Blog

Notes from building Trigli.

AI customer service, support automation, and what we learn from running a small SaaS. No SEO fluff. Just stuff that's actually true.

October 15, 202613 min read

Live Chat for a One-Person Business: What to Automate and When to Say Nobody Is Online

Live chat looks like the one channel a one-person business cannot run, because it implies someone is sitting there ready to answer. You are not, and you should not pretend to be. Here is how to run chat as a solo operator: what to let the AI handle, where it should stop, when to show the email-capture fallback, and why "no agents online, leave your email" beats a fake "an agent will be right with you."

October 1, 202612 min read

How to Write a Refund Policy an AI Can Actually Answer From

Most refund policies are written for one job: covering the business if something goes to a dispute. That is a different document from one an AI can answer a customer from. Here is how to rewrite the second kind, with before and after examples you can copy the shape of.

September 15, 202612 min read

AI in your real inbox vs a separate helpdesk: the case for replying from your own address

Almost every AI support tool asks you to do one thing before it helps you: move your support into a new inbox. A helpdesk address, a portal, a second place to check. For a small team, that migration is the most expensive part of the whole project, and it is easy to skip. Here is the case for connecting AI to the inbox you already run, when replying from your own address is the right call, and the specific situations where a separate helpdesk earns its keep.

September 8, 20269 min read

Per-Resolution AI Pricing: Run This Math Before You Sign

Intercom Fin currently charges $0.99 per resolved conversation. That sounds fine until your support volume spikes during a launch or a bug. I ran the numbers at three different volume levels so you can see the trap before you sign a contract.

September 1, 202613 min read

The Support Questions You Should Never Let AI Answer

Most advice about AI support argues over auto-send versus draft. There is a category underneath that argument: messages the AI should not answer at all, where a draft is not caution enough and the human should own the reply from the first read. Angry customers, exceptions to policy, and anything touching money. Here is why those three families are different, and the routing that gets them to a person.

August 25, 20269 min read

Measure AI Customer Support Quality: A Small-Team Guide

Most support metrics were built for human agents. When you add AI to your inbox, the old numbers stop telling the full story. This is a practical guide for founders and small teams who want to know if their AI is actually working - without an analytics engineer, a CSAT platform, or a per-agent dashboard.

August 15, 202610 min read

What "AI Reads Your Docs" Actually Means: Chunking and Retrieval

People upload a 60-page policy PDF, watch the AI miss a question that is answered on page 41, and assume the tool is broken. It is not. The AI never read page 41. It reads a handful of retrieved chunks, and that one fact changes how you should write every doc you give it.

August 11, 20269 min read

Why Your AI Support Bot Sounds Wrong (And How to Fix It)

Most small teams set up an AI support bot, watch it answer questions correctly, and still get complaints that it sounds off. Too stiff. Too eager. Not like us. This post is the honest guide to what actually goes wrong with AI tone, how to audit it yourself in an afternoon, and why the real fix is simpler than any conversation design framework.

August 4, 20268 min read

How to Keep Your AI Support Bot Knowledge Base Current

If your AI support bot is flagging everything for human review or confidently quoting a policy you changed three months ago, stale docs are the culprit. Here is the scrappy, low-overhead sync routine that founders and small teams actually use to keep their knowledge base current when the product moves fast.

August 1, 202612 min read

Draft-first AI email: why "AI writes, you approve" wins the first 90 days

Every AI support tool wants to auto-send on day one because it demos better. For a real inbox, that is backwards. The first 90 days should be draft-first: the AI writes, you approve, and every approval teaches you where the AI is trustworthy and where it is not. Here is what that arc actually looks like week by week.

July 28, 20269 min read

Open-Source Customer Support: Self-Host vs SaaS Break-Even

Open-source customer support tools like Chatwoot and Papercups are genuinely good software. But free to download is not free to run. Here is the honest break-even math for a 1-10 person team deciding whether to self-host or just pay for a hosted tool.

July 21, 20269 min read

AI Customer Support Setup: Stop Guessing, Flag Unknowns

The fear is real: AI support sounds great until it confidently tells a customer the wrong refund policy. Here is the operational playbook for structuring your knowledge base, automation rules, and review thresholds so your AI drafts on what it knows and flags everything else instead of making things up.

June 30, 20269 min read

WISMO Tickets at Black Friday Volume: A Playbook

WISMO tickets will bury a small team in November. This is the exact setup I would use to handle 300+ order-status emails with one person: how to build the knowledge base, which emails to auto-send vs. draft vs. escalate, and where automation breaks down so you are not caught off guard.

June 16, 20269 min read

SLA Tracking for Small Teams: A No-Fluff Setup Guide

You have outgrown a shared Gmail inbox but a full Zendesk rollout feels like overkill for a team of three. Here is a practical guide to defining SLA tiers, assigning ticket priority without a dedicated ops person, and making sure nothing gets buried when everyone is wearing five hats.

June 2, 20269 min read

WISMO Overload at Peak Season: Handle It Without Hiring

WISMO questions - where is my order - can hit 60-70% of your total support volume during peak season. Staffing for that spike costs $3-4K a month per rep, and per-resolution AI tools get expensive exactly when you need them most. This is the exact routing setup I use to auto-draft order-status replies, keep edge cases in human review, and get through Black Friday without hiring anyone.

May 26, 20269 min read

Draft-and-Approve AI Support: A Category Decision Framework

Full automation is not the goal. The goal is knowing exactly which ticket categories your AI can handle solo, which ones need a human glance before sending, and which ones should never touch the queue. Here is the category-by-category decision framework I built into Trigli's automation rules - and how to apply it to your own support stack today.

May 19, 202610 min read

AI Support Knowledge Base: Build One That Actually Works

The Klarna reversal, the Jeff Geerling post, the Hacker News pile-on - they all trace back to the same root cause: an AI with no grounded knowledge base confidently making things up. Here is how to fix that before it embarrasses you.

May 2, 20269 min read

Klarna AI Support Reversal: What Small Teams Should Learn

Klarna made headlines replacing support agents with AI, then quietly hired humans back. Founders are now asking: does AI customer support actually work, or was it hype? The honest answer is neither. AI handles the right ticket types extremely well and falls apart on the wrong ones. Here is the practical breakdown for small teams deciding how much to trust AI with their inbox right now.

May 2, 20269 min read

Autonomous AI Customer Support: When to Send vs. Approve

Klarna built a fully autonomous AI support system, then quietly started rehiring humans. The lesson is not that AI failed - it is that fully autonomous replies fail at the edges. Here is a practical framework for 1-10 person teams: exactly when to let AI send on its own, and when to make it wait for a human to approve first.

May 2, 20268 min read