Published: 23 July 2026 · Last updated: 23 July 2026
Build an AI feedback loop that learns from every result and improves future workflows automatically without adding more to your workload.
You've got numbers somewhere that would tell you exactly what's working.
The open rates, the sales figures, the stuff sitting in a spreadsheet you've kept for years.
It's all there.
And every week it loses out to the fires, the questions, the people who need you – so it never quite turns into anything.
Same here.
I've tracked my podcast email numbers since the very start.
And most weeks it was the same line – "we'll look at it next week."
Then I got AI to read the lot and tell me what was working. That part was easy.
The hard part was the next step.
Someone still had to take those lessons and work them into how the emails get written – reading the report, working out what it meant, updating the brief by hand.
And that's the bit that always got stuck.
So I stopped treating the report as a job to do.
Now I have the findings fed straight back in, and the AI checks that before it drafts a thing.
The next one starts off better before anyone's even looked at it.
Nobody has to sit and make sense of a spreadsheet now.
The work just improves itself more each week.
And this isn't really about email.
It's about anything in your business you can measure and want to get better at – your sales follow-ups, your onboarding, your hiring.
Point AI at your data, then feed what it learns back into the instructions it works from.
It even fixed something I'd worried about for a while.
The judgement behind those emails used to sit with one person.
Now it's built into the system.
I'm two weeks in, so I won't oversell it.
But this might be the most useful thing I've worked out all year.
Have a listen – I'll walk you through exactly how it fits together.
KEY TAKEAWAYS: How an AI Feedback Loop Drives Continuous Improvement
- Turn Dormant Data into Decisions: Don’t let your open‑rate spreadsheets gather dust; use AI to surface patterns and recommendations you can actually act on to improve how your emails perform.
- Make AI Your Subject Line Coach: Let AI compare hundreds of your own emails to reveal which structures, hooks, and topics your audience really clicks on. Tapping into this information to create better subjects is key for improving opening rates.
- Capture That Tribal Knowledge Using AI and Playbooks: Bottle the intuition of those people who consistently nail subject lines by feeding their know-how into your AI and shared playbooks. When you do, your whole team will easily and reliably produce great results.
- Close the Loop Weekly: Feed each week’s results back into your subject‑line data sheet so every send gets smarter.
BEST MOMENTS: AI Feedback Loop Insights You Can Apply Anywhere
00:25 – 💬 “This is one of those rare occasions where AI actually worked and gave us a result without hours of frustration.”
04:15 – 💬 “What really matters is not, ‘Are your emails written to a very good standard?’ It’s, ‘Do people actually open the emails that you write?’”
05:56 – 💬 “The email subject is arguably as important, if not more important, than the body of the email.”
07:49 – 💬 “Every time we have either good or bad email, that information is passed to this data sheet, which is then using the source to create the new emails.”
TIMESTAMPED OVERVIEW
00:00 Intro: Using AI for Business Insights
02:26 Using AI to Analyse Email Performance
07:42 Building a Self-Improving AI Feedback Loop
09:56 Why Humans Still Matter in an AI Workflow
Episode Transcript
Please note: This transcript was generated using automated transcription tools and may contain typographical errors or inaccurately captured words or phrases.
Dr Steve Day: In today\'s episode, I wanna share a real success that we\'ve had using AI to actively improve the way we\'re working. Without just creating more work for us to have to go and do. This is about running AI to generate reports on data that we\'ve got. And then to use that data to feed into our existing pipelines. And actually improve the way future work is done.
This is one of those rare occasions where AI actually worked and gave us a result without hours of frustration. So I wanted to share it because this principle could be applied to so many parts of your business.
Okay, in today\'s episode, I wanna share how we\'ve used AI to effectively give us tips and hints on how to improve things like our email open rates and click-through rates. This is for primarily my podcast and the promotion emails that go out. But it can be used for any type of marketing and many other parts of your business as well. And it\'s the principle I\'m gonna share towards the end of this episode which is absolutely key.
It\'s about getting the AI to not only give suggestions on how you can improve, which is in effect creating more work for you. But instead to use those suggestions to improve the existing workflows. Meaning the results from any reports and the recommendations, for example, actually automatically feed back in. And improve the actual work being done in the future.
This was, for me, a light bulb moment, for a cliche phrase. Of when I suddenly connected the dots to see how I can effectively use the power and the brain of the AI. Which is massively useful for analyzing, big chunks of data. Or in this case comparing open rates and click-through rates over like, 280-odd episodes of this email. And to get useful insights from that, to see trends, to give recommendations. But then to actually do something with it, and that was the key.
So primarily, we save all of our data for our open rates, our click-through rates, and any other useful stuff. And we put that into a spreadsheet. We\'ve been doing that for, well, pretty much since the start of the episode. It\'s useful data to see. We can see, where things have gone really badly wrong. And we used to create sort of, charts based in our Google Sheets. And you could see if there was any major blips.
But it didn\'t really catch the nuances. And often we\'d not actually find time to look at this report. So even though we were creating them, or look at the spreadsheet, I should say, even though we were creating them. We\'d be busy in the marketing meetings. And it would be like, \"Oh, we\'ll just- we\'ll have a look at the report next week,\" or, \"Have a look at the spreadsheet next week.\"
And often got pushed back and pushed back. Because to actually get anything useful from it took quite a lot of thinking and brainpower. And that often was not actually at the top of my priority of just getting stuff done. And putting out the fires in the business or answering the questions and helping the staff out. So it didn\'t get done as often as it should.
AI came in and actually gave us a super effective tool to summarize and give us real actionable, results. That we could then apply to future episodes and hopefully incrementally improve stuff. We\'ve seen some benefits of that. It\'s not been, like, mind-blowing how much things have improved, but it\'s definitely not getting any worse. And it is slowly incrementally getting better.
So the fact is we\'re actually getting the AI to in effect be our marketing manager for the podcast emails. And be able to pull in insights of anything we suggest that we should look at. So if we\'ve got, \"Hey, there\'s this great marketing guy that I follow or girl that I follow,\" \"Have a look at their emails, see how they\'re constructing stuff, how do ours compare with that?\"
Also, knowing our clients and what their pain points are and their challenges. So are we addressing those in the subjects that we covered? Are we, is there any holes in the subjects that we\'ve covered that we should be covering? What are the key words or the structure of the emails that actually work best? Is it leading with a fact, is it leading with a story, is it telling something shocking?
Like, what is actually working for my actual audience? Not going out there and asking people, \"Hey, how should I write an email subject?\" Yeah, that\'s a good starting point. But actually what really matters is not are your emails written to a very good standard. It\'s do people actually open the emails that you write?
And that can only be got through data. Because your audience is unique to you. And therefore you need to be tweaking emails based on the actual data you\'re getting and any information you can feed back. So if you get, for example, somebody replying to an email saying, \"Hey, I love this email,\" that\'s a good sign. That it\'s actually probably worth doubling down on something similar to that in the future.
So all that\'s good, and we\'ve got this great AI tool that\'s creating these reports and creating actionable results. But what now happens is we\'ve got to translate those actionable results into the input. So that next time we write an email subject, we are using those insights to improve that. And that\'s where the challenge was.
It was my virtual assistant having to interpret what the AI advised. And then apply that to a completely new subject. It\'ll be a new topic in the email, like this one, for example, like talking about how to use AI to do what we\'re talking about today. She\'s got to come up with or decide on, like, what the best email subject should be. And we use AI to help us generate these emails and the email subjects. So they draft everything, and then we review it and tweak it and then send it.
And so it\'s still, it\'s coming from me. It\'s using my brand voice, using my brand guidelines, using all the information I put into it. Everything I talked about in the copywriting engine in a previous episode. But it\'s still me reviewing this and reading through before I send it. Just to be absolutely clear. So it\'s saying stuff in my voice the way I would say it. I\'m just using it as a tool to help me draft things and speed up the process.
Same thing is true for email subjects. And the email subject is arguably as important, if not more important, than the body of the email in many cases. Because if no one ever clicks the email to read it, there\'s no point in having anything in the body. And it doesn\'t matter how good or bad it is. So obviously you need both to have a good email pipeline. But you\'ve got to start at the beginning and say, \"Look, get people to open the email.\" So doubling down, actually getting the email subjects good is a very good use of your time.
So herein lies the problem. Emails will be suggested based on previous subjects without any analysis going in. And it will take the, my virtual assistant to apply what they\'ve learned over the time of working with me. Which in this case Von\'s been working with me on the podcast emails for a couple of years now. So she\'s got an incredible amount of insights and knowledge.
And she\'s seen hundreds of emails go through. The ones that have been rejected, ones that have worked, ones that haven\'t. So that brain, that knowledge is hugely valuable to me. What I wanna do is to actually remove her from being the only person or thing in the business that has that skill. Because if she ever left me, I\'d be left with nothing.
And so I want the AI to be able to think like her. To see the insights that she\'s seen, to notice the trends, and then bring that back in. So the connection was really simple. We create an output file from our reports that was specifically about our email subject, guidelines. And this is a data sheet that is constantly updated every time we run the report, which is for us once a week.
That means that as new insights come in, they are automatically updated onto this report. That is available for the email writing AI copywriting tool. So therefore, it uses the report as a source for how it constructs emails going forward. So every time we have a either good or a bad email, that information is passed to this data sheet. Which is then using a source to create the new emails.
So this loop will hopefully, and we\'ve seen it\'s working, it\'s only, it\'s week two we\'re doing this now. But the principle is absolutely brilliant. We now may have to tweak a few, things to get it working, like with every time you try to do anything with AI. It\'s never as straightforward as you hope it would be. But I\'m hoping and I believe that this is gonna be an incredibly powerful feedback loop. So every adjustment we make, it will then actually get that as a source.
And probably the next, the meta learning of this as I\'m sort of realizing it as I\'m actually saying this out loud. Is that they need to actually be talking to each other more clearly as in these two AIs. So it\'s not just actually based on this report. It\'s almost like that when it does something that actually causes it to be good or bad, then we actually need to actually know what was it about the report. Like, what was it that changed that allowed it to go the right way or the wrong way.
So there\'s some sort of additional feedback loop that might have to come in to make this completely self-learning and self-improving. But for now, this principle of taking the output from our reports, feeding that as the input into the next time the particular thing is produced. That has just saved a huge amount of brain power needed by my virtual assistant. To actually get these emails improving on a weekly basis.
And of course, she\'s still there to read them and to reflect on them and say, \"Actually, this isn\'t looking good,\" or, \"Actually, I can see there\'s an issue here,\" whatever. Or, \"This looks really good.\" And so that human element I don\'t believe will go away. I don\'t want it to go away, actually. Because I don\'t want things to just go out without me ever seeing them.
It just, like, takes away the whole personalization of business. Of the fact that I\'m a coach and I want people to engage with me as a human. Because I wanna work with them. I wanna work with them on their, the reason they do work, the input they\'re gonna make. And I wanna connect and empathize on a human level. I can\'t do that if I just let go entirely and let the AI run things. I want to actually be, yeah, getting better results. But I wanna do it in a way that\'s totally in line with the way that I work.
So that\'s it. How to get AI to do your analysis, create reports. And then the key is to then feed that back into the production element of whatever the pipeline is. So that it\'s this continual self-improving loop going forward. And hopefully forevermore.
Cool. That\'s it for today. If you found this useful, then please do give me a review on your favorite platform. Hit me a comment in whatever you\'re currently watching or listening on. And do share this with anyone else you like and love. And please do hit subscribe so you don\'t miss out on future episodes where I try and help you live with more presence, purpose, and peace.
Thanks very much.
VALUABLE RESOURCES
You can put what you have learned in this episode to work immediately and check out Claude. Download my recommended tools at sys.academy/tools and join our paid community at join.sys.academy
LINKS TO CONNECT WITH THE HOST
- Podcast: https://www.systemizeyoursuccess.com
- Website: https://systemsandoutsourcing.com/
- Facebook Group: https://facebook.com/groups/systemsandoutsourcing/
- LinkedIn: https://linkedin.com/company/systemsandoutsourcing/
- Instagram: https://instagram.com/systems_and_outsourcing/
- YouTube: https://youtube.com/@drsteveday42
- TikTok: https://www.tiktok.com/@drsteveday42
ABOUT THE HOST
Steve moved to Sweden in 2015 and transformed how he ran his businesses—switching to a fully remote model. A former NHS doctor, with a background in computing and property investing, he now helps overwhelmed business owners systemise and outsource effectively. Through his courses and coaching, Steve teaches how to automate operations and work with affordable virtual assistants, freeing up time and increasing profits. He runs his UK-based businesses remotely with support from a team of UK and Filipino VAs, and is passionate about helping others build scalable, stress-free companies using smart systems and virtual support.
