The short version: Everyone tells small business owners to use AI to write their emails, but that's the wrong end of the problem. The real time sink is deciding what needs a reply at all, so the useful move is using AI to triage and summarise your inbox first, and only draft replies for the handful of messages that need one. Do that and you cut hours off your week without your clients noticing you've changed a thing.
The advice everyone gives, and why it doesn't hold up
Search "AI for email" and you'll get the same three tips repeated on every site: use ChatGPT to write your replies, set up a signature that sounds friendly, maybe try a tone-of-voice prompt. I tried all of it for about six weeks last year. Here's what happened. My replies got longer, my inbox got no smaller, and one client, a printing company owner in Reading I've worked with for years, messaged me asking "did you get someone else to write this?" She wasn't cross, just noticed. That's the tell. AI-drafted emails have a rhythm to them, a certain over-politeness, three-clause sentences that never quite land like a person typing fast between calls. People who know you can spot it in one paragraph.
The bigger issue is that writing was never the bottleneck. I can write a reply in ninety seconds. What takes twenty minutes is scrolling through 110 emails deciding which nine need me, which forty are FYI, and which sixty should never have hit my inbox in the first place. Nobody talks about that part because it's less exciting to write about than "AI writes your emails for you." But it's the actual maths of the problem.
My actual numbers
I tracked this for one week in early 2026, the way I tracked the data behind the blog post experiment I ran last year, because I don't trust a claim I haven't measured myself. Over five working days I received 587 emails, so roughly 117 a day. Of those:
- 68 percent were newsletters, automated notifications, or CC threads I didn't need to be on
- 21 percent needed a reply within 48 hours
- 11 percent needed me specifically, meaning a decision only I could make
That 11 percent, about thirteen emails a day, is the entire job. Everything else is noise I was paying attention to out of habit. Before I changed anything I was spending roughly 2 hours 40 minutes a day in my inbox. After building the system below, that dropped to 40 to 50 minutes, and the thirteen emails that mattered got faster, more thoughtful replies because I wasn't exhausted from reading the other hundred.
The client story that made this click
A client I'll call Dan, who runs a plumbing supply business near Maidstone with four staff, came to me last spring saying he was "drowning" and wanted an AI chatbot for his website to fix it. When we looked at where his time went, the website wasn't the problem. His inbox was. He was getting order confirmations, supplier updates, and internal CC'd threads mixed in with the two or three genuine customer queries that needed his judgement each day, and he was reading every single one in the order it arrived, top to bottom, every morning.
We didn't build him a chatbot. We built him three folders and a fifteen-minute morning routine using an AI summary tool built into his email client to generate a daily digest of anything that wasn't already filtered. Within three weeks he told me he'd got his mornings back. He wasn't answering more emails faster with AI-written text. He was reading far fewer of them, and only writing replies himself for the ones that needed a human decision.
The system, step by step
This is exactly what I run now, and what I set up for clients when this is the actual bottleneck rather than something flashier they've asked for.
- Step 1, audit for one week. Don't guess. Count what lands in your inbox for five working days and bucket it into three piles: needs me, FYI only, shouldn't be here.
- Step 2, kill the "shouldn't be here" pile first. Unsubscribe, mute threads, set up filters that skip the inbox entirely for anything automated. This alone usually removes 40 to 60 percent of daily volume with zero AI involved.
- Step 3, set up an AI daily digest. Most major email clients now have a built-in AI summary feature, and there are inbox tools built specifically for this. The digest should tell you, in a few lines, what came in and which threads need attention, not draft the replies.
- Step 4, only let AI draft for the FYI-adjacent middle layer. For routine, low-stakes replies (confirming a meeting time, acknowledging receipt of a document) a first-draft reply from AI that you skim and send is fine. Nobody reads tone into "confirmed, see you Thursday at 2."
- Step 5, never let AI touch the 11 percent. Anything involving a complaint, a negotiation, a favour, or a relationship you care about, you write yourself. This is where the drafting advice everyone gives you does the most damage, because it's exactly the emails where sounding like a real, slightly tired person builds trust, not the ones where speed matters most.
Setting this up takes about half a day, not five minutes, which is the other thing the "just use AI" articles skip over. You have to go through your filters, your folders and your subscriptions, and that bit is boring admin, not prompting.
The uncomfortable part nobody wants to hear
Here's the bit that sits under all of this and that most productivity advice quietly avoids: replying fast is not the same as being responsive, and a lot of small business owners have started confusing the two. AI has made it possible to fire off a polished-sounding reply in seconds, and that speed has made some people worse at their actual job, because a fast hollow answer to a client's real concern is worse than a slower, honest one. I've had suppliers reply to me within four minutes with something clearly AI-generated that didn't answer my question, and it made me trust them less, not more. Email overload, as researchers at organisations like the information overload field have documented for years, was never really solved by writing faster. It's solved by reading less rubbish in the first place.
The other honest bit is that this system only works if you're willing to sit with fewer emails and not fill the gap with something else. I've watched business owners free up ninety minutes a day and then just... use it to check email again, out of habit, in shorter bursts throughout the day, which cancels out most of the benefit. The point of triage isn't to check your inbox less well. It's to check it less often, on purpose, and trust the system enough not to peek every twenty minutes.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Where this connects to the bigger AI-for-business picture
I'm not a developer and I've never pretended to be, which is exactly the point I made in the piece I wrote about using AI daily without being technical. Email triage is a good entry point for exactly that reason: it doesn't require you to learn a new platform or rebuild your website, it just requires you to be honest about where your day goes, the same honesty I had to apply when I was working out why my own website had stopped paying me and what to fix first. Small, unglamorous fixes like this tend to move the needle faster than the big flashy AI projects everyone wants to start with.
If you'd rather have someone set this up for you rather than spend a Saturday untangling five years of email filters yourself, this is exactly the kind of practical, low-risk project an AI consultant can scope and cost in a single short call, because it's cheap to build and the return is measured in hours back every week, not some vague promise of "efficiency." And if you're curious what other quietly useful tools exist beyond the big household names, I keep adding to a running list of lesser-known but handy sites, several of which are inbox and summarising tools nobody talks about because they're not flashy enough to go viral.
Frequently asked questions
Should I let AI reply to client emails automatically?
No, not for anything involving a complaint, a negotiation, or a relationship you care about. Automatic AI replies are fine for routine confirmations and low-stakes admin, but the moment a reply matters to the relationship, write it yourself even if it takes longer.
Does using AI for email make small business owners sound less genuine?
It can, and clients notice more often than you'd think. AI-drafted text has a recognisable rhythm, over-polite phrasing and evenly weighted sentences, that people who know you well will pick up on within a paragraph or two.
What's the fastest way to reduce email overload without AI at all?
Unsubscribe and filter first. In most inboxes, 40 to 60 percent of daily volume is newsletters, automated notifications, and CC threads that don't need a reply, and cutting those removes far more time than any AI drafting tool will save you.
How long does it take to set up an AI email triage system?
Roughly half a day for the initial audit, filters, and folder setup, then a week to bed in the habit of checking a daily digest instead of the raw inbox. It's a slower start than people expect, but it holds up long after most "AI writes your emails" tricks stop feeling novel.
Related reading: AI Phone Receptionists for Small Business: What They Cost and Where They Quietly Lose You Money and I Got AI to Chase My Late Invoices. Here's What Happened.
