Who this is for
Your store or service business has an inbox that never empties. Two or three people spend their day answering the same questions in slightly different words: where is my order, can I return this, do you ship to Australia, what are your hours. The real problems, the damaged parcel, the wrong item, the angry customer, wait behind them in the queue.
Support automation is for the business where the volume is the problem, not the difficulty. It takes the known cases off your team's desk entirely and turns the rest into pre-written drafts, so the people you have spend their time on the tickets that need a person.
What the automation does, in order
1. It reads and classifies every incoming message
Every ticket or email that arrives is read and labelled: order status, return request, shipping question, product question, complaint, wholesale enquiry, spam, and the categories specific to your business. The label is written back to your helpdesk as a tag, so your existing views and reports keep working. Priority and sentiment are set at the same time, so an angry message about a late order goes to the top of a person's queue rather than into the automatic lane.
2. It answers the cases with a known answer
For the ticket types you have agreed in advance, the automation looks up what it needs and replies:
- Where is my order: it matches the sender to an order in your store, reads the fulfilment status and the carrier tracking, and replies with the current status, the tracking link and your published delivery window. If the parcel is late against your own promise, it does not reassure; it routes to a person.
- Returns and exchanges: it checks the order date against your return window, confirms whether the items are eligible under your published policy, and sends the return instructions or label link. If your store supports it, it opens the return request for a person to approve.
- Shipping zones and times: it checks the destination against your shipping zones and rates table and replies with the options, costs and times you publish. Countries you do not ship to get a clear no and a note of any alternative you offer.
- Hours, locations, simple product facts: answered from your site and catalog, with a link.
Every automatic reply is signed as automated and offers a person in one click. Every one is logged with the data it looked up.
3. It drafts a reply for everything else
For tickets outside the automatic list, the automation writes a draft in your tone using your past replies, your policies and the order data, and leaves it in the ticket as an internal note or a saved draft. Your team reads it, edits if needed, and sends. A draft that would have taken eight minutes to write from scratch takes one to check.
4. It keeps a review queue
Anything the automation was not sure about, anything a customer replied to after an automatic answer, and a sample of automatic replies each day land in a review view inside your helpdesk. During the first 30 days we read this queue with you weekly. After that, it is yours, and most teams spend ten minutes a day on it.
5. It updates your systems
Tags, priorities, order notes and return requests are written back to the helpdesk and the store, so the record of what happened lives where it always has, not in a separate tool.
What it connects to
- Helpdesks and inboxes: Gorgias, Zendesk, Freshdesk, HubSpot Service Hub, and shared Gmail or Outlook mailboxes.
- Stores: Shopify and Shopify Plus, WooCommerce, Magento 2 and BigCommerce through their APIs. Custom Laravel, Node or .NET applications through your database or API.
- Carriers: through the tracking data your store already holds, so no extra carrier accounts are needed in most cases.
- Anything else with an API: a returns portal, a warehouse system, a subscription app. Each extra connection is priced separately.
The connecting service runs on your own cloud account. The model account is yours. You can revoke our access the day after launch and everything keeps running.
What stays human
Some tickets should never be answered by a machine, and we write that into the rules rather than hoping the model behaves:
- Money: refunds, partial credits, price adjustments, chargebacks, disputed charges. The automation can draft the reply and prepare the data; a person decides.
- Judgement: damaged or wrong items, late orders past your promise, complaints, legal threats, press and wholesale enquiries.
- Anything ambiguous: a message that mentions two orders, a reply to an earlier thread, a question the rules do not cover. These get a draft and a tag, not an automatic reply.
- Anyone who asks for a person: routed immediately, with the conversation attached.
The split between automatic and drafted is yours to move. Most owners start cautious and widen the automatic list after a month of reading logs.
A worked example
The numbers below are illustrative. They show what a typical first month looks like, not a result we are claiming for a client.
WooCommerce store selling replacement parts, Freshdesk helpdesk, three people on support. About 1,100 tickets a month: roughly 40% order status, 15% returns and eligibility, 10% shipping destinations, the rest product fit, complaints and trade enquiries.
- 1,100 tickets a month opened by a person; first reply averaging 14 hours; order status tickets answered by copying tracking numbers by hand
- Return eligibility decided from memory, with three different answers to the same question in one week
+ Automation live in week 4 after matching on a 240-ticket sample; order status, return eligibility and shipping zones on the automatic list
+ Month 1: about 460 tickets answered automatically, about 640 reaching a person with a draft and tags already set
+ Running cost in month 1: $120 model fees, $15 hosting, on the store's own accounts; the extra $600 over the base price was a second connection to the returns portal
What it costs
| Work | Includes | Fixed price |
|---|---|---|
| AI fit checkThe first step | One-page plan: which ticket types qualify for automatic answers, which get drafts, running cost, fixed build price | Free |
| Support automation | One helpdesk or inbox connection, one store connection, classification and tagging, automatic answers for the agreed types, drafts for the rest, review queue, 200-ticket sample test, 30 days of tuning | from $4,500 |
| Extra connectionEach | A second helpdesk, store, returns portal, warehouse or subscription system | $600 to $2,000 |
| Website assistant built alongsideShares the rules and store connection | Live answers on your site for the same question types, with hand-off into the same helpdesk | from $3,500 |
| Care planOptional, cancel any month | Monitoring, rule and connector updates when your helpdesk or platform changes, monthly log review, small changes, a named engineer | $250 to $1,000 a month |
Half the fee at start, half at launch, by wire transfer or PayPal. Model fees and hosting are billed by the providers to your own accounts; the fit check gives you a monthly estimate from your ticket count. We do not take jobs under $2,500.
How it runs
Free AI fit check (2 working days)
You tell us the helpdesk, the store platform and the monthly ticket count. If you can export a month of tickets, we classify them and show you which types are ready for automatic answers. You get a one-page plan with a fixed price and a running cost.
Scope and rules (1 week)
Written scope: the automatic list with a rule for each type, the hand-off conditions, the tone sheet, the review queue design, and the sample test target. Model and cloud accounts created in your name. Half the fee paid.
Build and sample test (2 to 3 weeks)
Connections built against a staging copy of your store and a sandbox of your helpdesk. The automation is run against 200 or more past tickets and its output compared with what your team did. Daily written log.
Launch in draft mode, then automatic (1 week)
Live first with everything as drafts, so your team sees what it would have sent. Then the agreed types switch to automatic. Second half of the fee at launch.
Tuning (30 days)
We read the review queue and logs with you weekly, correct rules and tone, and widen or narrow the automatic list. Then the care plan, or nothing further.
Start with the fit check
Get a free AI fit check: tell us your helpdesk, your platform and your monthly ticket count, and get a one-page plan in two working days. Or email info@riantsoft.com with the subject line of the ticket you see most.
Questions people ask
Will customers know they are talking to an automation?
Automatic replies are signed as automated, in the wording you choose, and every one carries a one-line way to reach a person. We do not recommend hiding it. In our experience customers care about a correct answer in minutes far more than about who wrote it, and an honest signature avoids the complaint that follows when someone discovers a bot pretending otherwise. Drafted replies are sent by your staff under their own names, because a person has read and approved them.
Which tickets get answered automatically on day one?
Only the ones on the agreed list, usually order status, tracking, return eligibility and instructions, shipping destinations and times, opening hours and simple product facts from your catalog. Each type has a written rule: what data to look up, what the reply must contain, and when to stop and hand off. Everything else gets a draft. After 30 days of reading the logs together, you can move more types onto the automatic list, and you can take any type off it in one setting.
How accurate does it have to be before launch?
Before launch we run the automation against a sample of 200 or more past tickets from your helpdesk and compare its classification and replies with what your team actually did. Automatic reply types go live only when they match on the sample to a level you sign off, and the rules say "draft, do not send" whenever the message contains anything the rule does not cover. The first weeks run with a higher share routed to drafts, and we widen the automatic set as the logs justify it.
Can it write in our tone of voice?
Yes. We give it your existing macros, a few dozen replies your team is proud of, and a short style sheet: how you greet, how you sign off, which words you avoid. Drafts and automatic replies follow that. During tuning you mark replies that sound wrong and we adjust. It will still not promise anything you have not published, however friendly the tone.
Does this cover live chat and phone as well as email?
Email and helpdesk tickets are the core, because that is where the volume and the delay are for most stores. Live chat on your site is better served by the website assistant, which can share the same rules and the same store connection, and the two are often built together. Phone is out of scope; we do not build voice systems.
What does it cost to run each month?
Model fees depend on volume. A store with 1,000 tickets a month, each classified and either answered or drafted, typically sits in the range of a few tens to a couple of hundred dollars a month on your own provider account. Hosting for the connecting service is usually under $20 a month. Your helpdesk and store plans are unchanged. The fit check includes a running cost worked out from your actual ticket count.
We do not have a helpdesk. Can it work from a shared Gmail or Outlook inbox?
Yes. The automation can read a shared Gmail or Outlook mailbox, label and reply from there, and keep drafts in a folder for review. A helpdesk makes the review queue and reporting easier, and if your volume is over a few hundred messages a month we will say so in the plan and name one or two options that fit your platform. We will not sell you one as part of the build.