How We Automated Shopify Customer Service, Instagram DMs, and Blog Content With Koira

An honest look at the AI workflow tool we’ve been running quietly behind the scenes for the last few months — what it does, what it doesn’t, and why we picked it over hiring an agency.

If you’ve ordered from us recently and got a fast, accurate reply on Instagram at 11pm — that wasn’t us at the desk. It was Koira, the self-driving AI tool we use to handle most of the customer service, social, and blog workflow at SLR Speed.

We’ve had a few customers and a few other Shopify store owners ask us how we’re keeping up with response times, posting frequency, and Google ranking without growing the team. The honest answer is one tool doing roughly four jobs. This is a write-up of exactly what we’ve automated, what the workflow looks like end-to-end, and the trade-offs we’ve found in the real world.

The four things we automated with Koira

Before we found Koira, four jobs were eating our week:

1. Customer emails — Gmail and Shopify Notifications inbox combined, roughly 80–120 messages per day during a launch week, most of them “does this fit my setup?” or “where’s my order?”
2. Shopify Inbox messages — the chat widget on the storefront. Live questions where the customer is mid-checkout and wants an answer in seconds, not hours.
3. Instagram DMs and comments — peaks around drops and reels going up. Hundreds of messages, half of them duplicate questions about availability or compatibility.
4. SEO blog posts — long-form content for Google ranking. We knew we needed to be publishing twice a week to compete with the bigger names, and we were barely managing twice a month.

Adding a full-time customer service hire was on the whiteboard. So was hiring an agency to write blogs at $4–$6K/month. We did the math on both and they didn’t work for where we were. We needed something cheaper that actually got the work done.

Why Koira and not Zapier, ChatGPT, or an agency

Quick context on why we didn’t go with the obvious options:

• Agencies charged $4K–$15K/month for a writer or a social manager who didn’t know our products. The on-ramp was 4–6 weeks of feedback loops before they wrote anything that sounded like us.
• VAs were cheaper ($2K–$4K/month) but had the same context problem — they’d either answer questions wrong about fitment or escalate everything back to us, which defeats the point.
• Zapier + ChatGPT sort of works for two-step automations like “new email → AI drafts a reply.” But the moment we wanted to actually look up the Shopify order, check the carrier’s tracking, and reply in the customer’s original channel, we were stitching together five tools and breaking every other week.
• Other AI inbox tools charged $0.15–$0.80 per response because they re-take screenshots and re-prompt AI for every single message. At our volume that’s $22–$120/month per workflow before any AI writing cost — and they only do one channel.

Koira sits in a different category. It’s self-driving software that runs the same websites we use by hand. Sign in once to Shopify, Gmail, Instagram, our blog admin, our shipping tool — and it takes over from there. No API setup. No integrations to configure. If we can use the site by clicking through it, Koira can.

Workflow 1 — Shopify customer emails (Gmail + Shopify Notifications)

This is the workflow we trained first. Most of our inbound email is one of four patterns: where’s my order, does this fit, return request, custom request. Each one Koira handles end-to-end across multiple sites:

Where’s my order:

Gmail → Koira reads the email → Pulls the customer’s actual Shopify order → Hits the carrier’s tracking page → Drafts a reply with the live status → Queues it for our approval → Sends.

That whole chain used to be a tab-juggle that took 4–7 minutes per email. We had to switch from Gmail to Shopify admin to the carrier’s tracking site to a draft reply and back. Koira does the same sequence in about 20 seconds and we just tap approve when the draft looks right.

About four weeks into running it, we turned approval off for shipping-status replies — the answers were always correct and the customers liked the speed. Now those go out fully autonomous and we just see them in the activity log.

Does this fit my [setup]:

This was the killer one. Compatibility questions are where customers buy or bounce, and we used to lose them to slow replies. Koira reads the product the customer is asking about, cross-references it against our compatibility data (we point it at a shared doc), drafts a yes/no answer with the relevant SKUs, and queues it. Response times went from 6–14 hours overnight to about 4 minutes.

Workflow 2 — Shopify Inbox (storefront chat)

Shopify Inbox is the chat widget on our storefront. The bar there is harder — the customer is mid-checkout, they want a 30-second reply or they’re gone.

We pointed Koira at it and gave it the same product context. It now handles the first reply on every conversation. Most go end-to-end without us. The ones that need us — usually anything custom or anything where the customer sounds upset — Koira pulls into our approval queue with a flag, so we see them inside a minute.

Our Shopify Inbox response time dropped from a weekly average of 2h 18min to under 90 seconds. Conversion rate on chat-initiated sessions jumped about 31% over the following six weeks. We weren’t expecting the conversion lift — that came from just being there when the customer asked.

Workflow 3 — Instagram DMs and comments

Instagram is where most of our discovery happens. Reels go up, people DM, half the messages are “link?” or “available?” or some variation. The other half are real conversations we want to be in.

Koira handles both. For the routine asks, it reads the DM, figures out which product the customer is referring to from the reel they replied to, sends the link (with our affiliate parameter for tracking), and replies on the comment thread if there are public asks too. Same workflow runs on Instagram Stories replies. It even handles the brand-collab DMs — sends our media kit, quotes our rate, books a call on our Calendly. We just see the booked call on the calendar.

For the “real conversations” Koira surfaces them in our queue with a one-line summary and we step in. The win is we now reply to every DM the same day instead of letting them rot in the requests folder.

Workflow 4 — SEO blog posts

This is the one investors get most excited about, because the math is unreal.

We publish three SEO blog posts a week now. Before Koira we published two a month. We didn’t hire anyone. The flow is:

We give Koira a topic (or it picks one based on Google Search Console gap analysis) → Researches it across reference sites → Drafts the post in our voice (it’s read every previous post we’ve published on slrspeed.com, so the tone matches) → Generates the featured image with AI → Drafts the meta title and description with our target keyword → Queues for our approval → Publishes to Shopify Blog with the right tags and category.

Drafts cost us roughly 8 minutes of editing each instead of 4 hours of writing. Our organic traffic is up 47% in the four months since we started. Google Search Console’s indexed-pages count went from 184 to 412.

The biggest thing we didn’t expect: quality went up, not just quantity. Because Koira reads our entire site before drafting, it doesn’t contradict itself across posts the way a freelance writer rotating through topics did.

What it actually costs us

The math against the alternatives we considered:

We’re on the Koira Pro plan at $199/month. Per-run cost on emails, DMs, and blogs is zero because it’s unlimited at that tier. The AI runs on the platform’s key for training and through our own connected key for runtime, so we’re not paying per-message like we were with the other AI inbox tools we trialed.

The setup actually took an afternoon

This was the part we were most skeptical about. Every other tool that promised “easy automation” took us a week of meetings to get configured. Koira was honestly an afternoon:

1. Signed up at koira.ai, picked the Pro plan.
2. Downloaded the desktop app (Mac, in our case) and signed in.
3. Pointed it at the sites we wanted it to use — Shopify admin, Gmail, Shopify Inbox, Instagram, our blog editor. Signed into each one in their browser the same way we always do. Koira saw the session and remembered it.
4. Described each workflow in plain English — “answer customer emails about shipping by looking up the Shopify order and the carrier’s tracking, then replying in Gmail.” That’s literally the prompt we gave it.
5. It figured out how to run each one. We approved a few drafts to teach it our voice. After about a week it was sounding exactly like us.

There’s no API integration. No webhooks. No Zapier middle layer to break. It drives the same browsers we do, sees the same buttons we see. That’s why it works on sites that don’t have APIs (looking at you, Instagram).

The honest downsides

To keep this balanced, the parts that aren’t perfect yet:

• Your computer has to be on for the local browser tasks. For us that’s fine — we have a computer always on. But it’s worth knowing.
• The first week needed babysitting. Mostly because we hadn’t pointed it at all our context yet. Once we gave it our shipping policy doc, our compatibility chart, and a folder of past customer emails, the accuracy went up sharply.
• It will flag things it’s not sure about instead of guessing. That’s a feature for us but at first it felt like “why are you bothering me with this?” The answer was almost always “because a wrong answer here would have cost you money.”

What we’d tell other Shopify owners

If you’re a Shopify store doing more than $30K/month and you’re drowning in customer service while trying to also do social and SEO — this is the closest thing to hiring four people without the cost. The pitch isn’t “AI replaces your team.” The pitch is “the boring repetitive stuff that doesn’t need a brain runs itself, and you spend your time on the parts that do.”

Things we’d recommend if you’re thinking about it:

• Start with one workflow. We started with shipping replies because they were the most repetitive and the lowest risk. Once that was running clean we added the next.
• Keep approval-required on every workflow for the first 2 weeks. You’ll catch the cases where it’s missing context, and the next iteration improves.
• Give it your style. Drop in your shipping policy, your compatibility chart, past examples of how you answer questions. It reads those and matches your voice.
• Look at the activity log every Friday. We do a 10-minute review to see what went out, what got escalated, and what we’d change. Total weekly overhead.

Closing — the platform behind it

We’ve been quietly running Koira in production for a few months now, and we’d be happy to vouch for it to any Shopify owner reading this. The team at Koira (koira.ai) has been responsive, the product keeps shipping new features (we got cross-site flows last week, which let one workflow touch Gmail, Shopify, and Shippo in a single sequence), and the pricing is fair.

If you want to try it, they have a free tier. You can sign up at koira.ai and have a workflow running on your own Shopify store the same day. No code. No API project. Just describe what you want automated.

That’s the entire pitch. The boring busywork runs itself. You go ride.

— SLR Speed
SLR Speed is not an investor in Koira. We pay for our subscription like any other customer. We’re sharing this because the tool genuinely changed how we operate and we’ve been asked enough times to write it up.