It is 4pm on a Saturday. Two of your part-timers have swapped shifts in a WhatsApp group you are not in, one has gone home sick, and the roster on the wall stopped being accurate at lunchtime. You have been told AI can fix this. What nobody explains is that the four tools you are about to shortlist charge in four completely different ways, and at 25 staff the gap between them is over HK$12,000 a year for the same job.
What does AI staff scheduling actually do?
AI staff scheduling builds your roster automatically from three inputs: who is available, what your labour rules allow, and how busy you expect to be. Instead of dragging names into slots, you set the rules once and the software proposes a full week that you edit and publish.
Three capabilities sit behind the marketing word.
--- Auto-scheduling. The software fills open shifts from declared availability, skills and cost limits, so a week takes minutes rather than an evening.
--- Demand forecasting. It reads your past sales or footfall data and predicts how many people Saturday actually needs, which owners get wrong in both directions.
--- Labour cost optimisation. It shows the wage cost of the roster before you publish it, not after payroll has run.
Everything else, clock-in apps, shift swaps, leave requests and messaging, is ordinary workforce software that existed long before AI. That distinction matters commercially, because on most platforms the AI features are not in the cheapest plan.
How much do Deputy, Connecteam, When I Work and 7shifts cost in 2026?
Published list prices in August 2026, converted at HK$7.8 to the US dollar. Every figure below is per month unless stated, and the plan that contains the AI is flagged, because it is rarely the entry plan.
Deputy, charged per user
--- Lite US$5 per user (HK$39): scheduling, timesheets, clock-in, shift swaps, payroll integrations. No auto-scheduling.
--- Core US$6.50 per user (HK$51): adds auto-scheduling, demand forecasting, labour optimisation, wage budgets and biometric clock-in. This is the cheapest plan that contains the AI.
--- Pro US$9 per user (HK$70): adds single sign-on, location hierarchies and 24/7 chat, with the Analytics+ and Messaging+ add-ons bundled in.
--- Add-ons per user per month: HR US$2, Messaging+ US$1.95, Analytics+ US$1.50. Minimum spend US$30 per month per invoice on monthly plans, with roughly 10% off for paying annually.
Connecteam, charged as a flat fee for the first 30 users
--- Free plan for up to 10 users.
--- Basic US$29 (HK$226) for up to 30 users, billed annually.
--- Advanced US$49 (HK$382) for up to 30 users.
--- Expert US$99 (HK$772) for up to 30 users.
--- Users 31 and above are charged per head, reported at about US$0.50, US$1.50 and US$3 per user per month on the three tiers. Monthly billing costs roughly 20% more than annual, and pricing is per hub, so Operations, Communications and HR are separate purchases.
When I Work, charged per user
--- Essentials US$2.50 per user (HK$20) for a single location.
--- Pro US$5 per user (HK$39), adding labour sharing and custom forecasting.
--- Premium US$8 per user (HK$62).
--- Adding time and attendance raises those to US$4, US$7 and US$10 per user respectively.
7shifts, charged per location
--- Free plan for a small single location.
--- Essentials US$39.99 per location (HK$312), reported as covering up to 30 employees.
--- Pro US$79.99 per location (HK$624), reported as covering up to 60 employees.
--- Premium US$134.99 per location (HK$1,053). Payroll adds a US$39.99 base fee per location plus US$6 per paid employee.
Employee caps on per-location plans are reported inconsistently across review sites, with some sources describing unlimited employees per paid location. Ask the vendor to confirm your cap in writing before you commit, because it is the number that decides whether the price holds as you hire.
Why does the billing shape matter more than the sticker price?
Because in this one software category, the word "user" means your entire part-time workforce rather than the people using the software. One manager builds the roster. Everybody on it is billed. That single quirk produces cost differences of five times or more on identical work.
Work out where the shapes cross over and the shortlist narrows itself in about a minute.
--- Per user against a flat fee. Connecteam Basic at US$29 covers up to 30 people, and Deputy Lite at US$5 per head passes US$29 at six staff. From your seventh employee onward the flat-fee tool is cheaper, and it keeps getting cheaper with every part-timer you add.
--- Per user against per location. 7shifts Essentials at US$39.99 per site passes Deputy Lite at eight staff. A restaurant with 30 people pays the same as one with 12.
--- Cheapest plan against cheapest AI plan. On Deputy the jump from Lite to Core is US$1.50 per user, trivial until you multiply it by 25 heads and 12 months: HK$3,510 a year for auto-scheduling and forecasting alone.
Two billing details deserve more attention than they get, and both punish exactly the business type most likely to buy.
Deputy's own pricing FAQ states that a user added five days before month end is charged for the full month, that archiving a user mid-month still incurs the full month, and that unarchiving a previously archived user is billed at the full monthly rate regardless of the day. For a shop that brings the same students back every summer and every Chinese New Year, that is a recurring cost attached to a pattern the business cannot change.
What does this cost a Hong Kong restaurant with 25 part-timers?
Between HK$2,714 and HK$21,060 a year for the same roster. Take a 25-person restaurant in Sham Shui Po, one location, mostly part-time floor and kitchen staff, where the owner spends about four hours a week building the roster and handling swaps.
Annual cost at 25 staff, list prices, one location:
--- Connecteam Basic: HK$2,714 (HK$109 per employee per year)
--- 7shifts Essentials: HK$3,743 (HK$150 per employee)
--- Connecteam Advanced: HK$4,586
--- When I Work Essentials: HK$5,850 (HK$234 per employee)
--- 7shifts Pro: HK$7,487
--- Connecteam Expert: HK$9,266
--- Deputy Lite: HK$11,700, or about HK$10,530 paid annually
--- Deputy Core, the cheapest plan with the AI: HK$15,210, or about HK$13,689 paid annually
--- Deputy Pro: HK$21,060
That is a 5.6 times spread between the cheapest option and the cheapest option that actually includes AI scheduling, on identical headcount.
Now put it next to the two numbers that decide whether any of it is worth buying.
The owner's own time. Four hours a week is 208 hours a year. Valued at HK$350 an hour, that is about HK$72,800 of the most expensive attention in the business, spent on a task with no revenue attached. The entire subscription gap between cheapest and dearest, HK$18,346, is a quarter of that.
The wage bill the roster controls. Suppose the restaurant schedules 1,400 part-time hours a month at an average HK$60 an hour, against a statutory minimum of HK$43.1 an hour from 1 May 2026. That is HK$1,008,000 a year being allocated by hand on a Saturday afternoon. Trimming 3% of scheduled hours by matching staff to demand is HK$30,240 a year, more than twice the most expensive subscription on the list.
The conclusion is not "buy the expensive one". It is that the prize sits in the wage bill and the owner's calendar, so refusing to pay per head for software one person operates is the cheapest way to reach it.
Where does each tool actually win?
Each of these products beats the others somewhere, and one of them beats all of them for the typical Hong Kong small business. A comparison in which one vendor wins every row is an advertisement, so here is the honest split.
Connecteam wins on price for 10 to 30 part-timers. A flat US$29 for up to 30 people, plus a usable free tier up to 10 users, makes it the lowest-risk purchase in the category for a single-site shop or restaurant. Its automation is shallower than Deputy's, and for most owners that will not matter.
Deputy wins on depth and on the clock. Its auto-scheduling, forecasting and wage budgeting are the most developed here, and biometric clock-in addresses buddy punching, a real and rarely discussed cost in Hong Kong food and beverage. With three sites and a payroll integration to protect, the per-head price starts to be defensible.
7shifts wins outright for a restaurant with 30 or more staff per site. Per-location billing means your headcount stops driving your software bill, and the product is built around food and beverage patterns rather than generic shift work.
When I Work wins on the lowest entry price. At US$2.50 per user for a single location, a team of eight that only needs a published roster and swap requests will not find a cheaper paid option that works properly.
Local vendors win on the thing global vendors ignore. Hong Kong and regional platforms, including Optix Solutions and Roubler, sell AI rostering with local labour-law compliance as the headline. Their pricing is quote-based, so they cannot be compared here on cost, but they belong on your shortlist for the reason in the next section.
None of them wins on Hong Kong statutory rules, Chinese-language rosters, or the WhatsApp problem. All four global tools are built for English-first workplaces in the United States, United Kingdom and Australia. Their compliance engines model those jurisdictions, their AI reads English shift notes, and none of them knows about a swap two of your staff agreed in a group chat.
What can AI rostering not do in Hong Kong?
Five limits, and the first one carries real money. Nothing here is a reason to avoid the category. All of it is a reason to ask harder questions before you pay.
1. It does not know the 468 rule. Since 18 January 2026, Hong Kong's continuous contract threshold has changed. After four weeks with the same employer, an employee is under a continuous contract if they worked at least 17 hours in a week, or, in any week under 17 hours, if they worked 68 hours or more across that week and the three preceding weeks. That status brings statutory entitlements including paid annual leave, sickness allowance and severance eligibility. An AI told to minimise labour cost will happily place a part-timer on 17 or 18 hours a week, and nothing in a global product will flag the moment a rolling four-week total crosses 68 hours. Ask any vendor in writing whether their compliance engine models the Hong Kong Employment Ordinance, and assume the answer is no until you see it. This is general information, not legal advice.
2. It cannot fix an understaffed roster. With five people available for six shifts, AI shows you the hole faster and more precisely. It does not fill it. Owners buy scheduling software hoping to solve a hiring problem roughly as often as a scheduling problem.
3. Demand forecasting inherits your sales data. If a meaningful share of your business is walk-in trade on a local point-of-sale system the tool cannot connect to, the forecast covers part of your day and guesses at the rest, confidently.
4. Per-user billing charges for people who work twice a month. Count the names that appear on the roster in a month, not the number you think of as staff. On per-head pricing, the occasional weekend helper costs the same as your full-time supervisor.
5. Shift swaps happen where the software cannot see. A large share of Hong Kong shift coordination lives in WhatsApp groups. A swap agreed there and never entered leaves your labour cost report, your attendance record and your four-week hours total all wrong at once. That last one matters more from 1 May 2026, when the monthly wage cap triggering the hours-recording requirement rose to HK$17,600, which covers most part-time staff.
Which one should you choose?
Answer one question first and most of the shortlist disappears: how many names appear on your roster in a typical month, and do they all work at one address?
--- Under 10 staff, one site. Connecteam's free plan, or When I Work at US$2.50 per head. Do not pay for AI scheduling to arrange nine people.
--- 10 to 30 part-timers, one or two sites, price is the constraint. Connecteam Basic or Advanced, at HK$2,714 to HK$4,586 a year.
--- A restaurant with 30 or more staff at one address. 7shifts Pro. Per-location billing is the right shape and the product speaks your industry.
--- Two or more retail sites, payroll integration, and a clock-in problem. Deputy Core, provided you genuinely need both the depth and the biometric attendance.
--- Rosters written in Chinese and swapped in WhatsApp. Buy nothing yet. Run one real month through two free trials and read the results yourself. This test disqualifies more shortlists than any price comparison.
--- Fewer than about 10 roster changes a month, everyone on fixed shifts. Buy nothing. A shared spreadsheet and a written rule about who approves a swap will outperform a subscription, because your problem is a habit rather than a tool.
When UD is the wrong thing to buy
UD's AI Staff Solution is not a rostering product and is not sold per seat, so if your question is who works Saturday, buy one of the four tools above and stop reading. It will cost between HK$226 and HK$1,268 a month and it will solve the problem properly.
UD becomes relevant only when the repetitive admin surrounding the roster runs to several hundred items a month: re-keying attendance into a payroll spreadsheet, chasing staff for documents across three channels, producing the same bilingual notices weekly, or reconciling hours between a clock-in app and an accounting system by hand. The honest first step is to count those tasks for two weeks, not to book a demo.
Frequently asked questions
Is there a genuinely free AI rostering tool?
There are free rostering tools, not free AI rostering. Connecteam is free up to 10 users and 7shifts has a free single-location tier, but auto-scheduling and demand forecasting sit on paid plans across the category.
Do these tools handle Hong Kong payroll?
Not natively. They export approved hours, and integrations are built around United States, United Kingdom and Australian payroll providers. Assume you or your accountant will bridge the last step, and ask about export formats before you buy.
Will my part-timers actually use the app?
This is the real adoption risk and it is worth testing before payment. Give the app to your three least technical staff during the trial. A cheaper tool they will use is worth more than a sophisticated one they ignore.
Can it help with the 468 rule if it does not understand it?
Indirectly, yes, and this is the honest case for buying one. A tool that records actual hours gives you the data to run your own four-week check, which is far better than reconstructing it from memory. The obligation to interpret the rule stays with you.
The verdict
In every other software category you pay for the people who use the product. In rostering, you pay for the people who appear in it, which is why the sticker price tells you so little and the billing shape tells you almost everything. Count the names on your roster, count your addresses, then pick the shape that does not charge you for growing.
Then spend the difference on the thing the software cannot do: knowing that a part-timer on 17 hours a week for four weeks has just changed status under Hong Kong law while your dashboard shows a perfectly optimised roster. UD stands with you, making AI human. Twenty-eight years here has taught us the right answer is often a HK$226 subscription and a written rule, and we will tell you when that is the case.
Related reading: what voice cloning means for your payment approvals, what Asana, ClickUp and monday really cost, and how AI credits are billed.
Reviewed by the UD AI team, Hong Kong, August 2026.
Counted your tasks and the number is high?
If the repetitive admin around your roster runs to several hundred items a month, that is a different purchase from a scheduling app, and we will walk you through it step by step, from counting the tasks honestly to matching the ones worth handing over. If the answer is that you do not need us yet, we will say so plainly.
