AI automation and the workflows around it
Automation here means two things. The first is plain workflow automation: when a lead form is submitted, send a WhatsApp message, create a CRM contact, append a row to a sheet and alert the sales person. The second is AI automation: the same workflow, but with a model reading the enquiry, classifying it, drafting the reply and deciding who should handle it. I build both, and most engagements end up with a mix. What I do not do is automate a channel nobody has made work manually first; the process has to exist before it can be encoded.
Ownership across the site is simple. Running a channel by hand, such as posting on social or sending an email campaign, sits under digital marketing services. Paid placement sits under Google Ads and Meta Ads. Anything that ranks pages, including visibility in AI answers, sits under SEO. Automating any of them sits here.
Lead, sales and customer workflows
Marketing channel automation
AI services
Operations, data and internal tools
Platform pages
Each platform page explains what I build on that tool and shows a shipped workflow: n8n for self-hosted, code-friendly workflows; Make for visual scenarios with complex branching; Zapier for the widest app coverage with the least setup; HubSpot for CRM-centred marketing and sales automation; GoHighLevel for agencies and local businesses that want CRM, messaging and funnels in one; ManyChat for Instagram, Messenger and WhatsApp flows; Klaviyo for eCommerce email and SMS; and Google Apps Script and Workspace automation for anything that lives in Sheets, Gmail and Drive.
How I scope and build an automation
- Process walk-through. I watch how the task is done today, by whom, with which tools, and count how often. If it happens fewer than a few times a week, I usually recommend not automating it.
- Decide the system of record. Which tool owns the customer record, which owns the money, and what happens when they disagree. Most broken automations trace back to skipping this step.
- Architecture on one page. Trigger, steps, branches, failure path and who gets alerted when it fails. You approve this before I build.
- Build in a test workspace with your real tools and sample data, including duplicates, empty fields and Nepali text, because those are what break in production.
- Go live with monitoring. Error alerts to email or WhatsApp, a run log, and a rollback plan.
- Document and hand over. A plain-language description of every workflow, the credentials it uses and how to pause it. Optional monthly maintenance covers API changes and new steps.
Tools I build on
n8n, Make and Zapier for connecting tools; HubSpot, Zoho and GoHighLevel for CRM; ActiveCampaign, Mailchimp and Klaviyo for email; ManyChat for messaging flows; Google Apps Script for anything in Google Workspace; and custom code where an off-the-shelf connector does not exist. I choose by your stack, your team’s comfort with the tool and the monthly cost, and every account is created in your name.
Pricing pointer
Setup is a fixed-price project quoted on request, depending on how many systems and branches are involved; ongoing maintenance and optimization is also quoted on request; platform subscriptions are extra and usually NPR 5,000 to 25,000 a month depending on the tools. Each service page gives its own range, and the packages page shows how automation combines with marketing retainers.
Proof
I run my own lead intake, reporting and client onboarding on the workflows described here, and I have built internal operations and task-management tools for an agency in Kathmandu and for personal use. In the fertility clinic case study, WhatsApp follow-up connected to Meta lead forms was part of what made 150+ leads manageable for a small front desk.
