Meet Osewa: agentic customer support for small teams

It’s 9am Monday. You open Gmail and 18 unread support emails are waiting. WhatsApp Business has 14 more. Instagram DMs has 9 nobody got around to checking. A Stripe webhook fired overnight and surfaced 6 refund disputes you didn’t know existed.

Two of your three agents are out today. One customer has emailed three times in two days asking the same question, because the previous reply landed in their spam folder, and you didn’t notice.

Welcome to running support at a small company.

The 50-tab problem

Here’s the workflow we keep seeing at companies under 25 agents:

1. Email comes in to a shared Gmail.
2. Agent opens the email, copies the customer’s address.
3. Pastes it into the CRM in a second tab to find their account.
4. Pastes it into Stripe in a third tab to check the last invoice.
5. Pastes it into the product admin in a fourth tab to read their settings.
6. Pastes it into Notion to find the runbook for this issue.
7. Writes the reply.

That’s 50 tabs across a day, easy. And every time the agent picks up a ticket they’ve never seen before, the customer waits while the agent rebuilds context from scratch.

The big help-desk tools — Zendesk, Intercom, Freshdesk — solve this with deep integrations and dozens of buttons in the agent UI. They’re built for organizations that can dedicate a person to configuring them. For a 4-agent team where the head of support is also the head of operations, that’s overhead you don’t have.

The other end of the market — shared Gmail, free-tier inboxes, “just use Slack” — keeps things simple but doesn’t solve the context problem at all. You answer faster but with less of the picture.

Why we built something new

Most support tooling is built for one of two shapes: the 200-agent enterprise that has a Director of CX and three help-desk admins, or the solo founder who just needs a shared inbox. Teams of 2 to 25 agents — where the founder is still answering tickets at 11pm — sit in the gap. The tools either drown them in configuration or leave them with no real workflow.

We built Osewa for that gap. It’s a customer support platform with one strong opinion: an AI agent should do the boring 60% of every reply, and a human should do the part that needs judgment.

Three design choices we keep coming back to.

1. One queue, every channel

Email, WhatsApp Business, Instagram DMs, web chat, and the customer portal all flow into a single ticket queue. Each ticket is a conversation, not a thread. If a customer messages you on WhatsApp and follows up by email two days later, Osewa pairs them automatically — same customer, same ticket, the agent sees the whole arc.

This matters more than it sounds. The single most common reason support feels broken to a customer isn’t slow replies — it’s repeating themselves to a new agent who can’t see what happened on the other channel.

2. An AI agent that drafts every reply

This is the part we spent the most time on.

On every incoming ticket, an AI agent reads:

• The customer’s full conversation history with you, across every channel.
• Their account in your product (we sync the relevant data per tenant, with row-level isolation so customer A’s data never leaks to tenant B).
• Your knowledge base.
• The runbooks you’ve written for similar past tickets.

…and drafts a reply before the human agent even opens the ticket. The agent can send the draft, edit a sentence, rewrite it from scratch, or ignore it. Osewa never auto-sends without a human’s explicit approval.

We’re stubborn about that. The customer is talking to a human, and the human is in the loop. The AI is a co-pilot, not a substitute.

For a typical SaaS team running on Osewa, the AI draft is usable as-is for about 60% of tickets, partial for another 25%, and unhelpful for the rest. That math will get better. But even at 60%, it cuts a 4-agent team’s first-response time from hours to minutes.

3. Built for the team you actually have

No required onboarding consultant. No 4-week implementation project. You connect a Gmail address and a WhatsApp Business number, point Osewa at your knowledge base, and you’re answering tickets the same afternoon.

The agent UI is deliberately compact — closer to Stripe or Linear than default Material UI — because every pixel of UI chrome is a pixel an agent isn’t reading the customer’s question. We’re working in a job where the user is interrupted every 90 seconds. The interface has to respect that.

What Osewa doesn’t do (yet)

Honest about the shortcuts: there are things Osewa doesn’t do today, and we’d rather tell you upfront than let you discover them mid-implementation.

• Voice (phone) support. It’s on the roadmap, not in the product. If voice is half your volume, we’re not your tool yet.
• Workforce scheduling and deep BI. We have basic SLA tracking. We don’t have shift planning, gamification, or the analytics suite Zendesk Explore gives you.
• No-code workflow builder. If your support process requires “manager approves every refund over $100 in three steps”, we can do it via webhooks today. The visual builder isn’t there yet.

If those are dealbreakers for you, we’d rather you keep using Zendesk and we’ll see you in a year.

Who Osewa is for, in one sentence

If you run a team of 2 to 25 agents, you handle WhatsApp plus email plus at least one social channel, and you’re tired of context-switching between five tabs to answer one ticket — that’s the team Osewa is built for.

How to try it

If that sounds like your team:

1. Email me at h@osewa.io with one line about your inbox shape — channels, agent count, tickets per week.
2. We’ll set up a sandbox tenant and walk you through the AI draft loop on a few of your real tickets. (Anonymized, on your terms.)
3. If it’s a fit, we go live in days, not weeks.

No demo gauntlet. No pricing-page roulette. No 90-day procurement cycle.

I’m Daniel, and I built the first version of Osewa because I got tired of watching small teams either pay enterprise prices or limp along on shared Gmail. There’s a real product in the gap, and we’re going to spend the next few years filling it.

If you’ve got a sharp opinion on what’s missing for your team — write me. I read every email.

— Daniel