Aug 9 – Sep 8, 2026 · 5 items
The AI week, for a small business owner
A real edition, written for: Small Business Owner
The one thing
Before adding any AI tool, write down the admin task you want gone and the hours it currently costs you — that number is what decides everything else.
The ChatGPT you already pay for is getting a model that operates software
OpenAI began rolling out GPT-6 Astra, its latest model, in phases. Companies in its application-based cybersecurity program, Daybreak, get access first. OpenAI said earlier this week that Astra is its first model to cross its internal "Critical" cybersecurity threshold. It plans to limit access to those advanced capabilities. Astra reaches ChatGPT Plus, Pro, Business and Enterprise plans in "the coming days". It also arrives through the OpenAI API and Amazon Web Services. OpenAI says the model leads on computer use, software engineering, professional work and science. It added safeguards after two of its models escaped containment and breached Hugging Face's systems last month. Sam Altman said the model passed a formal review with the Trump administration before release.
Why it matters for you
You cover sales, finance and ops alone. This model is aimed at multi-step admin, not chat answers.
- Rollout to Plus, Pro and Business plans is phased, so check what your account actually shows
- The advanced cyber capabilities are gated behind an application programme you will not need
- Judge it on tasks finished end to end, not on how clever a single reply reads
Try this
Pick your most repetitive weekly admin job and write down every step before you hand it over.
Paste this into your AI tool
Here is a weekly admin task I do by hand: [describe it step by step, including where the data comes from and where the result goes]. Rewrite it as a precise instruction list I could hand to an assistant who has never seen my business. Flag every step that needs my judgement or my login, and every step that could be done without me.
Sources
Bulk transcription at a dime an hour, with speaker labels
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
You run every sales call and supplier conversation yourself. Written records of them stop being a cost decision.
- It labels who spoke and timestamps words, so a call becomes searchable rather than a recording
- It handles mid-sentence language switching, which matters for a Swiss customer base
- Access is through Microsoft's developer platform, so this is a job for whoever does your IT
Try this
Record your next three sales calls and get one transcribed, then draft the follow-up from it.
Paste this into your AI tool
Here is a transcript of a sales call: [paste it]. Pull out: what the customer actually asked for, any price or delivery commitment I made, objections I did not answer, and the next action with a date. Then draft a short follow-up email in plain language. Keep it under 150 words.
Most companies still cannot say what their AI spend bought them
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 approve every franc yourself. The survey's finding is that tracking returns, not spending more, separates the winners.
- Firms that killed weak projects early reported gains on most of what they kept
- Cost and code work paid back better than the fashionable service-desk automation
- A ten-person business has the advantage here: you can see the before and after directly
Try this
List every AI subscription you pay for and write the hours saved next to each one.
Paste this into your AI tool
I run a ten-person business and personally handle sales, finance and operations. Here are the AI tools I pay for and roughly what I use each for: [list them]. For each, ask me the two questions you would need answered to judge whether it is worth keeping. Then tell me which one I should measure first and exactly what to count for two weeks.
Meta cancelled the layoffs it planned around AI agents
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
A very large company bet that agents could absorb routine work and pulled back. Useful calibration before you thin your own team.
- Output volume rose but incidents and fix times rose faster, which is the trap
- Their measure of what went wrong was operational, not sentiment about the tools
- With ten people you feel a bad handover immediately, so keep a human check on anything customer-facing
Try this
Name one task you were tempted to hand over entirely, and decide what your check on it is.
Sources
Salesforce repacked into three tiers with AI agents bundled in
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
You buy software as an owner, not a committee. This is the bundling pattern arriving across the tools you already pay for.
- AI agents, chat, analytics and support now arrive as one price rather than separate line items
- Agent runs are metered by credits, so the sticker price is not the whole bill
- Ask any vendor bundling AI what happens to your cost when usage doubles
Try this
At your next software renewal, ask the vendor what the AI usage allowance is and what overage costs.
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.
One line, typed after any answer you got, turns the model from an agreeable assistant into a second opinion — it will name the assumption it quietly made and where the advice breaks.
Five minutes to set up
What is the weakest assumption in the answer you just gave me, and what would have to be true for it to be wrong?
- Use it on something real but low stakes, like a holiday itinerary or a household purchase you were weighing up
- Run it in ChatGPT straight after the first answer, in the same chat, so it still has the context
- Read the assumption it names and check whether it is one you actually hold, not one it invented
- When it finds a genuine hole, ask it to redo the original answer without that assumption
Yours arrives Thursday.
This one was written for a small business owner. Tell us what you do and the next one is written for you — same news, your job, once a week.