AskWisely.ai

Train Gemini on Your Email Judgement Calls

Document your edge cases so the model learns when to hold, escalate, or delete

Your filters catch the obvious stuff, but you still spend time on borderline requests, vague complaints, and messages that need context you carry in your head. The system cannot learn from your decisions because you have never articulated the pattern.

That is the gap this drop closes. It is a operations skill built for Gemini, and it takes about ten minutes to set up the first time. After that it runs in under a minute.

Who should use it

Operations managers who have basic automation running but still handle grey-area emails themselves

How it works

The skill file does five things, in this order.

1. Collect your last fifty handled edge cases. Export or screenshot emails where you paused before deciding what to do.

2. Label each with your decision and why. Write one sentence explaining the principle behind hold, forward, archive, or respond.

3. Feed the set to Gemini as training examples. Paste the emails and your reasoning, then ask it to propose a decision tree.

4. Test the tree against new ambiguous mail. Run five fresh borderline cases through the prompt and check if the output matches your gut.

5. Refine the criteria where it diverges. Add clarifying conditions or exceptions and update the prompt with those nuances.

What comes back

A reusable decision prompt that handles the grey zone with your logic baked in

The mistake to avoid

Store the training examples in a shared doc so you can append new edge cases monthly. Each addition sharpens the model's ability to mirror your priorities without you repeating the same explanation.

Running it

Save the prompt as a Gem so it is reusable, or paste it into Gemini with your files attached. If the task touches Sheets, Docs or Gmail, run it from the Workspace side panel so it can read the file you already have open.

Where this fits

On its own, one skill saves an hour a week. The compounding happens when three or four of them run in sequence on the same input, the same transcript that produces a scope of work also produces the follow-up email and the project brief. That is the point at which it stops being a prompt and starts being an internal tool. If you want that wired into the systems your team already uses, that is the work 67 Digital does.

In the file

You are an email triage assistant trained on my decision patterns.

Below are examples of emails I handled and the reasoning behind each action:

{paste_your_labelled_examples}

Based on these examples, when a new email arrives, evaluate it against these criteria:

1. Urgency signals: {list_your_urgency_keywords_or_phrases}
2. Relationship tier: {describe_how_you_classify_senders}
3. Completeness: does it contain enough detail to act, or does it need clarification first?
4. Risk: could ignoring or delaying this create a compliance, customer, or reputational issue?
5. Precedent: have I seen this scenario before, and what did I do?

For each new email, output:

Action: [Hold / Forward to {team_or_person} / Archive / Draft reply]
Reason: one sentence explaining which criterion drove the decision
Flag: any missing information or context I should know

New email:
{paste_incoming_email}

Apply the decision tree now.

Get it built

Reading is free. Building is what changes the numbers.

AskWisely is published by 67 Digital, an AI and automation team in Dubai. We take the skills on this site and turn them into systems that run inside your business, connected to the tools your team already uses.

AI agents that actually ship

We build the agent, wire it to your tools, and hand you the keys.

Automation on your stack

n8n, Make, or custom. Connected to the CRM and inbox you already use.

Team training in Arabic & English

Half-day sessions that leave your team using AI on Monday.

Talk to 67 Digital →