Straight answer: AI writing tools have made business email longer, more frequent, and harder to skim, because everyone’s inbox is now full of polished messages that took someone thirty seconds to generate and now take you five minutes to read. The fix isn’t using AI to write more email. It’s using AI to triage the email you already get, so you only read what needs a human. I rebuilt my own system around that idea and it’s the one AI habit that’s saved me real hours every week.
The problem nobody flags when they tell you to “use AI for email”
Every AI marketing guide tells you the same thing: use ChatGPT or Copilot to draft your emails faster. What none of them tell you is what happens on the other end, in the inbox of the person receiving fifteen of those emails a day from fifteen different people who all did the same thing.
I went through my own sent folder in January and counted. In 2022, before AI drafting tools were normal, my average outbound email to a client or prospect ran to about 90 words. By late 2025 that had crept up to nearly 180 words, not because I had twice as much to say but because AI tools default to “helpful and thorough” unless you actively fight them for brevity. Multiply that by every small business owner, freelancer, and agency doing the same thing, and you get an inbox that’s technically full of well-written email and practically unusable.
This is the bit most “10 AI prompts for your inbox” posts skip: the tool that was meant to save you time writing has quietly cost everyone else time reading. It’s the same volume-over-quality trap I wrote about when I covered business lessons from Alex Hormozi, more output isn’t automatically more value, it just moves the cost from the sender to the receiver.
What my inbox looked like
By last autumn I was getting somewhere around 210 to 240 emails a day across two accounts. Not spam, real email: podcast requests, speaking enquiries, client threads, newsletter replies, LinkedIn connection follow-ups that had migrated to email, vendor pitches from AI tools wanting me to review them. I was spending, by my own honest count over a week using a time tracker, just under three hours a day just processing inbox. Not doing the work the emails asked for, just reading, deciding, filing, replying to the two-line ones.
Three hours a day is fifteen hours a week. That’s close to two full working days spent on triage, not output. For someone running a business that depends on me delivering client work, that’s not sustainable and it’s not something a faster draft-writing tool fixes, because drafting was never the bottleneck. Deciding what deserved my attention was the bottleneck.
The shift: AI as a filter, not a writer
The change that made a real difference wasn’t a smarter way to reply. It was building a system that reads everything before I do and sorts it by whether it needs me at all.
Here’s the practical setup, and you can build this in an afternoon with tools most small businesses already have access to:
- Step 1: Turn on AI summarisation at the inbox level. Gmail’s Gemini side panel and Outlook’s Copilot both now offer a “summarise this thread” function. I use this on any thread with more than three replies before I even open it fully.
- Step 2: Set up a triage tool that scores incoming email, not just filters spam. I moved to Shortwave for this. It uses AI to bundle email into categories (needs reply, FYI, newsletters, receipts) automatically, and it learns from what I act on versus archive unread.
- Step 3: Build a “does this need a human” rule. I set a simple filter: anything that’s a straightforward yes/no, a scheduling request, or a template question gets auto-drafted by AI and I just approve or edit before sending. Anything that’s a negotiation, a complaint, or a genuine relationship email, I write myself, no AI draft at all.
- Step 4: Batch, don’t drip. I check the triaged inbox twice a day, 8am and 4pm, for twenty minutes each. Everything else waits. This alone cut my daily inbox time from three hours to about ninety minutes.
- Step 5: Audit weekly, not daily. Once a week I look at what the AI triage got wrong, an important email filed as low priority, and retrain the rule. This took maybe fifteen minutes a week and the accuracy went from roughly 70% right in week one to closer to 95% by week four.
Ninety minutes a day instead of three hours is ninety minutes back, every single day. Over a five-day week that’s seven and a half hours, close to a full working day, given back to actual client work rather than reading.
The uncomfortable part: you’re feeding your business relationships into someone else’s model
Here’s the bit that gets glossed over in most productivity posts about this. Every AI email tool that reads your inbox to summarise or triage it is, by definition, processing the content of your business relationships, contracts under discussion, client complaints, pricing conversations, through a third-party model. Sanebox, Shortwave, Gemini, Copilot, all of them need to read the email to do the job.
For most solo founders and freelancers this is a fine trade. But if you’re handling client contracts, financial details, or anything under an NDA, you need to check what each tool’s data policy says before you plug your whole inbox into it, not just accept the default settings because the demo video looked slick. I checked this for my own setup and Shortwave’s enterprise-tier data handling is different from its free tier, which most users never read past the sign-up screen. Read it. Most people don’t, and that’s the honest gap in every “just use AI for your inbox” recommendation you’ll find.
Where this breaks down
Triage tools are not magic and they get things wrong in specific, predictable ways:
- They deprioritise short, blunt emails from important people because brevity reads as “low value” to a model trained on longer, more elaborate business email.
- They over-trust subject lines. A journalist once emailed me with the subject “quick one” that turned out to be a Forbes feature request. My triage tool filed it under low priority. I caught it by luck, not by system.
- They struggle with sarcasm and tone in complaint emails, sometimes flagging an angry client message as neutral because the words are polite even though the meaning isn’t.
None of this means don’t use them. It means treat AI triage the way you’d treat a junior assistant on their first month: useful, fast, wrong often enough that you still check the “low priority” pile once a day rather than never.
Why this matters more for small business than for big companies
A company with a marketing department and an ops team can absorb inbox chaos because the work gets spread across people. A solo consultant or a two-person agency can’t. If you’re the person answering client email, writing the proposals, and doing the delivery work, every hour lost to inbox triage is an hour that doesn’t get billed and doesn’t get spent on the thing that grows the business.
This is the same principle behind why Ahrefs built their entire marketing engine around systems that run without constant manual attention rather than around one person doing everything by hand. Small businesses that build a proper filter between “incoming noise” and “things that need me” free up the exact hours that let them do higher-value work, the kind that Napoleon Hill would have called definiteness of purpose, focused effort on the thing that matters rather than reactive effort on whatever landed in the inbox last. I’ve written before about the business lessons from Napoleon Hill and this is one that holds up decades later, unfocused reactive work is the enemy of anything that compounds.
What I’d tell someone starting from scratch
If you’re a small business owner reading this with a dread feeling about your own inbox, start smaller than my five-step version. Do just this: turn on thread summarisation in whatever email client you already use, Gmail or Outlook both have it built in now, no extra tool needed, and set two fixed times a day to check email instead of leaving notifications on all day. That alone, no new software, no cost, will get you most of the time saving.
Then, once that habit sticks for two weeks, add a triage tool if your volume justifies it. If you’re getting under 50 emails a day, you probably don’t need one. If you’re over 100, you almost certainly do, and the fifteen minutes a week to retrain the filters will pay for itself many times over.
The same logic that makes an interactive calculator more useful than a static blog post, it does the sorting work for the visitor instead of making them read everything, applies here. The goal isn’t more content or more email. It’s less friction between the person and the decision they need to make.
Frequently asked questions
Does using AI to summarise email save time or just move the work around?
It saves real time when it’s used for triage and summarising long threads, roughly ninety minutes a day for me after a month of use. It does not save time if you use it to generate longer, more elaborate replies, because that just adds to the volume everyone else has to read.
Which AI email tools are worth paying for as a small business owner?
Gmail’s Gemini panel and Outlook’s Copilot are free or included in existing business subscriptions and cover basic summarisation. Shortwave and Sanebox are worth the monthly cost, typically $10 to $30 a month, only once your daily volume is consistently over 100 emails and manual sorting is costing you more than an hour a day.
Is it risky to let an AI tool read client emails for triage?
It can be, depending on what’s in those emails. If you handle contracts, financial details, or NDA-covered material, check the specific data policy of the tool’s tier you’re on rather than assuming all tiers handle data the same way. For general business correspondence the risk is low, but it’s worth ten minutes of reading before you connect a full inbox.
What’s the single fastest change to make if I don’t have time to build a whole system?
Turn on built-in thread summarisation in Gmail or Outlook and set two fixed inbox-checking times a day instead of leaving notifications on. Both take under five minutes to set up and cut most of the reactive time loss without adding any new tool or cost.
Related reading: AI Meeting Note-Takers: What Nobody Tells You Before You Turn One On and Mailchimp Marketing Strategy: How They Built a Brand That Wins.
