Today's briefing
When the Answer Box Starts Selling Ads
OpenAI is building advertising into ChatGPT, with hooks into Shopify and HubSpot. What that changes for a shop, a school or a clinic — and what it doesn't.
The one that matters
OpenAI Starts Building an Ad Business Inside ChatGPT — and Wires It to Shopify and HubSpot
OpenAI has published its plans for advertising inside ChatGPT. The announcement describes three things: ‘Sponsored Agents’, a set of tools for marketers, and integrations with HubSpot and Shopify — the CRM and the storefront software that a very large number of small businesses in India and Canada already pay for. OpenAI has not published pricing, launch dates, or which countries see it first, and its description of how a sponsored agent actually appears to a user is thin. What is clear is the direction of travel: the company that taught millions of people to ask a chat box for a recommendation is now building a way for businesses to pay to be part of the answer.
What this actually means for you
For twenty years, being findable meant being high in a list of blue links. A list has room for ten results, which is why local search, reviews and directory listings mattered so much. An assistant does not give you a list. It gives one paragraph, maybe three named options. If advertising becomes part of how those three options are chosen, the number of businesses that can be ‘found’ for a given question shrinks — and a slot in that paragraph becomes something with a price on it.
The Shopify and HubSpot integrations are the part worth watching, because they are plumbing rather than announcement. If you run a 15-person apparel brand selling through Shopify in Bengaluru, or a Toronto services firm whose entire sales pipeline sits in HubSpot, the machinery for advertising and selling through an assistant is being built into software you already use and already know how to log into. That lowers the effort of trying it from ‘hire an agency’ to ‘tick a box’. Low-effort channels get tried, and channels that get tried get budget.
What it does not mean
It does not mean your customers are buying through ChatGPT today. OpenAI has published no numbers on how many purchases start in an assistant, and nothing in this announcement says otherwise. For a 600-student school in Pune, parents still choose on board results, fees, distance and what another parent said at the gate. For a 30-bed clinic in Surrey, patients still come from a GP referral, insurance coverage and Google Maps. None of that changed this week.
The overstatement is already arriving in your inbox. Agencies are selling ‘AI search optimisation’ or ‘GEO’ retainers — commonly ₹40,000 to ₹1,00,000 a month in India, CA$1,500 to CA$3,000 in Canada — and promising you a place in AI answers. Nobody can promise placement in an ad product whose mechanics, pricing and eligibility have not been published. If a pitch includes the word ‘guarantee’, it is a pitch, not a service.
What to do this month
- Spend twenty minutes, free. Ask ChatGPT, Gemini and Claude the five questions a customer would ask before choosing you. Write down what they say about you and about your three nearest competitors. That is your baseline.
- Fix the facts, not the marketing. Assistants repeat whatever is written about you in public. Half a day of an admin's time correcting hours, address, services, fee ranges and phone numbers on your own site, Google Business Profile and the directories you appear in is worth more than any retainer right now.
- If you are on Shopify or HubSpot, put this on the agenda with your account contact and ask what it will cost. Ask. Do not commit.
- Add one line to your enquiry form: ‘How did you hear about us?’ with an option for an AI assistant. In six months you will have real data instead of opinions.
- Sign nothing longer than three months on anything that mentions AI visibility.
The honest summary: this is a channel worth understanding before it exists, not one worth funding before it does.
Also worth knowing
-
Claude Adds Docs and Slides, and Folds Its Separate Work Mode Into One Window
Anthropic has added document and presentation creation directly inside Claude chats, with export, editing and sharing, and merged its regular chat and Cowork modes into a single window. For a business already paying for Claude, this removes a real friction point: the first draft of a staff policy, a parent circular or a monthly board deck can come out of the same conversation where you discussed it. It is a convenience upgrade rather than anything deeper, and the output still needs someone who knows your business to check it before it leaves the building.
The Verge AI ↗ -
A Coffee Shop Made a Menu Poster With AI and Got Angry Messages
A small café owner used AI to produce a menu poster and was met with a wave of hostile DMs from customers. This is the most practical story of the week for anyone running a customer-facing business: the cost of AI-generated marketing material is not the tool, it is the segment of your customers who notice and mind. Before you use it on anything public-facing, ask whether your customers are the kind who would care — for a machine shop's spec sheet, nobody will; for a café, a boutique or a design-led brand, some will.
Hacker News ↗ -
Anthropic Launches a Version of Claude Aimed at Financial Advisors, With Schwab and BlackRock
Anthropic has released Claude for Advisors, built with Schwab and BlackRock. If you run a wealth management or financial advisory practice, expect your clients and your compliance officer to start asking about this within the quarter. For everyone else, the useful signal is the pattern: vendors are now shipping industry-specific versions with named institutional partners, which means a version aimed at your sector is a matter of time, and waiting for it may beat building your own.
forbes.com ↗ -
Anthropic Says a Russian Operation Misused Claude for Espionage and Drone Work
Reporting this week describes a Russian team misusing Claude for espionage, disinformation and drone-related work, including training on scraped war footage. The point for an ordinary business is not geopolitics — it is that the same cheap, capable tools you are evaluating are available to people attacking you. Assume the phishing email that reaches your accounts clerk this year is grammatically perfect, in their language, and references a real supplier; train for that, because spotting bad English no longer works.
Euronews.com ↗ -
Mistral and Mozilla Team Up on Private, Multilingual AI in the Browser
Mistral and Mozilla have announced a partnership around private, multilingual AI browsing — details in the announcement are limited, so treat specifics with caution. The two words worth noting are ‘private’ and ‘multilingual’: for an Indian business operating across Hindi, Tamil or Marathi, or a Canadian one with bilingual obligations, an option that does not route everything through a US assistant is genuinely useful. Nothing to act on yet, but worth a look when it ships.
Hacker News ↗ -
OpenAI Publishes Guidance on Tying AI Spending to Actual Business Results
OpenAI has released material on using ChatGPT Work and Codex analytics to see who is using AI, what it costs and where training is needed. The underlying admission is the interesting part: enough companies are paying for seats without knowing whether they are used that the vendor is now shipping dashboards for it. If you have bought AI licences for staff, pull the usage numbers this month — the common finding is that a third of seats are dormant and can be cancelled.
OpenAI ↗ -
OpenAI Sets Out How It Will Report When Its Models Misbehave
OpenAI has published a framework for tracking, investigating and disclosing model misalignment, along with six reports of unexpected or concerning behaviour. This is housekeeping rather than news, but it establishes something you can use: when you next negotiate an AI contract, ask the vendor what they disclose when their system behaves badly, and how you would find out. A vendor with no answer to that question is a vendor whose failures become your problem silently.
OpenAI ↗ -
Novo Nordisk and Anthropic Partner on Drug Discovery
Novo Nordisk will work with Anthropic to apply Claude to drug discovery. Included here only so you can dismiss it with confidence: this changes nothing for a school, clinic, factory or shop, and any consultant citing pharma research partnerships as a reason you need an AI strategy is selling on borrowed prestige. Judge AI by what it does to a task you actually perform this week.
ca.finance.yahoo.com ↗
How this briefing is put together
Every morning we read the day's AI announcements and reporting from the companies themselves and from the technology press, then pick the handful that actually change something for a working business. The analysis is ours and it is written for owners and managers, not engineers. Every story links to its original source above — read them, and disagree with us where we've got it wrong.