Nepal
Most "AI" pitched to Nepali businesses right now is a chatbot demo. Arcetis builds automation that replaces specific manual work — the kind that shows up in a hospital, a CRM, or a booking desk.
For most Nepali businesses, useful AI and automation isn't a generic chatbot — it's removing a specific manual bottleneck: appointment booking that currently runs through phone calls, lead follow-up that currently depends on someone remembering to reply, or reporting that currently means manually compiling spreadsheets. Arcetis has built a hospital management system on React, TanStack, and Convex, along with CRMs and booking platforms, so our automation work comes from shipped systems, not a slide deck of AI buzzwords.
"AI" has become a sales word in Nepal faster than most businesses have had a chance to figure out what it should actually do for them. We'd rather start from the other direction: what manual, repetitive work is currently costing a business time or losing it customers, and can software remove that specific bottleneck. Sometimes that's AI in the strict sense — a model doing something genuinely hard, like summarizing or classifying. Often it's simpler automation — a booking form that writes directly into a calendar and notifies staff, instead of a phone call that might get missed.
Arcetis has built this kind of system before shipping it as a sales pitch: a hospital management system using React, TanStack, and Convex, CRMs for service businesses to track leads and follow-ups, and booking platforms for appointment-based businesses. That background shapes how we scope AI and automation work for new clients — we ask what the system needs to actually do in production, under real usage, not just in a demo.
The clearest wins we see are in lead and inquiry handling (routing WhatsApp, Messenger, and form leads into one tracked place instead of three separate inboxes), appointment and booking systems for clinics, salons, and service businesses, and internal reporting that currently means someone manually pulling numbers together each week.
We're cautious about recommending AI for its own sake. If a rule-based automation — a form, a notification, a scheduled report — solves the problem more reliably and cheaply than a model-based one, we'll recommend that instead, even though it's less exciting to pitch.
There's a second side to AI that matters for Nepali businesses: how they show up inside AI tools like ChatGPT and Google's AI Overviews when a potential customer asks a question. This is AEO/GEO (answer engine / generative engine optimization) — work Arcetis has focused on since the practice began in 2019, positioning us among the first in Nepal to build for it deliberately.
In practice this means structuring a business's content so an AI assistant can quote it accurately and specifically when asked a relevant question, rather than defaulting to a bigger, better-structured competitor or ignoring the business entirely.
Both, depending on what the business actually needs. We've built full custom systems (a hospital management platform, CRMs, booking tools) as well as scoped automations using existing AI services where that's the more sensible, cost-effective fit.
Automation is often more valuable for a small business, since a missed lead or a double-booked appointment costs proportionally more when the team is small. We scope projects from around NPR 20,000, sized to what the business can realistically use.
AEO/GEO is about how a business's content gets quoted or surfaced inside AI answers (ChatGPT, Google AI Overviews) rather than traditional search results. We've worked on this since 2019 and pair it with automation work, since both are about how AI increasingly sits between a business and its customers.
Depends on the job. We'll use an existing platform when it genuinely fits and is more cost-effective; we build custom (as we did with our hospital management system on React, TanStack, and Convex) when the workflow is specific enough that an off-the-shelf tool would be a compromise.