Runtime as featured inForbesRead the article

Supporting customers

Every escalation, already investigated.

Agents answer customers from the payment record. Technical cases escalate to agents with more access, which read your logs, code, and database like an engineer would. Refunds and account changes stay with your team. One harness, running in your cloud.

Trusted by teams running mission-critical operations.

RainNixtlaMonoLineLeap
Search
support-escalations
MessagesCanvasFiles
Today

APP 10:02 AM

Lumen Payroll says their webhooks stopped at 9:40. Their keys and endpoint check out, and the API logs are outside my access. can you take it?

Zendesk #48213, with my account checks attached.

APP 10:06 AM

Release 4.18 cut webhook retries from 5 to 1, so 412 events failed on Lumen’s slow endpoint. 3 other merchants are affected.

Logs, error trace, and the commit attached.

Approval needed

Open a PR restoring 5 retries

👀 2
DK

Dev 10:09 AM

Good find. let the 4 merchants know their events will replay once it ships.

Message #support-escalations

Aa

Every channel, every tier

Customers reach a support agent by WhatsApp, text, email, or phone. When a case needs more access, the agent escalates it in Slack to an engineering agent that can read logs and code.

@mention an agent in any channel or thread. Click a channel or an agent in the window to explore.

Payment questions aren’t FAQ questions.

Customers ask about declines, held funds, and transfers that haven’t landed. The answers live in systems your CX team can’t see, so tickets bounce between support, ops, and engineering.

The answer isn’t in the help desk

The help desk has the conversation. The processor, the ledger, and the logs have the answer.

Escalations wait in line

Support can’t tell what needs engineering, so cases arrive without context and sit in a queue.

Wrong answers cost trust

Business customers notice a confident, wrong reply. Every answer has to match the record.

One harness, one view of every customer.

The same customer can open a ticket, end up in an engineering escalation, and bring it up again in their account review. On Runtime, all three teams work from the same memory and the same record.

  1. 01

    One memory across teams

    A payout traced by support is already in context when the account manager hears about it.

  2. 02

    Answers your team can check

    Every reply cites the records, queries, and policies behind it, so reviewers can verify it before it’s sent.

  3. 03

    Skills that sharpen with every case

    Agents learn from resolved tickets and propose updates to their own playbooks and macros.

Autonomy you grant, step by step.

Start with drafts. Refunds and account changes always stay with your team.

  1. 1

    Investigate

    Agents trace the payment.

    Your team · Writes every reply

  2. 2

    Draft replies

    Each reply cites its sources.

    Your team · Approves each reply

  3. 3

    Answer routine questions

    On topics you approve.

    Your team · Approves refunds

  4. 4

    Follow up

    Agents confirm the fix.

    Your team · Reviews trends

Customer data, with clear boundaries.

Read about our security
AuditedSOC 2Compliant

Scoped access at every tier

Front-line agents see the payment record, and escalation agents can also read logs and code. Each sees only what the requesting role may, and never another customer’s data.

Sensitive data stays private

Card numbers, account numbers, and SSNs are masked before any prompt, log, or reply. For the most sensitive work, run open-weight models inside your network.

Your cloud, your models

Run agents in your own AWS, GCP, or Azure account, or self-host the platform. SOC 2 audited, and we never train on your data.

Supporting customers: common questions

01Do we need to replace our help desk?
No. Agents start from the tickets in your current help desk, like Zendesk, Intercom, Salesforce, or Pylon, and read the payment systems behind them. Replies land back where your team already works.
02How is this different from a support chatbot?
A chatbot answers from help-center articles. Runtime agents work like support engineers: they check the customer’s actual payments, accounts, and logs within their permissions, and show the evidence behind every answer.
03Can agents reply to customers on their own?
Only where you allow it. Most CX teams start with drafts their team reviews, then let routine answers, like payout status, go out on their own once quality holds up.
04How do agents escalate?
When a case needs more than the payment record, the front-line agent hands it to an agent with more access, like one that can read logs and code. Engineering only gets the cases the evidence says need them, with the findings attached.
05How do we keep one customer’s data separate from another’s?
Tenant and role scopes on every data source and tool decide what each agent can see. Test them on past tickets before agents touch live conversations.
06What should a pilot look like?
One queue, one question type, historical tickets first. Measure handle time, escalation rate, and reply accuracy against your team, then expand as the results hold up.

Start with your most common question.

See agents answer a “where’s my money” ticket, hand engineering an escalation with the evidence attached, and draft a client update, each ready for your team’s review.