Skip to content

AI agents & automation

The shared inbox that sorts itself

An operational inbox receiving hundreds of messages a day always ends the same way: as an archive nobody opens. The triage agent reads every message, spots what needs action, answers the recurring, and routes the rest to the right person with the context already extracted.

Get my free diagnostic

The problem it removes

How it works

  1. 1

    Measure before sorting

    We start by measuring the real flow: who writes, about what, what actually needs action. Sorting rules are built on that data — not on impressions.

  2. 2

    Sort, extract, route

    The agent classifies each message, extracts the useful data (references, amounts, deadlines) into your tools, and routes to the right person. Urgent items are flagged within minutes.

  3. 3

    Answer the recurring

    Acknowledgements, document requests, standard questions: the agent replies on its own within the approved fence. The rest lands pre-qualified in your team’s queue.

What the agent does

Measure

of message volume and types before any automatic sorting

Urgent

flagged to the right person on arrival, not when the batch is read

Traced

every message classified, routed and logged

What the agent does by design, not measured averages: your results are measured during the diagnostic and the pilot, on your own data.

Frequently asked questions

What if the agent misclassifies an important message?+

Rules are calibrated on your real flow and start in assisted mode: the agent proposes, your team validates, the agent learns. Autonomous sorting is only switched on once reliability is demonstrated — and everything stays browsable, nothing is deleted.

Does it work with Gmail, Outlook, a shared inbox?+

Yes — Gmail, Microsoft 365, IMAP, shared inboxes and aliases. The agent works inside your existing mailbox, without changing the address your clients know.

Does the agent read my whole history?+

Only what is needed, defined with you: usually the incoming flow from go-live, plus a sample of history to calibrate the rules. Minimal access, logged, GDPR-compliant.

Is this the problem eating your team’s time?

Tell us how you work today — 30-minute call, then a free written diagnostic of what this agent would change for you, with numbers.

Get my free diagnostic

Free resource

The self-assessment grid: 20 tasks AI can automate

Sales, admin, support, operations: the 20 tasks AI agents already handle in SMEs — with, for each one, the tell-tale sign that your team is concerned.

This solution, industry by industry

Email triage — Transport & logistics

The operations inbox of a 45-person freight forwarder takes 350 emails a day: pickup requests, delivery disputes, demurrage notices, customs documents, subcontractor notifications — all mixed with internal threads.

  1. The agent classifies every incoming message continuously: urgent pickups to the relevant area dispatcher, demurrage notices to the overseas desk with the contractual countdown in the subject line.
  2. On a customs file, it extracts the MRN number and tariff heading from the attached document and pushes them into the matching transit file, no retyping.
  3. Dispute emails trigger an immediate acknowledgement to the client and a pre-filled card (reference, date, nature of damage) in the relevant team’s queue.

No more demurrage notice discovered three days too late at the bottom of the inbox: every message category reaches the right person within minutes, references already extracted.

Order of magnitude

  • · 350 emails a day in the operations inbox, roughly 60% of which can be classified without human judgement
  • · 1.5 minutes on average to read, qualify and forward a message manually

That is about 210 messages a day routed automatically — in the order of 5 hours of daily sorting removed, with true emergencies (demurrage, customs) handled in minutes rather than hours.

Indicative estimate built on average sector assumptions — it gets recalibrated on your actual volumes during scoping.

See the industry page →

Email triage — E-commerce

A brand receiving 250 messages a day across the contact inbox, customer service and three marketplace messaging systems. Every morning, one person spent their first hour and a half on pure routing.

  1. At opening time the sorting is done: marketplace messages under a 24-hour deadline sit at the top of the queue, timestamped, remaining time displayed.
  2. A supplier invoice arrives as an attachment: references and amounts extracted, it is already in the accounting circuit before the team logs in.
  3. An influencer partnership pitch is recognised and routed to marketing with the media kit attached in reply — without passing through a support queue that does not want it.

Nobody “does the inbox” anymore: everyone receives their queue pre-qualified, and platform-imposed deadlines stop being a source of penalties.

Order of magnitude

  • · around 5,000 messages per month across all inboxes
  • · 30 seconds of manual reading and routing per message on average

Close to 42 hours of sorting erased each month — the equivalent of a full working week, every month.

Indicative estimate built on average sector assumptions — it gets recalibrated on your actual volumes during scoping.

See the industry page →

Email triage — Real estate

120 messages a day in this typical agency’s inbox: portal leads, tenants, contractors, surveyors, notaries — and the urgent water damage waits its turn behind the newsletter.

  1. A contractor’s estimate arrives for an ongoing water damage: matched to the work order and the unit concerned, it is pushed to the manager for approval — no more orphaned attachments.
  2. A tenant sends one email mixing a departure notice with a service-charge dispute: the agent splits it into two cases, each routed to the right circuit with an acknowledgement.
  3. By 5 PM, nothing sits unanswered: every sender knows their request is logged, with whom, and by when — even when the substantive answer comes later.

The inbox stops being a lottery where the urgent queues behind the trivial; every morning the team opens a flow already sorted, referenced and acknowledged.

Order of magnitude

  • · 120 messages a day
  • · an average of 90 seconds’ reading and filing per message
  • · 5 working days a week

Around 3 hours of sorting a day removed — 15 weekly hours returned to the team, not counting the portal leads and management emergencies that no longer get lost. An order of magnitude measurable from the observation phase on your own flow.

Indicative estimate built on average sector assumptions — it gets recalibrated on your actual volumes during scoping.

See the industry page →

Email triage — Healthcare & clinics

The secretariat of a surgical clinic runs two mailboxes — contact and admissions — 150 messages a day where admission files sit next to supplier advertising.

  1. An incomplete admission form is detected on arrival: the patient receives the exact list of missing documents, not the evening before surgery.
  2. A request for a stay estimate reaches the admissions office with the administrative file pre-filled.
  3. An email mentioning post-operative bleeding is passed to the care team as a priority — routed, never interpreted, never answered on substance.

Admissions get prepared upstream instead of being fixed in the lobby on the day, and nothing touching on care lingers in a generic inbox.

Order of magnitude

  • · 150 messages a day spread across 3 secretaries
  • · about 1.5 hours of sorting each, daily

Over 20 hours of filing a week handed back to reception and patient files — while the sorting itself runs continuously, weekends included.

Indicative estimate built on average sector assumptions — it gets recalibrated on your actual volumes during scoping.

See the industry page →

Email triage — Recruitment & staffing

In high season, an agency’s application inbox takes 150 CVs and messages a day — and the assistant handling them does nothing else, and still runs late.

  1. Overnight Sunday to Monday, 84 applications land after the weekend’s ad run: by 8 AM all are in the ATS, deduplicated, with the five profiles matching open orders flagged on top.
  2. The 6:50 AM withdrawal for an 8 AM assignment triggers an immediate alert to the account manager — not a 9:30 discovery.
  3. Timesheets received by email flow to invoicing, matched against the corresponding assignment contract.

The application flow gets processed as it arrives whatever the season, and the assistant requalifies profiles instead of doing data entry.

Order of magnitude

  • · 150 applications and messages a day in season
  • · 3 minutes of manual handling each: opening, ATS entry, acknowledgement

On the order of 7 processing hours a day absorbed (150 × 3 min) — a volume no single assistant could hold, and one that no longer forces choosing which CVs to open.

Indicative estimate built on average sector assumptions — it gets recalibrated on your actual volumes during scoping.

See the industry page →

Email triage — Wholesale & B2B distribution

The order-desk inbox of a technical wholesaler takes 250 messages a day for four clerks — orders, acknowledgements, disputes and supplier advertising all mixed together.

  1. At 8:05 the queue is clean: sixty orders pre-keyed into the ERP, twelve price requests routed to the reps, the advertising archived.
  2. A crookedly scanned order: the agent extracts what it can, marks the two unreadable lines, and the clerk completes it in thirty seconds instead of retyping everything.
  3. A client replies “fine by me” to the twenty-second message of a thread: the approval is attached to the pending quote, which becomes an order.

Nobody “does the inbox” first thing anymore. Clerks start straight on the exceptions; retyping disappears from the job.

Order of magnitude

  • · some fifty orders a day to re-key, at 8 minutes each
  • · 150 messages to sort and route, at 30 seconds each

Close to 8 hours a day handed back to the order desk — order entry alone accounts for more than 6 of them.

Indicative estimate built on average sector assumptions — it gets recalibrated on your actual volumes during scoping.

See the industry page →