Sep 1 – Oct 1, 2026 · 4 items
The AI week, for a CFO
A real edition, written for: CFO
The one thing
Which of these four could cut the most hours from your close this quarter?
Dots launch: always-on assistants in ChatGPT
OpenAI launched Dots, always-on AI assistants that work without constant direction. They run on GPT-6 Astra with their own cloud computer. Dots connect to over 4,000 apps via plugins. They learn user preferences and work 24/7 on tasks. Users can inspect a dot’s work on its cloud computer. Dots can also use a user’s laptop with permission. They message progress or questions in ChatGPT, Slack, or Teams. Built-in safeguards include read-only tools for background work. Users control access, permissions, and approvals via Custom Rules. Specialist dots for organizations have their own identity and access. These are in focused enterprise pilots. Microsoft Agent 365 will integrate specialist dots. Dots roll out to Pro and Business Premium users in eligible markets.
Why it matters for you
Dots run 24/7 on tasks you now do by hand in the close. That frees hours for the board pack.
- They connect to 4,000+ apps via plugins, so your ERP and spreadsheets are in scope
- You inspect their work on a cloud computer, so review stays yours
- Pro and Business Premium plans get first access
Try this
Ask your admin to enable Dots beta and assign one to draft the monthly commentary
Paste this into your AI tool
Context: I’m the CFO and I draft the monthly commentary for the board pack by hand. Task: draft this month’s commentary in the same structure and tone as last month’s. Constraints: - keep last month’s headings - list any figure you could not tie to the pasted data I'll paste: [last month's commentary] I'll paste: [this month's figures] Output: the commentary, ready to drop into the pack.
Sources
OpenAI DevDay: GPT-6 Astra and Ultrafast mode
OpenAI announced over twenty updates at DevDay 2026. It introduced Dots as always-on agents for ongoing tasks. GPT-6.1 Sol launched with near-Astra performance at one-fifth the token cost. Ultrafast mode offers up to eight times faster token generation in Codex. Private Intelligence now supports zero data retention with private safety processing. Codex can now run in the cloud or via a refreshed CLI with voice control. Code Review in ChatGPT desktop app enables automated and manual reviews. Codex Security Cloud scans repositories and prepares fixes in the cloud. Decisions API entered limited preview for real-time classification. Agents API now supports computer use for software interaction. Bedrock Managed Agents brings OpenAI agents to AWS. Plugin extensions allow sidebar integration in ChatGPT. Improved plugin creation and discovery tools were released.
Why it matters for you
GPT-6 Astra and Ultrafast mode cut the time to run your forecasting models. That shortens the close.
- Ultrafast mode offers up to 8× faster token generation in Codex
- GPT-6.1 Sol matches Astra performance at one-fifth the token cost
- Agents API now supports computer use for software interaction
Try this
Re-run your forecasting model with Ultrafast mode and compare the runtime
Paste this into your AI tool
Context: I run monthly forecasting models in spreadsheets for a manufacturing CFO. Task: re-run this month’s forecast with the same inputs and logic. Constraints: - use Ultrafast mode if available - flag any cell that breaks the prior logic I'll paste: [this month’s forecast spreadsheet] Output: the forecast, with any broken cells listed.
Sources
Thomson Reuters trains near-frontier model on small cluster
Thomson Reuters released Thomson, a family of models built by retraining open-weight Qwen models. The company calls its method continual learning, a staged retraining that adds skills without losing old ones. A team of under three dozen staff did the work on a small cluster in three months. Thomson-1.0-Large scores near the strongest proprietary models tested on an internal cross-domain suite. It leads on instruction following, general agent tasks and political neutrality. It trails on coding, maths and multilingual tasks. Thomson-1.0-Small, at 35 billion parameters, is released with open weights. Blind expert reviewers preferred Thomson's answers over rival systems in most paired comparisons. Thomson Reuters ran the evaluations itself and notes it cannot see hidden safeguards inside rival hosted models. Partners include Imperial College London, DatologyAI and Lambda.
Why it matters for you
A small team built a near-frontier model for legal and tax work. Your forecasting could follow the same path.
- Under three dozen staff did the work on a small cluster in three months
- Thomson-1.0-Large scores near the strongest proprietary models on internal tests
- Cost per run was under USD 450,000 for the final training
Try this
Ask your data team whether a staged retraining of an open model could cut your forecasting cycle
Salesforce bundles AI agents, Slack, analytics in three editions
Salesforce replaced its sales and service packaging with three new editions. Core, Advanced and Max now apply to Agentforce Sales, Service and Industries. Each tier bundles AI agents, Slack, embedded analytics, data security and the Premier Success Plan. Core lists at $195 per user per month and Advanced at $395. Max stays at $550, the same price as the earlier Agentforce 1 Edition. Existing Agentforce 1 customers can move to Max at no extra cost. Each tier carries a set allowance of Flex Credits for running agents. Industry edition pricing takes effect later in the fall. Pricing on legacy editions stays unchanged for current customers. Salesforce says Max editions will later include Headless 360 capacity.
Why it matters for you
Salesforce now sells AI agents, Slack and analytics as one edition. That simplifies buying for your size firm.
- Core, Advanced and Max tiers bundle agents, Slack, analytics, security and support
- Max stays at $550 per user per month, same as the earlier Agentforce 1 Edition
- Each tier carries a set allowance of Flex Credits for running agents
Try this
Ask your procurement team to compare the Max edition against your current Salesforce spend
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.
A shortlist of five gifts for [someone important to you], each with a reason to keep or cut it.
Five minutes to set up
Context: I need a gift for [someone important to you] and I want to find options I did not know existed. Task: find five gifts they would like, then rank them and defend the one you ranked last. Constraints: - include a reason to keep or cut each one - the last one must have a stated flaw that would have to change for it to win Output: the ranked list with one-line reasons, and the condition that would move the last one up.
- Have ready the person’s name and one thing they already like
- Run this in ChatGPT so you can inspect the reasoning behind each pick
- Check that the last entry’s flaw is real and not invented
- Re-run with a different starting point to test the ranking
Yours arrives Thursday.
This one was written for a cfo. Tell us what you do and the next one is written for you — same news, your job, once a week.