The Do’s and Don’ts of Mercury Retrograde, AI Version

I pay attention to patterns.

Mercury retrograde gives me a named season to expect a little friction, read more carefully, and move with more patience. This one runs from October 24 through November 13, 2026.

I still sign contracts. I still start projects. I still travel. I still live my life. I simply bring more intention to the second look.

This time, I am giving my AI the same treatment.

For Mercury retrograde, I am putting my AI in retrograde too.

Your AI may know a business that no longer exists. The offer changed. The website changed. The customer changed. Your standards changed. The AI kept working from the last version you taught it.

An obviously wrong answer is easy to catch. Stale context is harder because it was once true.

This article was developed by the Seed & Society A.I. blog employee from Makeda Boehm’s approved podcast episode. Makeda approved the source episode and did not personally review this article line by line.

What Mercury retrograde means here

In astronomy, Mercury retrograde is the period when the planet appears to move backward against the sky from our view on Earth. In astrology, the period is commonly associated with communication, agreements, travel, and technology.

I use astrology as a reflective framework. Communication mix-ups, contract surprises, travel delays, and technology problems can happen at any time. I do not treat every glitch as proof that Mercury caused it.

The value for me is the checkpoint. A named season reminds me to slow down, check twice, and respond with more grace when friction appears.

AI needs that kind of checkpoint because it can be functioning exactly as designed while using information your business has already outgrown.

What it means to put your AI in retrograde

Putting your AI in retrograde means returning to the work, reviewing what changed, correcting the sources behind the system, and deciding what still needs human judgment.

It does not require shutting down your tools or auditing every piece of your business at once. The reset begins with one consequential or recurring workflow.

These are my do’s and don’ts of Mercury retrograde, AI version.

The do’s and don’ts of Mercury retrograde, AI version

Do: Use the season as a checkpoint

Slow down long enough to notice what you may have been moving too quickly to see.

Look for the places where your business, tools, offers, standards, or decisions changed while the AI kept working from an earlier version.

Don’t: Blame every glitch on Mercury

Check the process too. A broken link may be a broken link. A wrong answer may come from an outdated source. A customer-facing mistake may reveal that the workflow no longer matches the business.

The reflection becomes useful when it leads to a real correction.

Do: Check the AI work that carries consequence

Review the output that represents you to another person or affects a meaningful decision.

Check names, dates, numbers, links, sources, deadlines, commitments, and customer-facing language. Then read for meaning.

Does the answer say what you mean? Does it reflect the decision you actually made? Is it making a promise you approved?

Don’t: Trust an answer because it sounds confident

A fluent answer is not automatically a verified answer. AI can present stale information with the same confidence it uses for current information.

Review should match consequence. A private brainstorm and a customer promise do not need the same level of checking.

Do: Revisit the AI tools you already pay for

Open one tool you liked three or six months ago and see what it can do now.

AI moves at the speed of daily. The tool may have new models, better integrations, stronger controls, useful automations, or an easier way to complete work that once required a complicated workaround.

You may already own more capability than you are using.

Don’t: Buy another tool before you check

Start with the stack you already have. Look at the current features, settings, permissions, and integrations before adding another subscription.

A new feature only becomes useful when it supports a real job. Record what changed and which workflow it could improve.

Do: Read your business like a customer

Review your website, landing pages, sales pages, marketing emails, automated messages, intake forms, and recurring workflows from the customer’s point of view.

Does the public version of the business match the actual version? Is the offer current? Does the link go to the right place? Is the next step still the next step?

Don’t: Keep sending an offer or promise that changed

Automation is excellent at repeating instructions. That includes instructions your business has already outgrown.

Update the source before the same expired offer, link, promise, or instruction reaches another person.

Do: Compare AI activity with a real result

Put the AI activity beside the outcomes the work is supposed to support.

Did it save time? Did it support revenue? Did it improve the quality of the work? Did it create a useful option? Did the content reach the right people?

Don’t: Confuse more output with more progress

A system producing more drafts proves that it produced more drafts. The business result depends on what those drafts helped someone do.

Count the outcomes that matter to the workflow, including accuracy, response, time saved, revenue supported, useful choices created, and the amount of correction still required.

Do: Update what AI knows about your business

Refresh the source documents behind the work. Review the current offer, audience, voice, rules, examples, proof, priorities, and permissions.

Then teach the AI what changed.

If the business moved, the context has to move with it.

Don’t: Give AI old information and expect current work

Stale context often looks polished because it was once accurate. Your AI may keep using an old biography, price, policy, customer description, or offer until the governing source changes.

Correcting one final draft may solve today’s problem. Correcting the source can improve every later draft that depends on it.

Do: Keep important decisions human

Define what AI can prepare and what a person must decide.

AI can compare, organize, summarize, research, draft, and flag what appears to be missing. Those are useful responsibilities.

The final call on contracts, money, customer promises, public claims, and relationship-changing messages belongs with the person who has the context, authority, and life that choice will affect.

Don’t: Let AI make a decision you have to live with

Convenience does not make a consequential decision less consequential.

Write the human handoff into the workflow. Make it clear where preparation ends, who reviews the work, and who has permission to decide.

Do: Change one workflow at a time

Choose one recurring workflow close to money or customers. Update its source, run it again, compare the result, and preserve the correction.

Don’t: Change everything and lose the lesson

When everything changes at once, you cannot see what improved the result.

One workflow gives you a useful comparison between the old context and the current one.

Your 30-minute Mercury retrograde AI reset

Pick one recurring workflow and complete these steps:

1. Name the source. Identify the document, database, project, or instruction the workflow uses.
2. Find what changed. Check the offer, audience, rules, examples, links, proof, priorities, and permissions.
3. Update the source. Preserve the current information where the workflow actually reads it.
4. Run the same workflow again. Keep the task consistent so you can compare the old and new results.
5. Check the result. Review factual accuracy, meaning, usefulness, and the required level of human judgment.
6. Keep the correction. Save the improved source or instruction so the next run begins with better information.

Start with the workflow closest to money or customers. That may be a sales email, proposal, landing page, intake process, customer automation, or content system that communicates a current offer.

Why this is Context Training

Context Training is a living practice because the world it describes keeps moving.

Say what you need. Show what matters. Check the work. Teach what it missed.

The loop continues because your business, audience, tools, decisions, and standards continue to change.

Mercury retrograde gives me a return date. It reminds me to expect friction without treating every friction as failure. It reminds me to return, review, and respond with more grace.

Your AI can only work from the world you taught it.

If your world changed, teach it again.

Frequently Asked Questions

When is Mercury retrograde in October and November 2026?

Mercury is retrograde from October 24 through November 13, 2026.

Does Mercury retrograde cause AI errors?

This article does not make that causal claim. Communication problems, technology failures, stale sources, and workflow mistakes can happen at any time. Makeda uses Mercury retrograde as a reflective checkpoint that prompts a more intentional second look.

What does it mean to put AI in retrograde?

It means returning to the work, reviewing the tools and sources behind it, correcting stale context, comparing activity with results, and confirming which decisions still require a person.

Should I stop signing contracts or starting projects during Mercury retrograde?

Makeda does not. She continues signing contracts, starting projects, traveling, and living her life. She brings more intention to review, communication, and the second look.

What is stale AI context?

Stale AI context is information that was once accurate and no longer matches the current business, customer, offer, policy, priority, or decision. It may include an old price, link, biography, promise, example, rule, or audience description.

What is the first AI workflow I should review?

Start with one recurring workflow closest to money or customers. Choose work where stale information or a missed decision could create a real consequence.

Do I need to review every AI output personally?

Review should be proportionate to consequence. Customer-facing, financial, contractual, public, and relationship-changing work needs stronger human review. Low-consequence internal exploration can use lighter checks.

Can I use this reset outside Mercury retrograde?

Yes. The practice works whenever your business, tools, policies, offers, or priorities change. Mercury retrograde makes the return date memorable and gives this particular reflection its meaning.

Listen to the full reflection

Listen to Your AI Needs a Retrograde on the Seed & Society podcast.

Get the Context Training book.

This article is adapted from Season 3, Episode 15 of the Seed & Society podcast. Explore more practical writing in The Connectors Market.