A week of work, turned into half an hour. We start from where your time goes, not from a feature list.
Before any talk of AI: every company runs the same set of processes — drafting, collecting feedback, sign-off, order taking, distribution, reporting. The names differ, the six exist everywhere. And the time rarely goes into the doing. It goes into the back-and-forth, the alignment, and the redoing.
| Process | How you do it now | With AI wired in | Where the time goes |
|---|---|---|---|
| Drafting | Open a file, think of hooks, type, fix typos, one at a time | Give one topic, get several versions, edit the one you keep | Half a day becomes minutes |
| Collecting feedback | WhatsApp, email, verbal — versions scattered across three places | Collected in one place, turned into a single change list | Several rounds become one pass |
| Sign-off | Read every line for wording, facts, regulated phrasing | Your house style and banned words flag only the lines to check | Reading it three times becomes checking flagged lines |
| Order taking | Type the quote, issue it, copy it into another sheet | Generated from the conversation, held until you confirm | Fifteen minutes becomes one |
| Distribution | Paste into each platform, reformat each time | Written once, reformatted per platform | Pasting one by one becomes one dispatch |
| Reporting | Catch up on numbers at month end, write it up by hand | Numbers pulled daily, report assembled at month end | Half a day becomes automatic |
So our assessment does not ask “what do you want to use AI for”. It asks “which three things did you do more than three times last week”. Answer that, and there is something to work with.
One meeting. We look at how you work today and narrow it to the two or three processes that eat the most time. No quote, no pitch at this stage.
We demo on one or two of your real tasks, not on a prepared sample. Bring your own files.
One process runs for a set period while we measure two numbers: time to complete, and error rate. Without a baseline you cannot show a difference later.
Only when the pilot shows a difference do we extend to other processes. Accounts, permissions and an export path are handed over at the same time.
If the pilot does not show a difference, we stop at the pilot. That line is in the quote, not only on this page.
The unit we price on is the process, not the seat. The reason is simple: once a process is wired in, running it for three people and for ten costs us almost the same, so charging per head would not hold up.
| Basis | What you pay for | Why it is calculated this way |
|---|---|---|
| Per process | Each process that gets wired in | Adding users after go-live does not add cost |
| Setup and ongoing, separately | Assessment and configuration (one-off): usage and maintenance (ongoing) | You can take the setup only and run it yourself afterwards |
| Three levels of integration | Web entry, Managed Dedicated Host, local install | Deeper integration means more setup and more long-term control |
| No long contract | Per project, or monthly | Stop whenever you want, with nothing left to pay |
| Quoted after assessment | A quote that states scope and fee | We do not publish a number that has nothing to do with your case |
We deliberately do not put a figure on this page. “Handle our quotes for us” can mean one product line or ten, and the scope differs several times over. A published number would only fail to match by the time you are assessed. Levels and billing are set out under three levels of integration and billing.
We do not print “300% uplift” figures here. Every company starts from a different baseline, and the pilot exists to measure yours. Below are three common shapes — the process types, the constraint, and where we would usually start.
| Scenario | Processes that eat time | The constraint | Where we usually start |
|---|---|---|---|
| SME marketing team (two to three people) | Drafting, distribution, reporting | One person runs the whole chain; what is missing is concurrency | Web entry, moving to Managed Dedicated Host once volume is steady |
| Trading and wholesale | Quotes, orders, follow-up | Highly repetitive, and one mistake is expensive | Managed Dedicated Host, with a confirmation step on quotes and orders |
| Professional services (accounting, consulting, insurance) | Document handling, reports, reply templates | Client data is sensitive and cannot leave the building | Local install, or Managed Dedicated Host |
If your case is not on that list it does not mean it does not fit — it means the assessment needs a closer look at the processes.
Ask us these eight too. If any answer is vague, say so. See also best AI agent platform in Hong Kong and frequently asked questions.
Per process, not per seat. You tell us which processes eat the most time; after the assessment we issue a quote stating the scope and the fee. Setup (assessment, configuration, pilot) and ongoing (usage, maintenance) are listed separately, so you can take the setup only. We do not publish a fixed figure, because the same process wired for one product line and for ten differs several times over in scope.
Four steps: assessment (one meeting), demo, pilot, rollout. From assessment to the first process showing a real result is usually measured in weeks, not months. Pilot length depends on the process, and the timetable is written into the quote.
Not for web entry or Managed Dedicated Host. We handle setup, updates and backups. Local install needs a machine you control: we install and configure it, and day-to-day reboots and backups can be yours or ours.
Stated plainly, by level. Web entry: data sits on our shared managed system, controlled by permissions and a full action log. Managed Dedicated Host: you hold the highest privilege on that host, but the host runs on our cloud machine, so technically we can read the disk, and you can export the whole host at any time. Local install: the machine is on your premises and we cannot see it. All three include confirmation prompts and an audit log.
ChatGPT is a single SaaS chat service: one way to use it, data on the vendor cloud. MOKAGI offers three levels of integration including a local install; agents can operate a browser, a desktop and social platforms to finish a task rather than only replying with text; and memory and configuration export as a package you keep.
Yes. The local install is the third level, with the same features as Managed Dedicated Host, and it can run local models. The core code is open source.
No long contract. Per project, or monthly. If the pilot shows no difference we stop at the pilot, with nothing left to pay.
Often most worthwhile there. With few people, one person runs several processes at once, and the time goes into back-and-forth and rework. Start with one or two of the most painful processes rather than everything at once.