( fernwood — ai support engineer )
trained on 40M+ resolved tickets, tuned per workspace
( 01 · model card )
( capabilities )
tickets triaged monthly
average deflection rate
380ms median latency
Faster than a human can open the ticket.
/ standard
SOC 2 Type II
Every conversation encrypted in transit and at rest, with per-workspace data isolation and zero model training on customer content by default.
( 02 — how fernwood works )
( ingest )
Fernwood reads your knowledge base overnight
Connect Zendesk, Intercom, or a help center in one click. Fernwood indexes docs, past tickets, and macros — no manual training data required.
( triage )
Every ticket gets scored before a human sees it
Intent, sentiment, and urgency are classified in under half a second, so your queue is already sorted by the time an agent opens it.
( resolve )
Routine tickets close themselves, safely
Refunds, password resets, order status — Fernwood resolves the repeatable 40% end-to-end, with guardrails your team defines.
( learn )
Every edit your team makes retrains the model
When an agent rewrites a draft, Fernwood learns the correction and applies it workspace-wide within hours, not release cycles.
( 03 — inside a ticket )
Fernwood drafts a reply in real time
The moment a ticket lands, a fully-cited draft appears in your helpdesk sidebar, sourced from your docs and past resolutions.
Confidence scoring flags what needs a human
Anything below your trust threshold routes straight to an agent with the draft pre-filled — never a blind auto-send.
One click sends, edits, or escalates
Agents approve, tweak, or hand off to engineering — Fernwood logs the outcome and adjusts its confidence model instantly.
Weekly reports show exactly what it saved you
Hours reclaimed, deflection rate, CSAT delta — a plain-English summary lands in your inbox every Monday.
( fernwood in production )
tickets resolved to date
average deflection rate
median first-response time
routing accuracy
( from support leaders )
“We cut first-response time from six hours to under a minute without hiring a single new agent.”
Priya Nandakumar
Head of Support, Lattice Robotics
01 / 04
( early access — closing soon )
We onboard 20 new workspaces a week, white-glove, with a dedicated setup call. Join the list and we will reach out within 48 hours.
no credit card required, cancel anytime
people already joined
( faq )
01
Which helpdesks does Fernwood integrate with?
Native integrations with Zendesk, Intercom, Front, and Help Scout. A REST API and webhook layer cover anything custom in under a day of engineering time.
02
Does Fernwood train on our customer data?
No. Each workspace gets an isolated fine-tuned layer on top of our base model — your tickets never leave your tenant and are never used to train shared models.
03
How accurate is the auto-resolve feature?
Auto-resolve only fires above a confidence threshold your team sets per category — most workspaces start conservative and expand coverage after week two.
04
What languages does it support?
English, Spanish, French, German, Portuguese, and Japanese out of the box, with drafts matching the customer's original language automatically.
05
How long does setup take?
Most teams are live within a day — connect your helpdesk, let Fernwood index your knowledge base overnight, then review its first drafts the next morning.
06
What does pricing look like?
Usage-based, priced per resolved ticket rather than per seat — so adding agents to review drafts never increases your bill.