Ingest every signal.
Email, chat, product usage, billing events — pulled in continuously and threaded onto each customer's journey, not just when they reach out. Your AI agent always has the full context.
Early access — limited cohort
Customer support that runs on journeys, not tickets.
support-arc is an AI-native support platform organized around each customer's ongoing journey. The AI resolves the repetitive volume on first contact; your team focuses on the cases where judgment actually matters.
Onboarding a small group of teams now. We'll reply within two business days.
Watch it work
A minute inside support-arc.
A new message lands. The agent reads the journey, drafts a grounded reply, and decides whether to ship or hand off. No human touch on the easy ones — and a clean handoff on the rest.
0:00 / 1:00
Inbox · acme.io5 open
SK9:42
MT9:38
DR9:31
JLjust now
Select a journey
A new message just arrived. Watch the AI pick it up.
Inbox · acme.io5 open
SK9:42
MT9:38
DR9:31
JLnow
Jordan L. · acme.io · journey day 23
"Hi — trying to set up the Slack integration but the OAuth callback keeps timing out. We're on the Team plan if that matters."
Agent reasoning
Reading the prior journey9 events · 2 jobs · 1 prior conversation120ms
Identifying the jobSlack integration setup240ms
Searching knowledge3 sources · OAuth callback whitelist410ms
Inbox · acme.io5 open
SK9:42
MT9:38
DR9:31
JLnow
Agent reasoning
Reading the prior journey9 events · 2 jobs120ms
Identifying the jobSlack integration setup240ms
Searching knowledge3 sources matched410ms
Drafted reply
confidence —
Sent automatically · 0.9s
Inbox · acme.io4 open
SK9:42
MT9:38
DR9:31
JLnowresolved
Sara K. · enterprise · journey day 47
"This is the third time we've hit the export limit this week. We're paying for Enterprise — what's going on? Considering moving to a competitor."
Agent reasoning
Reading the prior journeyenterprise · 3 prior export tickets90ms
Detecting churn risk signal"considering competitor" · enterprise tier180ms
Below confidence thresholdjudgment call · routing to human220ms
Routing decision
confidence 62%
Repeat enterprise complaint with explicit churn signal. Drafted context, but a human should send.
Routed to Maya · with context attached
Resolved automatically
71%
High-confidence replies sent without a human in the loop.
Median first response
3.2s
From inbox arrival to reply landing in the customer's inbox.
Routed with context
29%
Handoffs to humans — no cold ticket, full journey attached.
The shift
Tickets were scaffolding for human bottlenecks.
Macros, queues, tags, SLA timers, round-robin assignment — every part of a modern support tool exists because humans are slow. When the AI is the primary actor, most of that scaffolding stops earning its keep.
Yesterday
A queue of tickets, worked by humans.
- Every email becomes a ticket with a status, owner, priority, and tag
- Agents type the same answers from saved replies
- First response time is the headline metric
- Context lives in someone's head and dies on vacation days
- AI is bolted on as a chatbot at the front door
support-arc
A journey per customer, with jobs the AI works.
- The AI resolves on first contact when it's confident — instantly
- Related messages across email and chat stitch into one job automatically
- Outcomes — effort, resolution, journey health — are the metrics
- Every event threads onto a living journey the AI re-reads each turn
- Humans review only the cases that need judgment
How it works
Three layers, working together.
Email and in-app chat in. Resolutions out. What happens in between — and what your team actually sees — is the interesting part.
AI works the job.
The agent identifies the underlying job the customer is trying to get done, gathers context from the journey and your knowledge base, and drafts a grounded reply. Clear and low-stakes goes out instantly.
Humans handle judgment.
A small, prioritized queue surfaces only the cases that actually need a person — uncertainty, escalations, patterns. Every human decision teaches the system, so trust grows category by category.
Who it's for
A narrow fit on purpose.
We've built support-arc for a specific shape of company. If this sounds like you, we'd love to talk. If it doesn't, we'd rather be straight with you than oversell.
Company
Modern SaaS teams with self-serve customers
Support team
A handful of agents handling a growing volume
Channels
Primarily email and in-app chat
Stack
Modern billing, an analytics tool, and documentation the AI can learn from
Champion
A support, CX, or operations leader who is technical and ready to rethink their stack
Probably not yet
Phone-heavy operations, highly regulated industries, or enterprise procurement cycles
See it
Try the agent on a real message.
Paste a customer message — or pick one of the examples — and watch the AI identify the job, draft a reply, and decide whether it's confident enough to send.
Customer message
support-arc agent
Pick an example or type a message, then run the agent.
Live demo · grounded only in the example knowledge base
From inbox to autonomy
In production in seven days.
No long migrations, no re-tagging years of history. We connect to what you already use and the AI learns alongside your team.
Day 1
Connect.
We hook up your inbox and chat, point the AI at your help center, and ingest the last 90 days of conversations as context.
Day 2–3
Shadow mode.
The AI drafts a response to every incoming message, but nothing is sent. Your team reviews drafts side-by-side and corrects course.
Day 4–6
Assisted mode.
The AI sends replies on the categories where confidence is high. Everything else still routes to a human, with full context attached.
Day 7
First auto-resolutions.
The agent is closing real conversations end-to-end. Your team is reviewing a small queue of judgment cases instead of grinding through volume.
Why we're building this
A note from the team.
We started support-arc because we kept watching support teams — sharp, thoughtful people — spend their best hours typing the same answer for the hundredth time. The AI is good enough now to handle that volume, and it should. What it shouldn't do is take judgment away from the people who are best at it.
If you're skeptical, good. We'd rather show you in your own inbox, on your own conversations, than make claims you can't verify.
Request early access — The support-arc team
Early access
Get on the early access list.
We're onboarding a small group of teams who want to put AI-native support into production and shape the product as we build it. Tell us a bit about your team and we'll be in touch.
What to expect
- Direct, hands-on onboarding.We'll work with you to get the AI set up against your real channels and knowledge base.
- A real line to the team.Regular working sessions with the people building the product. Your feedback shapes what ships next.
- Your data, exportable.Full CSV export of customers, journeys, jobs, and conversations whenever you want it.
- Trust grows with you.The AI starts cautious and earns autonomy as it learns from your team — never globally, never silently.
Prefer a quick call? Book a 20-minute walkthrough or write to hello@support-arc.com.
FAQ
A few honest answers.
How is this different from existing support tools with AI bolted on?
Most tools were designed around a queue of tickets worked by humans, and added AI as a chatbot at the front door. support-arc starts from the other end: the AI is the primary handler from day one, the data model is the customer's journey, and humans review only the cases where their judgment changes the outcome.
What happens when the AI is uncertain?
It says so. The case lands in your review queue with the AI's draft, its reasoning, and the journey context attached. A human approves, edits, or takes over — and that decision teaches the system, so trust grows category by category over time.
Can we get our data out if it doesn't work?
Yes. Full CSV export of customers, journeys, jobs, events, and conversations, with relationships preserved. Data portability is a feature — your data is yours.
Which channels do you support today?
Email and in-app chat at launch. Other channels are on the roadmap; if a majority of your volume is on phone, SMS, or social, we'll let you know honestly whether we're the right fit yet.
How does the AI take action in our other tools?
It reads broadly to gather context, but writes are conservative by default. Drafting replies and updating journeys happens automatically; anything that touches an external system — refunds, account changes, proactive outreach — is queued for human approval until you say otherwise.
How long does setup take?
Most teams are up and running in under a week. We'll connect your channels, point the AI at your help center, and start in a low-autonomy mode where every reply is reviewed — then relax that as the team gets comfortable.