 October 2026

Anthropic opened a research preview of Claude Duet on 21 October, an agent that starts each day with a short brief of what needs your attention and asks before it sends, pays, signs or deletes anything. OpenAI made its Decisions API generally available the next day, returning a fixed answer to each question you set, and Google let Sheets users question a full BigQuery data warehouse in plain words.

PHP, which runs a large share of websites and internal tools, also shipped emergency fixes for two critical flaws, one of them found with help from an OpenAI model.

What’s New

Anthropic opened Claude Duet (21 Oct) as a research preview, by waitlist, on Pro and Max plans in the US. Each morning it sends a brief on the most important items across your connected apps, then checks in every few hours and messages you only when something needs you. Before it sends a message, makes a payment, signs a document or deletes anything, it shows what it plans to do and waits for approval. Anthropic says it can still misjudge what’s urgent.

On 23 Oct Anthropic made chat and Cowork one product on Claude Enterprise. When a chat needs files or code, the work moves to its own cloud machine and “keeps going after someone closes their laptop”. It turns on by default for existing Enterprise organisations on 16 November.

OpenAI made its Decisions API generally available (22 Oct). Instead of a written answer, it returns a choice, a score or a probability for each question you set, up to 400 questions in one request, from text, images or structured data. Box, Datadog, Ramp and Speak trialled it before release. OpenAI also reported (19 Oct) that GPT-6.1 Sol now converts COBOL into modern languages with the original behaviour kept in most of its test cases, against about one in twenty for the model before it .

PHP released fixes on 21 Oct for two critical flaws that let an attacker run code on a web server without logging in. Researchers at the security firm Calif used an OpenAI model to find one, and Amazon saw an exploitation attempt four days before disclosure.

Google added plain-language questions about BigQuery data in Sheets (19 Oct). Gemini runs the query against the full warehouse, not just the rows in the sheet, for people with permission to query that data. Meeting recaps in Google Chat (rolling out from 16 Oct) post the summary, decisions and action items with owners into the team’s chat after a Meet call that used “Take notes for me”.

Elsewhere, Slack released Frontline (19 Oct), an add-on for Enterprise+ plans that gives deskless staff up to 10 channels for their role, sign-in by employee ID or QR code, and AI agents that answer questions on the spot. Village Hotels reports cutting issue resolution from 48 hours to under 20 minutes. Asana’s Q3 release (22 Oct) connects its 30+ prebuilt AI Teammates to Gmail, Outlook, Slack, Teams and Google Drive, and lets one automation span tasks, projects and portfolios.

What It Means for Your Projects

Start the day from the decisions waiting on you, not the inbox. On a software release, a morning brief drawn from email, the tracker and team chat can list the three approvals outstanding and the supplier who missed a date, with each chaser drafted and held until you send it.

Ask every tender response the same questions. Turn the mandatory requirements in a construction or government tender into 40 yes-or-no questions and run every response through them, so evaluators start from a compliance matrix, read the “no” answers first and score the rest themselves.

Put a plain question to the project history before setting contingency. A PMO with several years of timesheet and cost data can ask which work packages ran more than 20 per cent over on contractor hours, and bring that list to the estimate review instead of a gut figure.

Check meeting actions against the plan the same afternoon. When decisions and owners land in the project channel after a site coordination meeting, compare them with the RACI and the schedule before the day ends, and correct an owner while people still remember the conversation.

Size a legacy migration from one converted module. On a banking or public-sector replacement programme, convert one representative module and run the old and new versions against the same test cases, so the business case rests on a measured result rather than a vendor’s figure.

Where Things Stand

Most of this week’s releases hand back a short, fixed-shape answer rather than a long draft: a morning brief, a yes or no per question, a recap with owners, a figure from the warehouse. That suits project work. A fixed answer can be checked against the schedule, the contract or the RACI in minutes, where a long draft takes a full reread. What each one returns still depends on the questions someone wrote and the data someone connected, and on a project that someone is usually the PM.

Which check on your project asks the same questions every week, and who wrote those questions?

Yes, AI helped me to write this :)

Unsubscribe