Your inbox, already read.
- First reply overdue by 2h — SLA breached
- Sender is waiting on your team's reply
- Classified urgent (92% confidence) by rule 'Billing disputes'
- Topic
- Billing
- Priority
- Urgent
- Sentiment
- Frustrated
- Confidence
- 92%
Invoice #4482 dispute — Acme Corp
Customer disputes a duplicate charge of $4,820 on the April invoice.
W-9 request — Northwind Ltd
Vendor asks for an updated tax form before releasing payment.
Refund escalation — order 91277
Second follow-up; customer threatens chargeback if unresolved today.
Payment terms question — Lumen Co
Waiting on finance approval for net-60 terms extension.
Credit note issued — Bright Labs
Credit note CN-2210 sent and acknowledged by the customer.
Incoming work can arrive classified, summarized, and explained before your team opens it.
Your team starts at step three
By the time a human touches an email, the reading and the first draft are already done.
Northwind disputes $12,400 on INV-2417; they accept the revised terms if credit is issued this week. Waiting on Finance approval.
Every email arrives pre-read.
kriyaflo writes an intelligent summary for every email and every thread the moment it lands — what they want, what's changed, what's blocking. Your team scans a queue of one-liners instead of opening messages one by one, and long back-and-forth threads collapse into a single "here's where this stands."
- Email summaries — the ask, in one line, before anyone clicks
- Thread summaries — a 14-reply negotiation compressed into where it stands now
- Everywhere you triage — queue rows, board cards, context panel
- Per-inbox control — turn summaries on or off per inbox, and customize the summary prompt to surface what your team needs (amounts for AP, order numbers for support)
Thanks for confirming — we've approved the $12,400 credit against INV-2417 and it will post to your account this week. I'll send the updated statement once it clears.
Nothing leaves the inbox without a person pressing send.
AI-drafted replies. Human-sent, always.
Open an email and the draft is already there — a ready-to-send reply written from the full thread context by your own AI. Press Tab to drop it into the composer, edit if you want, and send. kriyaflo never sends email on its own: every reply crosses a human's hands first.
- Drafted on open — the suggestion is waiting when you get there
- One keystroke to compose — Tab accepts the draft; edit freely or send as-is
- Never auto-sent — the AI drafts, your team decides
- Provenance on every reply — each sent email is recorded as AI-drafted, AI-edited, or human-written
- Your AI, your tone — drafts come from the model you configured for that inbox
Your inbox never sleeps. Now neither does your triage.
Emails don't wait for business hours — customers write at midnight, vendors ship from other time zones, and Monday morning starts with 200 unread. kriyaflo continuously reads, classifies, prioritizes, and routes incoming email so your team can start with what needs attention — nights, weekends, holidays, and while your team is out of office.
Vendor escalation arrives
Ingested, deduped against the thread, sender domain verified.
Outage report — Node 04
Classified urgent · assigned to IT Ops · SLA clock started
Refund follow-up, second touch
Escalated priority, Jira ticket opened, owner notified.
Queue already ordered
18 emails triaged overnight — top of queue is what matters most.
3:47 AM, handled.
An outage report for Node 04 lands overnight — classified urgent, assigned to IT Ops, SLA clock started. It's the first thing on the queue at 9:00, not a lucky find at 2 PM.
Monday morning, pre-sorted.
No more weekend-backlog archaeology. Walk in to a queue already ordered by what actually matters, each email with a one-line summary.
Out of office ≠ out of the loop.
When the owner is away, nothing sits invisible — it's routed, tracked, and its SLA is still ticking where the whole team can see it.
The AI works the night shift. Your team keeps their evenings.
No black boxes. Every decision has a receipt.
Open any email and see exactly what the AI saw — the prompt, the model, the confidence per attribute, and every rerun. When an auditor, a manager, or your own gut asks “why is this marked urgent?”, the answer is one click away.
- Full prompt + response stored for every classification run
- Per-attribute confidence scores, not a single vague “AI score”
- One-click rerun with updated rules — history preserved, human edits untouched
- Complete event log: who (or what) changed anything, and when
[L1 org policy] + [L2 billing inbox] Classify the message below. Return category, priority, sentiment, owner…
Build rules in plain English. No code, no scripts.
Describe what to look for, decide what to set, and pick what to do. kriyaflo handles the pipeline behind the scenes — you just author the logic a non-technical operator can read and change.
Describe the trigger in plain English
No regex, no SQL, no scripting. Write what you want to match — sender, subject, body, thread history, attachments, or any combination — and kriyaflo turns it into executable conditions.
- Natural language conditions with AND / OR
- Match across message, thread, and attachments
- Reference custom business data or prior labels
"If body contains 'invoice' OR 'billing' AND sentiment is negative, set priority to Urgent and route to Finance."
Choose what the AI should decide
Pick the attributes you want set automatically: category, priority, sentiment, owner, tags, confidence threshold. Each attribute can be rule-driven, AI-driven, or locked to a human override.
- Per-attribute confidence thresholds
- Source trace for every decision
- Human overrides always win
Map decisions to real actions
A rule doesn't stop at a label. It fires the next step — a ticket, a channel post, a task, or a webhook — with the email context pre-filled so the downstream tool already knows the story.
- Jira ticket, Slack post, Asana task, webhook
- Context auto-filled from email and thread
- Multiple actions per rule
Jira is live today. Slack, Asana, and Notion actions are on the roadmap — built on the same framework, so adding them doesn't require re-platforming your rules.
Preview before you publish
Test the rule against live or sample emails. See exactly what would be classified, who would be assigned, and which actions would fire — before the rule touches a real inbox.
- Dry-run against recent emails
- Simulated action output
- Confidence and source for each attribute
Deploy per inbox with layered policies
Apply the rule globally, to one inbox, or to a specific alias. Layer org-wide policies on top, then inbox-specific overrides — so support@ and billing@ can share a base and still diverge where it matters.
- Global + per-inbox policy stacks
- Draft / publish / rollback versions
- Changes apply without touching your email server