Aug 9 – Sep 8, 2026 · 6 items
The AI week, for a data analyst
A real edition, written for: Data Analyst
The one thing
The models moved from answering questions to operating spreadsheets and browsers, which is the first change that touches the manual steps in your reporting cycle.
OpenAI's GPT-6 Astra 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, and this version operates spreadsheets and browsers directly. That is closer to the reporting steps you do by hand.
- Rollout starts with gated enterprise access, so your firm's ChatGPT tier decides when you see it
- It targets multi-step work like pulling figures and producing a document, not single answers
- OpenAI left out its own real-work benchmark, so treat the claims as untested on your data
Try this
Take one recurring report and write out every manual step, then test which ones ChatGPT can already do.
Paste this into your AI tool
Here are the manual steps I take each month to produce a reporting dashboard: [list your steps, e.g. export from SQL, clean in Excel, write commentary]. For each step, tell me whether a chat AI can do it today, whether it needs to operate software directly, or whether it must stay manual. Be blunt about which ones you cannot do.
Nvidia is buying Hugging Face for just under $13bn
Nvidia is buying Hugging Face, the open model hosting platform, for $12.9 billion. Jensen Huang and Clement Delangue confirmed the deal on CNBC. Delangue said all three founders and the whole team will join Nvidia. He said Hugging Face will run as an independent, neutral platform inside Nvidia. Huang said Nvidia compute will not be required to build or deploy through the platform. He confirmed a roughly $1 billion retention plan for staff, which Delangue declined to discuss. Huang said other bidders were involved but would not name them. Delangue said a summer cyberattack, in which unreleased models attacked the company, pushed the decision. He argued open weights let defenders match attackers. Huang cited a CrowdStrike partnership using Nvidia's Nemotron open models.
Why it matters for you
Hugging Face is where open models and datasets live. If your firm ever hosts a model internally, the ownership of that hub now matters.
- Nvidia says compute choice stays open, but analysts advise watching for tighter coupling later
- The deal is expected to close in the first half of 2027, subject to approval
- For a Swiss financial firm, vendor concentration is the question your risk people will ask
Try this
Ask whoever owns your data platform whether any internal model or dataset comes via Hugging Face.
Sources
Gartner: only a fifth of large firms have scaled AI beyond one team
Gartner found that only 22% of large organisations have scaled AI across several business units. It surveyed more than 1,300 leaders at firms with over $50 million in revenue, between January and April. Spending plans are undented, with 85% of technology leaders raising AI budgets next year. About 11% of respondents could not say what they spent on AI in 2025. Gartner's Tina Nunno warned that weak measurement tied to business outcomes wastes resources. Firms that track returns continuously and shut down weak projects reported gains on 81% of initiatives. Popular uses such as cybersecurity, threat detection and IT service desk automation often return less. The best returns came from IT asset and cost optimisation, synthetic data generation, and automated code generation. Separate reports from Infosys and Deloitte found similar gaps in measurement and readiness.
Why it matters for you
You build the dashboards that show whether spending works. AI spend is now the thing nobody can measure.
- A tenth of surveyed firms could not say what they spent on AI last year
- Firms that tracked returns continuously and killed weak projects saw far better outcomes
- Automated code generation and cost optimisation returned more than the popular security use cases
Try this
Sketch a one-page AI spend and outcome view using whatever cost data your finance team already has.
Paste this into your AI tool
I am a data analyst at a financial services firm. Design a simple one-page dashboard that tracks AI tool spend against business outcomes. List the specific metrics, where each one likely comes from, and the two or three that are hardest to source honestly. Keep it to things a mid-size company can actually collect.
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
This is a costing method, not a product. It is the shape of analysis you could run on your own firm's AI usage.
- The levers were routing each task to the cheapest capable model and capping session length
- Showing engineers the running cost in their terminal changed behaviour without a policy
- Cost per session, not total spend, is the metric that made the story readable
Try this
Model AI cost per completed task for one team, using per-session cost rather than total spend.
Paste this into your AI tool
Help me build a cost model for AI tool usage. The unit should be cost per completed task, not total monthly spend. Here is what I know: [number of users, monthly bill, rough tasks per week]. Show the formula, list what data I still need to collect, and name the assumption most likely to make the model wrong.
Sources
Microsoft's new transcription model drops audio costs to a dime an hour
Microsoft AI released MAI-Transcribe-2, a speech-to-text model, on Thursday. The launch price is 10 cents per hour of audio, called an early-bird rate. Microsoft has not named an end date or a standard price. The model covers 60 languages and handles noisy, overlapping real-world audio. It labels who is speaking, timestamps each word and accepts custom word lists. A verbatim mode keeps filler words for legal and compliance use. It also follows conversations that switch language mid-sentence. Microsoft claims first place on the FLEURS multilingual benchmark and second on Artificial Analysis. It says the model runs five to ten times faster than rivals from OpenAI, Google and ElevenLabs. The announcement says nothing about real-time transcription, speaker-labelling accuracy or data retention.
Why it matters for you
Meeting audio is data nobody at your firm analyses because transcription was expensive. That constraint just moved.
- It labels speakers, timestamps every word and handles language switching mid-sentence
- Swiss multilingual audio is exactly the case that broke older 43-language models
- Access runs through Microsoft Foundry, so this is a platform-team request rather than a self-serve trial
Try this
Ask your platform team whether Foundry is available, and name one recurring meeting worth transcribing.
Anthropic cut the cost of re-reading stored context by three quarters
Anthropic released Claude Fable 5.1, its most capable generally available model for coding and complex knowledge work. It is available to Pro, Max, Team and Enterprise subscribers and through the Claude API, Amazon Web Services, Google Cloud and Microsoft Foundry. Use requires 30-day data retention by default, and questions touching biology or cybersecurity are automatically answered by weaker Opus models instead.
Why it matters for you
Long analyses that re-read the same big dataset were priced out before. The repeat-read cost was the reason.
- Base input and output rates did not move, so only context-heavy work gets cheaper
- Use requires 30-day data retention by default, which your compliance team will ask about
- This changes the business case for AI over large document sets, not for one-off questions
Try this
Re-cost one long analysis job you dropped last year on the assumption AI was too expensive.
Paste this into your AI tool
I want to re-cost an analysis task I ruled out as too expensive. The task is: [describe it, e.g. reading 200 quarterly reports and extracting figures]. Walk me through how AI token costs actually work for this, where the cost concentrates, and what would make the bill much higher than my estimate.
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.
After any answer you are about to rely on, one line makes the model turn on itself and point at the part most likely to be wrong — usually the bit you were going to trust.
Five minutes to set up
Name the single weakest link in what you just told me, and the assumption you smuggled in without stating it. Be specific, not general.
- Use it right after an answer you plan to act on — a recipe, a route, a purchase, a plan
- Run it in ChatGPT in the same chat, so it has the answer in front of it
- Watch for a vague reply like 'individual results vary' and push back: ask which exact sentence it least stands behind
- When it names something real, ask it to rewrite only that part rather than the whole answer
Yours arrives Thursday.
This one was written for a data analyst. Tell us what you do and the next one is written for you — same news, your job, once a week.