01

Shopify gives merchants an AI workspace for redesigning their whole store.

Shopify is rolling out Canvas, a workspace that displays a store’s pages together and lets merchants change them with Sidekick, its AI agent. Owners can edit directly or describe what they want in chat, then preview the working pages with different products and screen sizes.

Sidekick can make coordinated changes across theme files, check its code and inspect screenshots as it works. Canvas is in early access for selected stores and works only on desktop, with Shopify-developed or custom themes. Third-party themes are not supported, and Canvas does not yet replace the existing editor.

Shopify / product announcement · Requirements and limitations · Oct. 1

02

A Main Street survey finds 43% of small businesses using AI.

Heartland Forward’s new survey finds that 43% of small businesses use AI, with adoption in heartland states matching the national rate. Among owners using it, 70% report a positive financial effect and 93% say it saves time. More than half of those users report saving at least five hours a week.

The training gap is specific: 83% of AI-using owners taught themselves by trying the tools. Among nonusers, 52% say they have not seen a need. Echelon Insights surveyed 691 owners online in September. These are owners’ self-reported experiences, which can help identify useful questions for a local business without guaranteeing the same returns.

Heartland Forward / research release · Oct. 1

03

Perplexity packages everyday business tasks for Amex cardholders.

Perplexity is giving eligible U.S. American Express small-business cardholders a collection of ready-made Skills inside Perplexity Computer. The workflows cover everyday work such as cash-flow forecasting and marketing campaigns, giving owners a defined task to start from instead of designing the whole process themselves.

Access requires a paid Perplexity Enterprise Pro or Enterprise Max subscription, both of which automatically renew. Eligible cardholders can also enroll for bonus Computer Credits and a targeted subscription offer. Check the eligibility and renewal terms before buying, review the permissions needed for business data, and have a qualified person check consequential financial outputs.

Upgraded Points / product and eligibility reporting · Oct. 1

04

Gemini Live adds spoken visual help for blind and low-vision Android users.

Google launched Guided Vision in Gemini Live for compatible Android phones. Users share their camera and ask about what is in front of them. Gemini describes objects and text aloud, answers follow-up questions and gives spoken suggestions for adjusting the camera when it needs a clearer view.

The feature was developed with the accessibility community and Aira, including more than 1,000 trusted testers. Google lists uses such as reading appliance controls or locating an item on a table. It works on Android 9 and later where Gemini Live is supported. It can make mistakes and is not intended for navigation or obstacle detection.

Google / accessibility announcement · Oct. 1

05

Google introduces Gemini 4 Argon, with access limited to cyber defenders.

Google announced Gemini 4 Argon for longer coding, research and professional tasks. The company says it can find and patch software vulnerabilities and sustain much longer outputs than its earlier models. Initial access is going to selected cybersecurity defenders through Google’s Fairwind Program while the company continues safety testing.

Google plans to expand access to paying API customers and AI Ultra subscribers, but it has not supplied a public release date. Its advertised introductory API price is $2 per million input tokens and $10 per million output tokens, with higher rates afterward. Those prices describe the planned rollout; most readers cannot use Argon yet.

Google / model announcement · Sept. 30

AI WORKFLOWS | FRIDAY: RAPID SKILL ACQUISITION

An AI-built practice tool can teach you when a customer problem needs a supervisor.

A WGU report released September 30 found that 60% of 3,128 U.S. hiring professionals surveyed say AI makes candidates’ real skills harder to evaluate. That gives Friday’s learning exercise a useful finish line: demonstrate a decision you can explain without AI supplying the answer.

Say you have just moved into customer service at an appliance-repair business. You need to learn which complaints you can resolve and which require a supervisor. A delayed technician, a disputed repair and a request for a replacement appliance call for different questions and different limits on what you can promise.

Have Claude Cowork retrieve the current service policy and approved resolved cases through your company’s connected apps, then build a private practice tool. Claude’s interactive artifacts can use AI to respond to your answers and save progress. This connected, saved-progress version needs a paid plan and enabled capabilities; company settings may restrict access. In the newer unified Claude experience, start from the main conversation.

You choose the skill and source boundaries. The agent gathers the material, creates fictional cases and adjusts the practice. Ask an experienced colleague to check its answer key against current policy before you rely on its feedback. Old resolutions may contain exceptions.

Give the agent this assignment

Build a private customer-service practice tool from the approved policy and case records. Identify conflicting or missing rules before building the answer key. Create fictional scenarios without customer details. Start with a diagnostic, then adapt practice to my mistakes. Withhold feedback until I commit to an answer; cite the policy behind each correction. Keep my progress in personal storage. Test the tool’s scoring against the approved answer key. Produce a short record of what I can explain and what still needs practice. Do not contact customers, change records or share my results.

For managers, review the fresh-case answers before assigning unsupervised customer work. A confident AI score alone cannot establish readiness.

Choose one complaint you currently hesitate over. Spend your first practice session learning where your authority ends.

AI TOOL SPOTLIGHT · Bracket

Bracket remembers the decisions behind client work.

Bracket keeps track of client work as conversations change. Connect Gmail, Slack, Figma or GitHub, and it organizes scope, decisions, deadlines and deliverables, flags changes and helps draft replies. Try it on one approved project: compare its account of the latest deadline with the original conversation before using it in a client response. The first project has a 14-day, no-card trial. Listed U.S. pricing is $12 monthly for up to ten active projects, or $2 for one project active for 60 days. Review its access permissions and independently check all proposed changes before approving them or sending any client replies.

Bracket / official product details · Plans and pricing · Verified Oct. 1

Work, learning and customers

Trust and public decisions

Building and checking AI

Research and real-world deployment