Bots that never clock out of work.

Pilot S$2,400 for one workflow. Fix Three S$7,000 for three loops. Your fee is design, install, guardrails, a month of watch per loop, training, and handover. Not included: Grok / xAI, Cursor, X, or any other model/API seats and token usage. Those bill to you. After Discovery I will tell you which seats you need. No Ops retainer on the site — Pilot and Fix Three stand alone.

Setup for a loop is 1–2 days. Then I watch it for the rest of that month before handover. One loop ≈ one month. Three loops ≈ three months if we run them one after another. I do not compress three loops into one month for the bundle price.

No — unless you switch that on. Default: the bot drafts, chases, files, and preps. Send, pay, delete, and publish wait for a named human yes. Money, legal, and customer-facing actions stay gated.

No new “automation SaaS.” The bot works on tools you already use (WhatsApp, Gmail/Outlook, sheets, Xero, calendars, vendor sites). You will need your own Grok/Cursor (and similar) seats. That is separate from my fee.

No. Discovery is a plain talk about the dull loop. I install on your stack, write a short runbook, train the person who owns the yes, and hand over. If you can describe the work, we can scope it.

Messy is normal. I watch how the work actually runs — including exceptions — before anything is built. If the loop has no happy path and no owner, I will say so and we stop. The bot fits the work. We don’t invent a clean process you won’t follow.

Hours and cost are baselined before install. After handover you get a Measured Savings snapshot (Pilot) or the same per loop (Fix Three). If the economic case isn’t there in Discovery, I decline the job.

Pilot if you have one painful loop and want proof with guardrails. Fix Three if you already know three loops and want the bundle rate. Pilot fee credits into Fix Three only if you sign within 30 days of signup. After that, Pilot is paid and done.

25% of measured savings for 12 months. The dollar amount is quoted after Discovery, once the baseline is on paper. This path is not “keep the whole admin office and add a bot.” It usually means a staffing change: 1–2 prompt / ops engineers instead of a large admin pile. If you want the office unchanged, use Pilot or Fix Three.

You do. Work runs on your accounts. You keep the credentials, the runbook, and the right to turn a gate on or off. I don’t hold your customer list or your wallet.

Yes — just me. One client at a time. When I’m in your install month, I turn other work away. There is no 24/7 Ops package. After handover, extra loops or a new watch month are a new quote. I am not an on-call helpdesk.

During the watch month I review failures and tighten the happy path. After handover, the runbook says who to ask inside your team. Model/vendor outages are on the vendor. Guardrails are there so a bad draft should not become a live email or a payment unless someone approved it.

That’s the usual stack. No API is not a stop. Browser-driven admin is in scope when we agree the clicks and the gates.

A named owner for the human yes, access to the tools on your seats, and time for one observation of the real loop (not a slide deck). If nobody owns send/pay, I won’t install.

Work stays in your tools and seats. I don’t need custody of funds. I don’t use your customer data to build a product for someone else. Don’t put secrets in chat logs; the runbook will say what the bot is allowed to see.

If the pain is retyping, chasing, filing, or portal-clicking, yes. If the pain is a new core system, a data warehouse, or replacing your accountant, no.

About 20 minutes. You describe the dull loop and what it costs. I say whether it’s a Pilot, Fix Three, Gain-Share, or an honest no. No build starts on that call.

Fees are in SGD. GST only if the invoice says so. Discovery is the place to confirm billing entity and whether remote-only is enough for your stack.