Thoughts & Perspectives
Building in public
AI for Manufacturing (2026): What a $2M to $50M Shop Can Actually Use This Quarter, Before Anyone Says "Smart Factory"
Type "AI for manufacturing" into a search engine and you get predictive maintenance, machine vision, and a digital twin. Type it as the owner of a 40-person shop and what you need is quotes out faster, order-status emails answered, work instructions that exist in two languages, and a maintenance log someone can actually read. This is that version: the five office jobs a general assistant does for a small manufacturer as of September 2026, the two things it must never touch, the plan that keeps your drawings out of a training set, and what it costs.
Gemini Usage Limits (2026): What Free, AI Pro, and AI Ultra Actually Cap, and When They Reset
Gemini usage limits as of September 2026: Google publishes no prompt count, only multipliers. Free gets standard limits, AI Plus 2x, AI Pro 4x, AI Ultra 5x or 20x Pro, all compute-based, refreshing every 5 hours until a weekly cap. What burns the allowance fastest, the 32K / 128K / 1M context windows, and what happens when you hit the wall, from Google's own pages.
Copilot for Outlook (2026): Summarize, Draft, and Triage Email, and Which Microsoft 365 License Turns It On
Copilot in Outlook as of September 2026: thread summaries with citations, drafts and coaching in the compose box, inbox prioritization, natural-language triage and rules, and meeting invites built from an email. Which features come with every Microsoft 365 business plan, which need the $30 (or $18 to $21) add-on, the limits, admin settings, and 6 daily prompts.
Copilot for Excel (2026): What It Actually Does, Which License You Need, and 7 Prompts That Work
Copilot in Excel as of September 2026: it edits your workbook directly in edit, plan, or chat mode, runs Python for analysis, and switches between Claude and GPT models. Which licenses include it ($30 add-on, $18 to $21 Copilot Business, $19.99 Premium), the limits Microsoft publishes, and 7 prompts with what they return.
AI Usage Limits by Plan (September 2026): ChatGPT, Claude, Gemini, and Copilot on One Page
Every AI plan's usage limit on one page, as of September 2026: ChatGPT and Codex, Claude, Gemini, and Microsoft 365 Copilot, one row per plan, with price, what's unlimited, what's metered, the window, and the reset. Read from each vendor's own pages, re-checked every Monday, with the plan I'd buy for a solo operator, a 5-person team, and a 20-person team.
ChatGPT Connectors (2026): Which Apps Connect, What Each Plan Gets, and How to Set Them Up Safely
ChatGPT connectors as of September 2026: OpenAI now calls them apps and ships them inside plugins. Which apps connect (Google Drive, Gmail, SharePoint, Outlook, Slack, GitHub and more), what Free, Plus, Business, and Enterprise get, the permission defaults, and a 6-step setup for a small business.
AI for Wealth Management (2026): What a Small RIA Can Actually Use, and the Two Rules That Decide What It Can Say
Most "AI for wealth management" content is written by a platform vendor for a firm with 400 advisors. This is the version for the owner of a small advisory firm: the four jobs a general assistant does every week (meeting prep and recap, client drafts from your templates, first-read summaries of statements and documents, the inbox), the one job it never does alone, what the SEC Marketing Rule and FINRA's own notice actually require of AI-written material, the plan choice that keeps client data out of a training set, and what it costs as of September 2026.
Copilot in Teams (2026): What It Actually Does in Meetings, Chats, and Channels, What It Needs, and What It Costs
Copilot in Teams, as of September 2026: it recaps meetings, catches you up on chats and channel threads, and suggests follow-up tasks. It needs a Microsoft 365 Copilot license on each person who uses it ($21 a user a month on Business plans), and under Microsoft's default meeting policy it waits for someone to start transcription. What it reads, what it can't, the setup checklist, and the 10-person math.
Claude Memory (2026): How It Works, What It Stores, How to Turn It Off, and Why Your Team Plan Has It Off
Claude memory, as of September 2026: on by default for Free, Pro, and Max, off on Team and Enterprise until an owner turns it on. How it saves topics as you chat, how it differs from chat search, what it refuses to store, pause vs reset, incognito chats, Cowork, and why your business context belongs in a project and a written doc instead of one person's memory.
How Much Does AI Cost for a Business in 2026? The Four Line Items Nobody Quotes You Together
How much AI costs a business as of September 2026, line by line: seats at roughly $20 to $25 per person per month, API usage if you build ($1 to $50 per million tokens on Anthropic's models), AI already inside Google Workspace and Microsoft 365, and one-time setup (about $4,500 to $25,000 for a first install). Worked budgets for a solo owner, a 10-person team, and a 25-person team, plus the hidden cost nobody invoices.
Claude Fable 5.1 (2026): What It Is, What It Costs, Which Plans Include It, and When to Use Opus 5 Instead
Claude Fable 5.1, as of September 2026: Anthropic's most advanced generally available model, announced September 1. $10 per million input tokens and $50 output on the API, twice Opus 5, with cache hits at $0.25. Not on Free or Team, included on Enterprise, metered separately on Pro and Max. What it is, the full price sheet, the workload math, and why Anthropic's own docs say to start with Opus 5.
AI Integration Services (2026): What You're Actually Buying, What It Costs, and When to Keep Your Money
AI integration services, as of 2026: what is actually inside an install (seats, connections, business context, approval-gated workflows, training), the three kinds of integration you get quoted, what each costs, and the five questions that separate a fixed-price install from an open-ended retainer. With my published prices: Assessment $2,500, installs from $4,500.
Ready to scale your product vision?
I help high-growth startups bridge the gap between technical complexity and business revenue.
Book a Strategic Call