Guides

Automating support on Gorgias, explained

Plain-English guides to running an AI support agent you can actually trust — the pricing models, the metrics, and the rollout choices that decide whether automation saves you money or quietly costs you customers.

Comparison

Glassdesk vs per-resolution AI support agents

Flat pricing vs $0.90–$1.50 per resolution, with real published rates for Gorgias AI Agent, Fin, and Zendesk — and an honest look at where each one is ahead.

Read the comparison →
Pricing

Flat-rate vs per-resolution pricing: the real math

Where's the break-even? The worked numbers on when a flat plan beats a per-ticket meter — and the hidden costs (double-billing, overages) to check first.

Read the guide →
Pricing

Gorgias AI Agent pricing: the full cost breakdown

$0.90 a resolution sounds simple until you find the overage rate, the bundle allowance, and the case where one conversation bills twice. Every component, with the arithmetic.

Read the breakdown →
Buyer's guide

Best AI agents for Gorgias: an honest comparison

Six tools compared on pricing model, auditability and Shopify depth — including, plainly, where each of the other five beats us.

Compare the six →
Buyer's guide

5 Gorgias AI Agent alternatives, honestly compared

Flat-rate, pay-per-task and enterprise options next to the built-in meter — each vendor's own published pricing, linked and dated, including where they beat us.

See the five →
Buyer's guide

Yuma AI alternatives: 5 honest options

Yuma doesn't publish pricing. Five options that answer the three reasons people look — price you can read, self-serve setup, auditable replies.

See the options →
Pricing

Freshdesk Freddy AI pricing: what a session really costs

500 sessions included, then $49 per 100 — plus the $29/agent Copilot add-on. Every line priced out, and the one term Freshworks never defines on its own pricing page.

Read the breakdown →
Buyer's guide

Best AI agents for Freshdesk: an honest comparison

Freddy, Fin, eesel and us — published rates only, priced out at 1,000 tickets a month, with the caveat that nobody agrees on what a billable unit is.

Compare the four →
How-to

Shadow mode: test an AI agent without risking a customer

Real tickets, private notes, zero customer exposure. What to measure over two weeks — and why a high escalation rate means your docs have a gap, not your AI.

Read the guide →
Benchmarks

AI deflection rate benchmarks: what the numbers mean

Why a deflection rate is meaningless without a denominator — and our own published eval results, with the full case mix attached.

See the numbers →
Explainer

What an AI confidence score actually means

A second model grades the first one's answer against your sources. That is the whole mechanism — and no, it is not a probability. The four questions that expose a decorative score.

Read the explainer →
Explainer

Your escalations are a ranked list of the articles you're missing

Deflection rate is the less useful half of the data. How to read unanswered tickets as a documentation backlog — and why "missing" and "unclear" need different fixes.

Read the explainer →
Explainer

Multilingual support: which languages we actually tested

Every vendor claims 100+ languages and publishes nothing. Here are our six ES/DE/FR eval cases in full — including a French prompt-injection attempt — plus the languages we have not tested.

See the test cases →
Explainer

"Where's my order?" is the one question an AI must not improvise

The most common Tier-1 ticket is also the one where a fluent guess does the most damage. What we read from your Shopify store, the four things we deliberately cannot do, and what happens when the lookup fails.

Read the explainer →
From real data

The smallest help centre answered the most questions

Three real DTC help centres, 15 Tier-1 questions, one retrieval pipeline. 13 pages beat 59 — and one store's warranty page was cited at 0.94 confidence while the warranty question still failed.

See what we found →
Free tool

Gorgias AI Agent cost calculator

Put your own monthly resolutions in and see the per-resolution bill next to a flat one — including the volume where per-resolution is genuinely the cheaper choice.

Run the numbers →
More on the way. Next up: what a full help-centre rewrite changes, measured on the same 15 questions. Want one sooner? Tell us.

The AI support agent you can see through

Every reply cited and confidence-scored, one flat price from $79, live on Gorgias in an afternoon. 14-day free trial, cancel anytime.

Start my flat-rate trial