How Relay works, from first message to resolved case

Relay is a customer care platform for AI and human support. It puts every customer conversation in one inbox, lets AI agents answer from your own policies within rules you set, and hands judgment calls to your team. Here is how it works, step by step.

1. Every channel lands in one queue

Relay brings private messages and public posts into the same queue:

• Email, WhatsApp, Telegram and calls.
• X, Instagram, Facebook, Reddit, YouTube and Quora.
• Google News, Google Business reviews, the App Store and the Play Store.

A complaint on X, an invoice question by email and a one-star Play Store rating sit side by side, each with tickets, routing, SLA clocks and an event log. Where a platform limits what software can do, such as LinkedIn messages, Relay shows that limit instead of hiding it.

2. Relay learns your business

Before it answers anyone, Relay is given context:

• Your policies. You upload playbooks, refund rules and brand voice as files, and Relay indexes them.
• The customer. History, tier and past cases.
• Your operations. Categories, regions, hubs, dealers and SLAs, so it knows who owns a case.
• The moment. Live themes, such as a citywide delay, that change what a good answer looks like.

That is why a reply can say "you can change the address once before dispatch" instead of "please contact support".

3. AI agents do the routine work

Relay runs a small team of agents:

• Triage Agent. Reads sentiment, sarcasm, category and urgency.
• Reply Drafter. Writes a reply grounded in the customer's history and your documents.
• Sarcasm and Tone Guard. Stops an automatic thank you going to a sarcastic five-star review.
• Escalation Router. Moves a case from L1 to L2 to a senior person.

4. You decide how much they do alone

Each agent runs in one of three modes: suggest-only, draft for approval, or autonomous. The playbook sits on top of that and locks categories to a person. For example:

• Refund, AWB or payment issues: person only.
• Perishable damage: a person, plus a claim.
• Late delivery: draft, then approve.
• Praise: acknowledge.

A person stands between the AI and anything that costs money.

5. Cases go to the right owner, with a clock

Routing follows your operations. A dealer complaint in Riverside goes to the Riverside dealer desk. A delay in the west region goes to the West hub. Each case has an SLA timer, and cases close to breach are escalated to a senior person.

6. Crisis alerts explain themselves

Relay groups mentions into themes and learns the normal volume for each. When a theme spikes, it scores the spike on six signals:

• Deviation from the theme's own baseline.
• Velocity.
• Concentration.
• Amplification.
• Sentiment intensity.
• Spread across platforms.

Every alert says why it fired and goes to the theme's owner. An incident board gives each live theme one owner, and an executive watchlist can track mentions of named people with their own alert rules.

7. Feedback and the contact centre

Relay also covers what happens after a case closes. It can raise a callback from a public complaint, send a CSAT survey after resolution and run NPS feedback with review requests.

8. Reporting leadership can read

The command center shows first response time, turnaround, SLA performance, resolution, reopen rate and share of voice against rivals. Alerts reach Slack, email or WhatsApp, and data flows out by REST API, webhooks, CSV export and reports.

9. Your data, your rules

Relay keeps data in your own database. Views are role based, every ticket has an event log, and customer-facing actions follow the approval rules you set. A Thinkiyo team sets Relay up with you, connects your channels and stays on after launch.

Who uses it

• Care teams get one queue, an owner and a clock for every case.
• Brand and PR watch themes and the executive watchlist.
• Dealer networks get complaints routed to the right location.
• Leadership sees SLA, resolution and reopen rate.

See how Relay handles this

relaycustomer careproduct guide