Ask Jorge.
Good morning, entrepreneurs.
This week we have a pricing change that could quietly hit your costs if you use AI coding tools, a Zapier workflow that keeps your client follow-ups from falling through the cracks, and a quick tip that will make every AI prompt you write noticeably sharper. Let's get into it.
Jorge’s Rundown:
GitHub ended flat-rate Copilot pricing this week, switching to token-based billing that could dramatically raise costs for small teams.
How to use Zapier and AI to automatically draft client follow-ups when quotes or proposals go quiet.
How adding three lines of context before any AI prompt produces noticeably better output from the start.
This Week in AI:
GitHub Copilot just ended flat-rate pricing for AI coding tools
GitHub has moved all Copilot plans to usage-based billing, replacing the fixed monthly rate with a token-consumption model called GitHub AI Credits. The change went live on June 1. Instead of a predictable flat subscription, users are now charged based on how many tokens they burn through as they work.
The details:
Smaller companies and individual developers are expected to see significantly higher costs, raising concerns about budgeting for the increased expenses. Dataconomy
Code completions and Next Edit suggestions remain included in all plans and do not consume AI Credits: the costs kick in when you use agent mode and more powerful models GitHub
A developer leaning on agent mode with a Pro+ plan could burn through their $39 monthly credit allowance in three to four days UsageBox
One Reddit user claimed their expenses would rise from around $29 to nearly $750 monthly under the new model Dataconomy
Copilot code review now also consumes GitHub Actions minutes, in addition to AI Credits, starting June 1 GitHub
The official community announcement thread has generated nearly 900 downvotes: unusual for a product people genuinely like Enterprisedna
Why it matters: Most solo business owners and freelancers using Copilot picked it precisely because the cost was predictable. That's gone now. If you or anyone you hire uses Copilot for coding, it's worth checking usage settings this week before an unexpectedly large bill arrives. The broader signal is also worth watching: the unlimited AI era is ending across the board, and the tools that felt like subscriptions are quietly becoming metered services.
Also this week:
Anthropic and the Gates Foundation committed $200 million over four years to use Claude in vaccine research, K-12 tutoring in sub-Saharan Africa, and agricultural support for smallholder farmers: four times the size of OpenAI's equivalent deal with the same foundation The Next Web
SoftBank pledged $87 billion to build AI data centres across France, the largest single AI infrastructure investment in European history Unrot
Apple registered genai.apple.com ahead of WWDC, the first time the company has publicly used "GenAI" as a label: its biggest AI moment since Siri launched in 2011 is one week away
Jorge’s Summary:
This one's worth paying attention to if you use any AI coding tools, or if you work with developers who do. GitHub quietly moved Copilot to usage-based billing this week and the backlash has been significant. The days of knowing exactly what you'll pay for AI tools are ending, and Copilot is the clearest example of that shift so far. Worth keeping an eye on your usage settings before the bill arrives.
Build with AI:
Turn your Gmail inbox into a client follow-up system that runs itself.
Most solo business owners lose deals not because they quoted wrong or delivered badly, but because follow-up slips. One missed reply. One job that sat too long. AI can fix this without you touching it every day.
Go to Zapier and create a new Zap. Set the trigger to Gmail, and choose "New email matching search." Set the filter to emails from client domains or with words like "quote," "proposal," or "invoice."
Add a second step using Zapier's built-in AI by Claude action. In the instructions field, write: "Read this email and decide whether it needs a follow-up. If the client hasn't replied to a quote or proposal in this thread, draft a polite two-sentence follow-up in my voice. If no follow-up is needed, write SKIP."
Add a Filter step: only continue if the output does not contain "SKIP."
Add a final Gmail step: create a draft email with the AI-generated follow-up. It lands in your Drafts folder, not the client's inbox. You review, tweak if needed, and send.
The result is a draft sitting in your outbox every time a client goes quiet. You stay in control of what gets sent. The system just makes sure nothing gets forgotten.
Jorge’s Summary:
If follow-up is something that slips for you, this workflow is worth the 20 minutes it takes to set up. You're not handing control to AI: every draft lands in your folder for review before anything gets sent. The system just makes sure nothing quietly falls through. Set it up once and it runs in the background every week.
Jorge’s Quick Win:
How to write better AI prompts by giving context before the task
Most people open ChatGPT or Claude and type the task. What they want. Nothing else. The output comes back decent but generic, and they spend ten minutes editing it into shape.
Try this: before you type what you want, give it three lines of context.
Who you are and what your business does. Who the output is for. One example of how you write or what good looks like for this task.
Then ask for what you want.
The model now has a frame. It knows you run a service business, not a tech startup. It knows your client is a tradie, not a CFO. It knows your tone is direct, not corporate. Every output from that point is calibrated to your situation, not a generic average of everything it has ever seen.
One extra minute of setup. Noticeably better output, every time.
Jorge’s Summary:
This one sounds almost too simple, but it's one of the most common reasons AI output feels off. The model doesn't know who you are or who you're talking to unless you tell it. Three lines of context before the task changes everything. Try it on your next prompt and see the difference.
