Sorting the shared inbox: an AI that reads info@ first, on servers in Switzerland
Monday, nine o'clock: info@ is full again. Somewhere between the newsletters is a client whose deadline is Wednesday. Whoever opens the inbox first decides, email by email, what matters. What if an AI had sorted them before anyone arrived?
The short answer: a local model reads each email and names its topics, its urgency, the deadline it mentions and what the sender asks for, so a person starts with the email that matters. It runs on servers in Switzerland, Meow's or your own. Meow measured this on 20 test emails, each in English, French and German: facts such as deadlines come out far better than judgement. How well it sorts your mail is something to test, on your mail.
Updated on 30 September 2026
Who answers first?
- 20 of 20
Deadlines the 8B model found in the 20 French test emails, and again in the 20 German ones. Facts like dates come out well.
- 5 of 20
Emails where the 8B model got topics, urgency and deadline all right at once, in English and in French; 3 in German. Judgement needs instructions written for your mail.
- Under 8 seconds
The 8B model's median time per email in each language, measured on an Apple M1 laptop.
How to test it
0 of 8 checked
A normal week of info@, newsletters and notices included.
Topic, urgency and deadline for each, before any model sees them. That's your answer key.
The topics your team actually routes by, and what makes an email urgent for you.
Sort the 50 emails and compare with your labels, field by field.
Examples from your own inbox and your own rules; test again, and keep a change only if the scores improve.
French, German, English and Italian mail don't score the same: count them separately.
The AI sorts and sums up; a person decides and answers.
Your mail server, your case system, or a folder per topic.
Ticked them all? The kit turns this into a test you can run next week: what it sorts well, sorting rules to start from, the test plan, and what to count before sizing a server. Get the kit
What it sorts well, and what to check
The kit covers six parts of the job. Two of them:
- An email with a deadline
- The 8B model found all 20 in French and in German; check the year it assumes.
- Which topics it touches
- No model got the exact set right in more than 12 of 20; give it your categories with examples.
- How urgent it is
- In the kit
- An email in another language
- In the kit
- A newsletter or an automatic notice
- In the kit
- A reply to the sender
- In the kit
What Meow measured
Three open models sorted 20 emails written like a fiduciary's client mail, each in English, French and German, and labelled by hand first. The largest, 8B, found every deadline in French and in German, but no model judged urgency right in more than 14 emails of 20. The full test, email by email, is on the Work page.
What it needs
On an Apple M1 laptop, the 8B model took under eight seconds per email, at the median. On a server, what counts is how many emails arrive at the busiest hour: the Meow Academy's planner shows how models fit on hardware.
AcademyHow to read model scoresAcademySize a model against hardware
Where does your firm stand?
The self-check takes two minutes and says which first step fits. Or talk it through in a free 15-minute call.
Book a free 15‑minute call
Pick a time that suits you and confirm it from your inbox. If the time no longer works on our side, you hear from us the same working day.
What happens in 15 minutes
- 0–5 min
Your situation
What AI should do for you, and what it must never see.
- 5–10 min
What's possible
Which tasks can run privately, roughly what it costs, and what should stay human.
- 10–15 min
The next step
A test on your own documents, an assessment, or nothing yet.
Prefer to write? Send a message
Tell us what you'd like AI to do, and what it must never see. We'll tell you honestly what's possible, what it costs and what should stay human. No details yet? A hello is enough.