Google Gemini in Chrome, AP2 Agent Payments, GPT-5 Codex

AI Headlines of the Past Week

1) Google weaves Gemini directly into Chrome; agentic features on deck

What’s new: Native Gemini access via a Chrome button brings cross‑tab analysis; an AI Mode in the address bar lands later this month; Google also previewed agentic capabilities (multi‑step tasks like booking and shopping).

Why it matters to SMBs

• Your teams will get AI assistance where they already work (the browser), reducing app‑switching friction.

• Cross‑tab reasoning can compress research, QA checks, and document reviews into minutes, not hours.

• Early “agent” previews signal a near‑term shift from chat to autonomous task execution in everyday ops.

Recommended next steps

1. Pilot Gemini-in‑Chrome with a 10–25 person cohort in Sales, CX, and Ops; define 3–5 “green‑lit” tasks (summaries, comparisons, first‑draft writing).

2. Create a browser‑AI usage policy (data handling, sensitive sites, copy/paste rules).

3. Track time saved per workflow for 2–3 weeks and decide where to scale based on measured gains.

2) Google launches AP2: open payment rails for AI agents (backed by AmEx, Mastercard, PayPal, Salesforce, Intuit)

What’s new: The Agent Payments Protocol (AP2) uses signed “mandates” to authorize purchases: an Intent Mandate for discovery and a Cart Mandate for checkout. It supports cards, bank transfers, and stablecoins (with Coinbase and others). 60+ financial/tech partners are involved and specs are open on GitHub.

Why it matters to SMBs

• Establishes clear accountability when agents spend on your behalf.

• Multi‑rail support (cards, ACH, stablecoins) broadens customer payment options and agent‑to‑merchant flows.

• Open standards reduce lock‑in and speed ecosystem adoption.

Recommended next steps

1. Map your approval flow to Intent and Cart mandates (who approves, thresholds, and alerts).

2. Ask your PSP/acquirer about AP2 timelines and sandbox access; nominate a low‑risk SKU for a pilot.

3. Update risk & audit policies (logs, human‑in‑the‑loop checkpoints, refund/chargeback handling for agent purchases).

3) How people actually use AI: personal use surges, “info‑seeking” grows

What’s new: OpenAI and Anthropic usage data show personal use of ChatGPT rising from 53% (Jun ’24) to 73% (2025); Claude skews to coding while ChatGPT sees more writing/decision support; adoption in lower‑income countries is growing 4× faster (ChatGPT). Users increasingly treat AI as search/information‑seeking vs. pure content generation.

Why it matters to SMBs

• Expect employees to lean on AI for answers and judgment calls, not just drafting.

• Training and governance should target decision support (assumptions, citations) in addition to writing.

• Consumer behavior is shifting too; marketing and support scripts should assume AI‑assisted buyers.

Recommended next steps

1. Embed “show your sources” and risk prompts in internal AI templates for decisions that affect customers or cash.

2. Re‑tune site FAQs and knowledge bases for AI retrieval (clarity, structured answers).

3. Track AI‑assisted vs. manual outcomes in support and sales to guide training and model choices.

UNLOCK AI POTENTIAL FOR YOUR BUSINESS

Join Cyrus for a free 30-min AI consulting call tailored for SMB executives to explore and implement AI solutions that enhance business efficiency and innovation.