Skip to content
AI customer support

Resolution,not deflection.

Craftlumen reads every conversation the way your best rep would — full account context, the right tool, a real answer — and only hands off the ones that actually need a person.

14-day trial on your real queue · shadow mode available · connects in minutes

Live chat · Order #48211Live
Hey, my order hasn't arrived and it's been 9 days. Can I get a refund?
AGENTLooking it up — it shipped Sept 3rd and the carrier's tracking stalled in transit on the 6th.
AGENTI've issued a full refund of $86.40 to your original payment method. You'll see it in 3–5 business days.
Resolved in 41s · no handoff
Resolution rationale94% conf.

Refund approved automatically — tracking shows no movement for 3+ days past the estimated delivery date, which policy covers.

Median first response
4.1s
Resolved without a human
68%
Channels connected
14
Conversations this week
128,400

Supporting 2.1M customers across 31 markets

NORTHBEAMD2C retail · AustinLoopcartE-commerce · TorontoFERNWAYMarketplace · BerlinDriftline & Co.SaaS · LondonCADENCESubscription · SingaporeHavenlyE-commerce · Sydney

Figures are aggregated across active customers on paid plans and refresh quarterly. Individual results vary by ticket volume and policy complexity.

How it works

Watch it decide.

Every case runs the same four checks — classify, look up, check policy, act — and shows its work. Nothing here is scripted per demo; it's the same trace a live case produces.

Decision trace41s end-to-end
  1. Intent classified0.3s

    Refund request · 94% confidence

  2. Order lookup0.8s

    Order API · shipped Sept 3, tracking stalled Sept 6

  3. Policy check1.1s

    Within 30-day refund window

  4. Refund issued1.6s

    $86.40 to original payment method

  5. Sent41s

    CSAT 5/5 · no handoff

Integrations

Where you already work.

Craftlumen sits inside your existing stack rather than replacing it — it reads and writes through the tools your team already uses every day.

Team chat

  • Channel-based alerts
  • Slash-command lookups

Helpdesk

  • Two-way ticket sync
  • Macro-aware replies

Live chat

  • Widget hand-off
  • Typing-state mirroring

Email

  • Shared inbox triage
  • Thread-aware replies

CRM

  • Order & account lookup
  • Customer history

Webhooks / API

  • Custom tool calls
  • Event streaming
Features

Built to finish the conversation.

Not a bolt-on chatbot. Every piece here exists to get a real conversation to a real close — or to hand it off cleanly when it shouldn't.

Resolves end to end

Not a router. Craftlumen reads the account, calls the right tool, and sends an answer a customer can act on — the same three steps a good rep takes, every time.

Can I get a refund for order #48211?
Refunded $86.40 — you'll see it in 3–5 business days.

Synced with your knowledge base

Re-indexes your help center and policy docs on every change.

28 languages, one voice

Replies in the customer's language, in your brand's tone.

Policy guardrails

Never promises a refund, credit or exception your policy doesn't allow.

Resolution analytics

CSAT, deflection rate and time-to-resolve, broken down by intent.

Hands off with the whole context

When a case needs a person, it arrives with the transcript, the reasoning, and a drafted reply already attached — never a cold ticket.

EscalatedBilling team
  • ✓ Full transcript attached
  • ✓ Reasoning attached
  • ✓ Draft reply attached
Customers

Run by teams who measure deflection honestly.

Craftlumen is used by support teams who report their real numbers upward — and need those numbers to hold up.

We didn't want a bot that says “i'll connect you with someone.” We wanted fewer tickets reaching a human at all. This is the first tool that actually does that.
Marcus Ferreira
Head of Support, Fernway
68% deflection
The handoffs are the tell. When it does escalate, the ticket already has the reasoning and a draft attached — our team just approves or edits.
Priya Nadarajah
VP Customer Experience, Northbeam
9s to handoff
It refuses to promise things our policy doesn't allow, which sounds small until you've been burned by a bot that overpromised a refund.
Owen Castellano
Support Ops Lead, Loopcart
0 policy exceptions
Same tone, 12 languages, no separate playbook for each. That alone paid for the subscription in the first quarter.
Ines Vaduva
Director of CX, Cadence
12 languages live
0.0M
Tickets resolved
0%
CSAT maintained
0s
Median resolution
0
Policy violations logged
Pricing

Priced for outcomes.
Not for headcount.

No per-seat fee, no minimum contract, no charge for agents who never touch a ticket. You pay for conversations that actually close.

MonthlyAnnual2 months free

Starter

For a small team testing the waters.

$40.83/month

Billed $489.96 annually

500 resolutions / mo · $0.15 per resolution after

Save $98.04 a year

Start free trial
  • 1 channel connected
  • Email + live chat widget
  • Business-hours coverage
  • Basic resolution analytics
Most popular

Growth

For teams running support as a real function.

$207.5/month

Billed $2,490 annually

5,000 resolutions / mo · $0.08 per resolution after

Save $498 a year

Start free trial
  • Unlimited channels
  • Every integration
  • 24/7 coverage
  • Priority routing + guardrail rules
  • Full CSAT & deflection analytics

Enterprise

For multi-brand or high-volume support orgs.

Custom

Scoped to volume and integrations

Unlimited resolutions

Talk to us
  • Everything in Growth
  • SSO & SCIM provisioning
  • Audit log & data residency options
  • Dedicated success manager
  • Uptime SLA
  • No cost per seat — everyone on your team can use it.
  • Resolutions are counted only when a conversation closes without a human reply.
  • Cancel anytime; your data export is always free.
Questions

What teams
actually ask first.

Still deciding? Our setup team will connect it to your real queue on a call and show you what it would have resolved last week.

Book a setup call
Setup team
Avg. setup call
22 min
First reply time
< 2 hrs

Engineers, not a ticket queue. Weekdays 07:00–20:00 UTC.

It resolves them. Craftlumen looks up the order, account or invoice through your connected tools, applies your policy, and sends an answer the customer can act on — a refund, a credit, a corrected shipping estimate. Routing is what it does when a case genuinely needs a person, not the default.

Every action is checked against your policy rules before it's taken, not after. If a refund amount, a promised date or an exception isn't explicitly allowed by your rules, Craftlumen drafts the case for a human instead of guessing. It cites the order, ticket or policy line behind every claim it makes.

It hands off — with the full transcript, its reasoning, and a drafted reply already attached, so your team edits or approves instead of starting cold. About a third of conversations end this way today, and that ratio is visible in your analytics, not hidden.

Live chat, email, and your team chat for internal alerts, plus your helpdesk, CRM and billing system through native integrations or a webhook. If a tool exposes an API, it can usually be added as a custom action.

28 at general availability, detected automatically per conversation. Replies keep your brand's tone across all of them — it's trained on your own resolved tickets, not a generic script translated on the fly.

Most teams are live within a day: connect your helpdesk and knowledge base, set your policy rules, and run it in shadow mode alongside your human team until you're comfortable switching it on for real replies.

Yes. The 14-day trial connects to your actual tools and can run in shadow mode the whole time — it drafts responses without sending them, so you can see exactly what it would have done before anything goes to a customer.

Start free trial

See what it would have resolved last week.

Fourteen days on your real queue, shadow mode available, no card. Most teams see their first clean resolution the same day.

14 days on your real queue · shadow mode available · no card required · cancel anytime.