Systeric systeric

Take-home

The Systeric take-home

A taste of the actual job. At Systeric we take real problems and validate, solve, and review them together, and this take-home is a small run of exactly that. You get a loose problem and you own it end to end: work out what is worth building, build the smallest thing that does it, and show how you would know it works. Then we dig into it together.

If you light up at a real problem to think through, not a puzzle with a trick answer, you will enjoy this, and that is exactly who we are looking for. The same problem runs for every level, intern to lead; what changes is the bar. Plan for three to four hours, no more. This is a thinking exercise, not a build-a-whole-app test, so keep the code tiny. Use any language and any tools you like, including AI. We care about your judgment, not a polished UI.

The case

Maya runs a small online clothing shop, on her own with one part-time helper. Orders come through the website; questions come through a single chat inbox. It is about 80 customer messages a week now and climbing. She answers them herself, between packing orders, and when a reply takes a few hours the customer has often bought elsewhere by the time she gets to it.

Here is a week of that inbox. Every week looks about like this:

2026-06-22T09:14 jen: hi! what time do you open today?
2026-06-22T09:31 marco: do you ship to Germany?
2026-06-22T09:33 aisha: what's your return policy?
2026-06-22T10:02 tom: where's my order? it's been a week
2026-06-22T10:05 lena: do you have the linen shirt in medium?
2026-06-22T11:20 raj: what are your hours on the weekend?
2026-06-22T13:40 nora: can I return something I bought last month?
2026-06-22T14:12 david: do you deliver to Germany?
2026-06-23T09:05 priya: hi, when do you open?
2026-06-23T09:48 sam: how long does shipping take?
2026-06-23T10:30 ella: is the linen shirt available in M?
2026-06-23T11:00 mike: order 1042 hasn't arrived, where is it?
2026-06-23T15:22 kate: what's the return window?
2026-06-24T09:10 leo: do you ship internationally?
2026-06-24T09:55 mia: what are today's opening hours?
2026-06-24T13:15 omar: still waiting on my order, any update?
2026-06-24T16:40 zoe: can I get that shirt in a medium?
2026-06-25T09:02 ivan: how do returns work?
2026-06-25T09:41 hana: what time do you close today?
2026-06-25T10:18 ben: do you post to Germany?

Most of it is the same handful of questions. Maya's instinct is "I need a chatbot." Maybe. Your job is to look at what is actually happening and decide what is worth building. Read the log closely: which questions repeat, why are they being asked at all, and which ones a faster answer would not really solve. Then build the smallest thing that takes real load off Maya, and be ready to defend why that, and not something bigger. We are deliberately not telling you what to build. Deciding that is the test.

Not sure where to start? A small script that drafts replies to the repeating questions is a perfectly good choice, and you will not be marked down for picking it. The sharpest answers go a step further and ask whether faster replies even dent the biggest pile (notice how often "where's my order?" comes up), but a clean, well reasoned version of the obvious thing is genuinely fine.

What to send back

Reply to our email with all three. The brief is the point; the build proves you can make it real.

  1. A brief, about a page. The real problem and who it hurts. What you decided to build and, just as important, what you decided not to. How you will know it worked: the one number you would watch, and what you would log to see it move. What could go wrong once it is live, and how you would catch it.
  2. A rough prototype of the core, not a product. Pick the single most important piece of your idea and make just that part real, as one small script that runs against the sample log. No UI, no deployment, no polish. We want to see the idea turn into something that actually runs, not a finished app. If your script is more than a page or two, you are building too much.
  3. Proof it works. Tests, a short screen recording, or logged runs that show it does what you claim, including the messy cases: typos, the same question worded three different ways, and a question you chose not to handle.

We care more about your judgment and how you would run this than about a polished build. Ideally get it back within a week. Any questions, just reply.