AI for Email: Four Ways to Get Your Inbox Under Control
A Practical Guide for Business Owners and Communicators
Email technology has really come a long way. How much of it you can use will depend on what tools you have access to, how your computer is set up, and honestly, your comfort level as well.
We have to say from the beginning that we are big believers in writing your own emails. They add that personal connection. But there are ways you can use AI to help you organise your emails, summarise them, and get through them faster.
And sometimes, we must admit, we do use it to help craft things. If you need to have a difficult conversation, AI can be very helpful in rewriting the copy so it conveys what you need to convey without any sort of upset.
Because here is the thing. Words on their own do not convey tone, and they do not convey body language. They can be interpreted in a whole range of ways. You have to be very careful with what you write in an email, and AI can help you with that.
A Full-Time Job You Never Applied For
The average professional spends around 28% of their workday on email. That is more than a full day, every week, just sorting through things, getting back to people and replying.
Most of us have multiple inboxes going, too. One for the business, personal ones, one for another job, one for client work. It adds up, and during busy periods the volume increases dramatically.
Can email be completely eradicated? We do not think so. But let us see what we can do to manage it.
The Golden Rule
Before we get into the tools, this is the part that matters most. We might let AI help us with tasks, but we are the last decision maker before anything goes out.
The technology that can go into your inbox now exists, it can write your emails and send them on your behalf. However, we think we will always have trust issues with AI on that one. We would always want to see what was going out.
So in this session we looked at four things: triaging, drafting, humanising and optimising. What we did not look at is emails that begin with "I hope this email finds you well," because that just makes our eyes twitch.
We posted about that on LinkedIn and some people said they always start their emails that way. But so does the free version of ChatGPT. It is so obvious. And you can sometimes lose people if they read that as a first line and think, oh, you did not even write this yourself.
Step One: Triaging and Summarising
This is where AI can be super, super helpful.
If you come back to work after a weekend, or after being out of office, your inbox is just full of all sorts of things. Sometimes there is a wall of emails and you think, I do not even know where to start.
You can ask it to tell you what the most important things are.
In Gmail, you click on that little star, which is Gemini already integrated into the platform, and you can ask something like "What are the most important emails I need to respond to today?" It scans your inbox and brings it out for you.
Microsoft Copilot is actually really good for this too. One habit we love is doing it on your phone over morning coffee. Ask "What do I need to do first today?" and it will give you a list. Then when you jump on your computer, you are already well prepared.
It is also good not only for email but with your calendar as well. Ask "What do I even have to do today?" and it will list things out for you.
Just remember: AI can misjudge urgency. Check before you archive.
Assistants Versus Agents
There is an important difference here, and it is worth understanding.
With something like Copilot and Gemini, you are the one prompting it. You ask, it drafts, you send. These are assistants and they are embedded in the inbox you already have.
With an agent, you can set it up so it already does that for you. Things like Claude Cowork and ChatGPT Work can work autonomously across your apps.
We know some people who have Cowork set up so that every morning it goes into their inbox and not only triages, but based on its knowledge of all their files and interactions and understanding of their voice, it writes drafts in response to every email that requires answering. Rather than going through and doing it yourself, you just go in and edit what has been drafted before sending it out.
Can you imagine the time saved on that? But you still need to check it, of course.
Worth knowing: agents have requirements. You need a recent enough operating system, enough grunt on your computer to run it, and you need to allow it to have access. If you are on a work machine, your organisation may well not permit an AI agent into your files or your emails. That is a conversation to have before you start.
Watch the full webinar recording below where Dr Karen demonstrates the triage and drafting features live in Gemini, and walks through exactly how each one works:
Now that you have seen those demos, let us get into the part that makes the biggest difference to how your emails actually read.
Step Two: Humanising Your Content
"I hope this email finds you well." That corporate porridge. That AI slop we all see.
It happens because we are not using AI well. It is lazy AI writing, rather than giving it the different ingredients it needs to write something that sounds like you.
There are two ways to fix it.
The first is to set up a dedicated space that learns your voice. A Claude project, a custom GPT, a gem in Gemini, or an agent in Microsoft Copilot. You give it previous emails to train on so it understands you. You will still need to give it context each time, but it starts from a much better place.
The second is the formula. When you use the drafting tool in your inbox, give it these four things:
Context + Relationship + Tone + One real detail
Context. What this email is actually about.
Relationship. Who they are to you. If you have a colleague you are really good friends with, you might still be professional, but you would write to them differently than to the CEO.
Tone. How you actually talk, aligned with that relationship.
One real detail. Something true about this person that it can work with.
Compare the two prompts.
A lazy prompt: "Write a follow-up email." It does not tell it or explain what it needs. You get "I hope this email finds you well. I am reaching out to touch base."
A formula prompt: context + relationship + warm tone + the detail they mentioned, a product launch. You get a reply that sounds like you, references their launch, and gets to the point.
The same formula works in Copilot, Gemini, Cowork and ChatGPT Work.
And if AI invents warmth or facts it does not have, cut them.
Step Three: Smarter Send Times
This one comes with a caveat. We tried this in both Gemini and Copilot and they are not brilliant at auto-optimising per recipient. They can see when people are opening emails, but at the moment they cannot quite do the step of analysing that and identifying the best times.
Where this works properly is in electronic direct mail tools like Mailchimp, which have AI embedded that analyses your database, sees when people open, and identifies the best time to send so they will actually open it.
That said, asking is still useful. We tried "Analyse my interactions with [contact] and identify the best times to send emails to him" and it came back with a recommended window, explained the time difference because he is based overseas, analysed our previous interactions, and listed the times to avoid. That is genuinely helpful.
Then use scheduled send in Outlook or Gmail to act on it.
One thing to keep in mind: one 6am reply does not make someone a 6am person.
Step Four: Subject Lines That Earn the Open
AI is very good at this. If you are sending EDMs and you want people to open them, the subject line is really important. Even in a more corporate setting where people know who you are, it still comes into it. We all get so many emails that you need something that is going to engage someone and get their attention.
Here is the approach:
Ask AI for 8 to 10 variants
Rank them against your rules: clear, value up front, under about 50 characters, no clickbait
Pick two and test them on real sends
We tried "I'm writing a promo email to share the details of my upcoming Lightning Lesson about finding your team's AI quick win. Give me five high-converting subject lines" and got back options like "Stop guessing, find your team's first AI quick win" and "Stuck on where to start? Join me this Wednesday."
What we often do is take a bit of one and stick it with another. It is just enough to get your brain working so you can come up with your own ideas as well.
One warning: AI optimises for what you ask. Say "no false urgency" or you will get it.
A Bonus Use Case: Finding the Email You Cannot Find
This came up from someone in the session and it is one of the best everyday uses there is.
Copilot is amazing for finding an email you cannot find. Sometimes you know you have had a conversation about a certain thing, but you cannot find it when you do a normal search. You can just say "I know I talked about this," and it will find it for you.
We also use it before meetings. "What is that person's name? Where am I going? Give me a three bullet point summary of what I am doing." It saves you from fossicking and searching.
AI Is Not the Same as Automation
These two get confused a lot. Automation is a whole other thing, built on rules, but it can be super helpful. You can set it up so that if someone emails you about a certain thing within your business, and there is normally the same sort of reply needed each time, that reply goes out automatically.
It is a bit different to AI, but it is still a really great way to go and it can save heaps and heaps of time.
What Still Needs You
AI gives you the first draft. You give it judgement. Here is what still needs a human eye:
Urgency calls the AI gets wrong
Warmth and facts the AI invents
Timing based on too little data
Subject lines that win the open but lose the trust
And remember, when you are emailing, it is you. It is directly you. Same as when you are posting on social media. It is a representation of you, so you have to make sure you are overseeing what it is doing.
Always add what we like to call your human sparkle to make it sound like you, not like the AI. We do not need any more robotic sounding emails out there.
Key Takeaways
Use AI to triage your inbox and turn "50 unread" into "5 decisions"
Understand the difference between assistants, which you prompt, and agents, which work across your apps
Use the formula for drafting: context, relationship, tone, one real detail
Ask AI to read your thread history for send timing, then use scheduled send to act on it
Get 8 to 10 subject line variants, rank them against your own rules, test two on real sends
Copilot is excellent for finding emails you cannot find and prepping for meetings
Never let it send on your behalf without you seeing it first
If you were going to pick one of these things this week, what would you have a go at?
Start with one step. That is all it takes to see the difference.
If you want support integrating AI strategically into your business, here are some ways we can help:
Join Dr Karen's AI Integration Sprint - helping senior leaders do the strategy work to integrate AI into their teams, certified by the AI Centre of Excellence.
Come along to our next free webinar, AI for Event Planning, at 4pm on Monday 24 August, where we look at different ways to streamline those practices
Join our free Facebook community - where we share different AI tools and articles, and go live every Friday afternoon talking about what has happened in AI this week
Book a chat with Dr Karen - let's talk about how AI can work for your specific situation
Because AI adoption without strategy creates noise. AI adoption with structure creates advantage.
