Research Period: 5 October to 12 October 2026
1. GPT-6 reaches every ChatGPT plan, with interactive answers
What changed
A fortnight after GPT-6 Sol and Luna arrived in the API (covered in wk40), OpenAI rolled the models out across all ChatGPT tiers on 7 October. GPT-6 Sol powers the paid plans (Plus, Pro, Business and Enterprise), while GPT-6 Luna serves Free and Go users from 8 October. The bigger news for day-to-day users is Intelligent UI: answers can now include interactive charts, forms, calculators and tappable buttons. There are no geographic restrictions.
Why it matters
Until now, changing one assumption meant writing a new prompt and waiting for a fresh answer. With Intelligent UI you adjust a figure in the calculator inside the response and the result updates on the spot. For an owner testing pricing, cash flow or staffing scenarios, that turns ChatGPT from a text generator into something closer to a quick, disposable spreadsheet you can share with colleagues.
Use cases
- Build a pricing calculator in a chat, then adjust margin and volume figures live with your team.
- Model three cash-flow scenarios side by side and tweak the assumptions without re-prompting.
- Generate a simple intake form to test what questions a new client onboarding process needs.
- Turn a monthly sales export into an interactive chart for a quick management meeting discussion.
- Compare Sol and Luna answers on a routine task to check which plan your staff need.
Sources: OpenAI: GPT-6 for everyone | 9to5Mac
2. Claude Haiku 5.5: Anthropic’s lowest-cost model gets much cheaper
What changed
Released on 7 October, a week after Sonnet 5.5 (wk41), Haiku 5.5 targets a different job: high-volume, routine work. Pricing is tiered. Requests under 100,000 tokens cost $0.10 per million input tokens and $0.50 per million output; above that, $0.50 and $2.50. Anthropic puts the average saving at around 75% against Haiku 4.5. It’s also the first Haiku with an adjustable effort setting, so you can favour cost or intelligence per task. The model ID is claude-haiku-5-5, available through Anthropic, AWS, Google Cloud and Microsoft Azure.
Why it matters
If you run automations that classify emails, tag records or summarise documents, this is a one-line model string change worth testing. One caution: Neowin reports that Haiku 5.5 consumes more tokens than its predecessor, so your real saving depends on your workload. Run a week of your own jobs and compare the bill before switching everything over.
Use cases
- Swap the model string in an existing email triage automation and compare a week’s API costs.
- Classify incoming support tickets by urgency at low effort, reserving higher effort for tricky ones.
- Summarise long supplier contracts in batches, keeping prompts under 100,000 tokens for the lower rate.
- Tag and tidy CRM records overnight where volume matters more than nuanced reasoning.
- Track tokens used per task so you can verify the claimed saving against your own workload.
Sources: Anthropic: Claude Haiku 5.5 | VentureBeat | Neowin: token consumption caveat
3. Google Workspace Studio adds logic and search steps
What changed
On 9 October Google announced two new step types for Workspace Studio, its no-code automation builder. Logic steps (nested conditionals) let Gemini read and classify content, then route the workflow down different paths. Search steps retrieve business information mid-flow, for example “Search for a Doc” to find a Google Drive document. It’s available on all Google Workspace plans, rolling out from 13 October for Rapid Release domains and 27 October for Scheduled Release. Note this is for Workspace accounts, not personal Google AI subscriptions.
Why it matters
Simple automations break down when every input needs the same treatment. Branching on what Gemini understands about an email or form, then pulling in the right document, means a single workflow can handle the messy variety of real enquiries. Since most members already pay for Workspace, there’s nothing new to buy.
Use cases
- Route incoming enquiries to sales, support or accounts based on what Gemini reads in them.
- Find the matching price list in Drive and attach it when a quote request arrives.
- Flag complaint emails for same-day follow-up while routine messages go to a weekly review.
- Pull the relevant policy document into a draft reply when staff ask an HR question.
- Check your release track in the admin console so you know when to start testing.
Sources: Google Workspace Updates Blog
Additional Noteworthy
Claude Dashboards, Motion, and Docs/Slides/Design out of beta: Dashboards (paid plans, beta) builds live data dashboards from plain-language prompts; Motion (Team and Enterprise, beta) creates animated explainers; Docs, Slides and Design are now on every plan. Enterprise admins must enable Dashboards and Motion. | The New Stack | FourWeekMBA
Claude for Startups expanded: Qualifying startups (founded under five years ago, or funded within two years) receive a year of Claude Team for up to five seats, $1,000 in API credits and virtual office hours. Apply at claude.com/programs/startups. | TechCrunch
ChatGPT visual ads (US only, one to watch): A new ad format appears during ChatGPT image generation, with click-to-chat and a self-serve Ads Manager with no minimum spend. It’s a US-only test; no UK or European access yet. | OpenAI: new ads format | 9to5Google
Personal Agent Protocol: Sierra and Meta propose an open standard for AI agents to log in and act inside business websites, backed by Shopify, Stripe and Walmart. See Strategic Awareness below. | The Next Web
Microsoft Work IQ preview: Grounds Copilot in live Dynamics 365 and Power Platform data. It needs Dynamics 365 and is preview only, not for production use. | Microsoft Dynamics 365 blog
Strategic Awareness
Two announcements this week are not tools to install, but they point to where things are heading. The Personal Agent Protocol, announced by Sierra and Meta on 6 October, aims to give customers’ AI agents a standard way to authenticate with your website and take actions there, such as placing an order or booking a slot. With Shopify, Stripe and Walmart involved, online sellers should expect “agent customers” to become a real channel, though the v0.1 spec isn’t published yet. Microsoft’s Work IQ, in preview since 30 September, shows the other side: assistants grounded in your own live business data. It only applies if you run Dynamics 365, but the direction is clear. AI assistants are becoming more useful when they can see your actual records, so it’s worth knowing where your business data lives and how cleanly it’s organised.
Sources: The Next Web: Personal Agent Protocol | Microsoft: Work IQ
Looking Ahead
Watch for the Personal Agent Protocol v0.1 specification, which will show what supporting agent customers actually involves. We’ll also keep an eye on whether ChatGPT’s ad formats reach the UK and Europe.
This briefing focuses on AI developments relevant to SME operations, emphasising practical applications in productivity, cost efficiency, and business capability. For questions or to suggest sources reach out on the group WhatsApp.
