Labour-Saving Automation in ASEAN Factories: Cost, ROI and Sequence

We installed one robot, and the headcount on the line did not change at all. In Japanese-owned and foreign-owned plants across Thailand and Vietnam, this comment comes up surprisingly often. The machine runs, cycle time is shorter, and yet the manning table looks exactly the same as last year. The cause is almost never the equipment itself. It is the order in which the work was tackled. When you look at labour-saving automation plant by plant, the factories that succeed have one thing in common: before buying anything, they put a number on how many people, how many hours and how much money sit on each operation. This article walks through that stocktake, then through the cost and payback period of each stage, in a form where you can drop in your own figures and get your own answer.

What labour-saving automation is, and how it differs from effort reduction and full automation

If the vocabulary is loose, the investment decision will drift. Start by separating three ideas.

The line most often blurred is the one between effort reduction and labour saving. If the operator’s job becomes easier but that operator still stands on the line, labour cost has not fallen by a single baht. Effort reduction is a necessary condition for labour saving, not a sufficient one. Conversely, labour saving does not always require automation. If an inspection done by three people can be done by two, purely by automating the record keeping and clarifying the pass/fail criteria, that is genuine labour saving.

This is also where “we automated but nobody left the line” comes from. The machine automates the main operation, but if you then need someone to feed work into it, someone to take parts out, someone to run over when it stops, and someone to copy its output onto a paper sheet, the headcount is unchanged. Labour-saving automation is best understood as removing the man-hours stuck around the main operation, not the main operation itself. Teams that frame it that way pick better investments.

What goes wrong when the three terms are mixed together

The classic failure is a capital request that says only “productivity will improve by 30% through automation”, with no basis for a headcount change. Productivity improvement is good, but unless output rises, the gain is absorbed on the shop floor as spare time and nothing improves in cost terms. A labour-saving proposal has to state four things together: which process, how many man-hours, from when, and how much money. If you cannot write those four, the investment is not yet ready for a decision.

Why labour-saving automation matters now: wage and workforce numbers in Thailand and Vietnam

“There is a labour shortage, so let us automate” is true but far too coarse to support an investment decision. Here are the actual numbers.

Wages are rising in Thailand while labour supply is thinning

In Bangkok, the minimum wage became a flat THB 400 per day for all industries under the Royal Gazette notification effective 1 July 2025 (it was THB 372 before the increase). In the manufacturing heartland, including Chonburi and Rayong, four provinces and one district had already moved to THB 400 from January 2025. Roughly 700,000 workers are covered (JETRO).

The step from THB 372 to THB 400 is THB 28, about 7.5%. On 300 operating days a year, that is THB 28 × 300 days = THB 8,400 per operator per year. For a plant with 100 operators, THB 8,400 × 100 = THB 840,000 per year of additional fixed cost. Throughout this article, where a US dollar figure is shown, it is converted at USD 1 = THB 32 and rounded; on that basis THB 840,000 is about USD 26,250 per year.

The supply side is tighter still. On 14 July 2026 the Thai cabinet approved, and on 19 July it gazetted, an extension of work permits for approximately 770,000 registered migrant workers from Myanmar, Laos and Vietnam. At the same time, Thai business associations are warning of a shortfall of roughly 500,000 workers across manufacturing, agriculture and construction (Mizzima / Bangkok Post). Registered Myanmar nationals number more than 2.3 million, accounting for 75% of CLMV workers. Manufacturing in Thailand therefore runs on a migrant workforce of that scale, and its supply moves every time policy changes. That is the structural position today.

Over a longer horizon, Thailand’s working-age population (15 to 65) stood at 46.85 million in 2022 but is expected to enter decline around 2026, while the share of people aged 65 and over is projected to rise from 13% in 2022 to 20% by 2029 (IDE-JETRO / Bangkok Shuho). Wages up, the base population down, and migrant labour carrying policy risk — all three are arriving at the same time in 2026.

Vietnam is moving in the same direction

Vietnam’s regional minimum wages rose by an average of 7.2% from 1 January 2026 under Decree 293/2025/ND-CP, promulgated on 10 November 2025. Monthly rates are VND 5,310,000 for Region I, VND 4,730,000 for Region II and VND 4,140,000 for Region III. On 16 July 2026 the National Wage Council agreed on a further average increase of 7.8% from 1 January 2027 and submitted it to the government (Vietnam Briefing / VietnamPlus).

If both steps are implemented as proposed, the compound effect against the 2025 level is 1.072 × 1.078 = 1.1556, that is about 15.6% over two years. Applied to the Region I monthly rate, 5,310,000 × 1.078 = VND 5,724,180 (based on the 2027 proposal, which is not yet final). The assumption that moving production from Thailand to Vietnam keeps labour cost structurally low is already eroding.

The equipment is idle. What is missing is people

This is the point worth emphasising most. Thailand’s Manufacturing Production Index (MPI) for June 2026 fell 3.1% year on year, the second quarter of 2026 was -1.79%, and average capacity utilisation was 57.47%. June’s drop was the steepest since November of the previous year, driven mainly by cuts in automotive and petroleum-related output (Xinhua).

Utilisation of 57.47% means 42.53% of installed capacity is sitting idle. In that situation, a capital request for “automation to increase production capacity” will not get past management, and rightly so. The capacity is already there.

What is justified right now is not automation that adds capacity, but labour-saving automation that produces the same volume with fewer man-hours. Get this distinction wrong and the request comes back marked “not urgent”. Write it instead as “this many man-hours removed, this much annual labour cost taken out of fixed cost”, and it becomes exactly the kind of investment that gets approved in a low-utilisation period.

Globally, robot adoption continues to accelerate. According to the International Federation of Robotics (IFR), new industrial robot installations worldwide reached 542,000 units in 2024, more than double the level of ten years earlier, with Asia accounting for 74% (about 401,000 units), Europe 16% and the Americas 9%. The operational stock has passed 4 million units, 2025 is expected to be up 6% at 575,000 units, and 2028 should exceed 700,000. This is not an argument to go and buy a robot. The right reading is that because adoption in Asia has already reached three quarters of the world total, there is now a large body of both successes and failures to learn from, and later movers gain the most from getting the sequence right.

For the overall sequence of factory automation and the wider picture of BOI privileges, see our separate guide, Factory Automation in Thailand: Implementation Sequence, Costs and BOI Incentives.

The most important step: taking stock of man-hours before you buy anything

Eighty per cent of the outcome is decided here, before any equipment is selected. The task is simple: take one shift of man-hours on one line, split it along four axes, and convert it into money.

First, fix a single hourly rate

To turn man-hours into money you need an hourly rate. As a rough guide from the projects we take on in Thailand, the annual cost of one direct operator typically comes to 1.5 to 1.8 times the minimum-wage base once overtime, social security, welfare, dormitory and transport are included.

• Wage base: THB 400/day × 300 days = THB 120,000/year
• Including on-costs: THB 120,000 × 1.5 to 1.8 = THB 180,000 to 216,000/year

For the calculations in this article we use a working figure of THB 200,000 per direct operator per year (about USD 6,250). With annual working hours of 300 days × 8 hours = 2,400 hours, the hourly rate is 200,000 ÷ 2,400 = THB 83.3/hour. From here on we use THB 83/hour (about USD 2.6/hour) consistently. When you run this for your own plant, replace those two numbers — annual cost per operator and hourly rate — with your actual figures. That alone converts every table below into your own version.

The four axes

Assume ten people on one shift (8 hours) on the target line, so the input per shift is 10 people × 8 hours = 80 man-hours. Split it into the following four categories, plus waiting.

From here on, man-hours and amounts are all per shift; annual figures multiply by the number of shifts and operating days.

Now extend it to a year. At 300 operating days on two shifts:

THB 6,640/shift × 300 days × 2 shifts = THB 3,984,000 per year (about USD 124,500)

That is the total labour cost carried by one line. And the target for labour-saving automation is the portion outside the 44 direct man-hours, that is 36 man-hours/shift, or 45% of the total:

(80 − 44) × THB 83 × 300 days × 2 shifts = THB 1,792,800 per year (about USD 56,025)

Here is the key point. Most factories have never put a price tag on this block of THB 1.79 million a year. Transport, recording and rework are all buried inside “the job” of the direct operator, so they are never recognised as reducible. The moment you break that THB 1.79 million into four parts and show it, the reaction from management changes. What a capital request needs is not “automation will improve efficiency”, but “of the THB 1.79 million a year of non-value-adding man-hours, this investment removes this many baht”.

How to run the stocktake (three days is enough)

You do not need a full industrial engineering study. This is the simplified procedure we actually use on site.

1. Narrow it to one line. Trying to do the whole plant at once always stalls. Pick one line with high manning, high defects and high overtime.
2. Run a time study on one shift for just two days. For each operator, record which of the five categories above they are in, in 15-minute blocks. No stopwatch is needed; 15-minute blocks are accurate enough to produce the share breakdown.
3. On day three, convert to money on a single sheet. Use the same format as the table above. At this point you have 90% of the material needed for the investment decision.

Having a system that captures output and downtime automatically improves both the accuracy and the speed of this exercise. The approach of starting from data collection is covered in Factory IoT Implementation Guide 2026.

Ways of doing the stocktake that do not work

• Talking only in averages. “Average utilisation 85%” tells you nothing. Until you break it down to who is doing what and when, the removable man-hours stay invisible.
• Measuring only on peak days. People move differently on peak days. Look at both normal and peak days, and base the investment decision on normal days.
• Asking operators to self-report. Without any bad intent, the share of direct work is reported too high. The rule is third-party observation, or capture from machine signals.

The five stages of labour-saving automation and what each stage costs

Once the stocktake is done, you choose the measures. Jumping straight to stage 4 or stage 5 is the single biggest cause of failed labour-saving automation. Every step up multiplies the investment by roughly an order of magnitude, lengthens the payback period, and tightens the preconditions required (standardisation, part accuracy, maintenance skills).

The five stages at a glance

These figures are indicative of the projects we take on in Thailand; they are not market statistics. They vary widely with specification, quantity, safety requirements and site conditions.

A representative case and payback for each stage, calculated in full

Ranges alone do not support a decision, so here is one mid-range case per stage, calculated end to end. The rate is THB 83/hour from the previous section, one direct operator is THB 200,000/year, and operation is 300 days × 2 shifts.

Stage 1: jigs (investment THB 80,000)

Locating jigs and one-touch clamps remove 1.5 man-hours per shift of changeover and positioning work.

• Man-hours removed: 1.5 man-hours × 2 shifts × 300 days = 900 man-hours/year
• Value: 900 man-hours × THB 83 = THB 74,700/year
• Payback: 80,000 ÷ 74,700 = 1.07 years (about 13 months)

Stage 2: poka-yoke device (investment THB 150,000)

An assembly-sequence interlock and a part-count sensor cut sorting and re-inspection work as well as escaped defects.

• Labour saving: 1.0 man-hour × 2 shifts × 300 days = 600 man-hours × THB 83 = THB 49,800/year
• Reduction in escaped-defect handling (sorting cost, post-shipment response): THB 100,000/year (mid-point of the THB 50,000 to 300,000 range typical of our projects)
• Total effect: 49,800 + 100,000 = THB 149,800/year
• Payback: 150,000 ÷ 149,800 = 1.00 years (about 12 months)

Stage 3: semi-automatic machine (investment THB 1,200,000)

A parts feeder plus automatic feeding removes one dedicated feeding operator per shift.

• Labour saving: 1 person × 2 shifts = 2 people → 2 × 200,000 = THB 400,000/year
• Quality gain (less sorting and fewer claims caused by feeding errors): THB 120,000/year
• Added cost (spare parts, consumables): −THB 40,000/year
• Net effect: 400,000 + 120,000 − 40,000 = THB 480,000/year
• Payback: 1,200,000 ÷ 480,000 = 2.50 years (30 months)

Stage 4: standalone automation with 100% inspection (investment THB 3,800,000)

Machine body THB 3,000,000 + vision inspection THB 500,000 + safety fencing and transfer modification THB 300,000 = THB 3,800,000. It removes the night shift and reduces inspectors.

• Labour saving (gross): 3 night-shift direct operators + 2 inspectors = 5 people × 200,000 = THB 1,000,000/year
• Added cost: one maintenance and changeover technician (200,000) + power, consumables and spares (120,000) = THB 320,000/year
• Net effect on a labour-only basis: 1,000,000 − 320,000 = THB 680,000/year → payback 3,800,000 ÷ 680,000 = 5.59 years
• Quality cost reduction (sorting, claims and re-inspection of escaped defects): +THB 300,000/year
• Overall net effect: 680,000 + 300,000 = THB 980,000/year → payback 3,800,000 ÷ 980,000 = 3.88 years (about 47 months)

The lesson from stage 4 is that on labour cost alone it takes 5.6 years, but with quality cost included it becomes 3.9 years. Whether 100% inspection automation can be justified is decided by the estimate of quality cost, not by labour cost. Submit it on labour cost alone and it will not be approved.

Stage 5: line linking (investment THB 12,000,000)

A project covering automated inter-process transport, automatic data collection and host system integration.

• Labour saving: equivalent of 8 people × 200,000 = THB 1,600,000/year
• Quality and yield improvement: THB 400,000/year
• Reduction in transport and work-in-progress inventory: THB 300,000/year
• Added cost: two maintenance staff (400,000) + power and similar (100,000) = −THB 500,000/year
• Net effect: 1,600,000 + 400,000 + 300,000 − 500,000 = THB 1,800,000/year
• Payback: 12,000,000 ÷ 1,800,000 = 6.67 years

Stage 5 is genuinely hard to justify on its own ROI. A figure of 6.7 years exceeds the 3 to 5 year hurdle most Japanese-owned manufacturers apply. Stage 5 makes sense only when it is done together with a capacity increase or a new model introduction, or when incentive schemes can be combined with it. Rather than aiming at stage 5 from the start, the realistic path is to build a track record through stages 1 to 4 and then plan stage 5 around a production increase.

Why you should not skip stages

There are three reasons.

1. Stages 1 and 2 fix the standard. In the process of installing jigs and poka-yoke devices, “what correct work looks like” gets defined physically. Without that definition, a stage 4 machine will assume a part orientation, dimension and loading sequence that does not match reality on the floor, and people end up standing next to it anyway.
2. Stage 3 reveals the split between machine and human. Running a semi-automatic machine shows you, with real data, where the boundary lies between what a machine can take over and what a person must judge. Skip it and you will over-extend the scope of automation, which lengthens payback.
3. Stages 1 and 2 pay back in about a year and fund what comes next. As calculated above, stages 1 and 2 pay back in roughly twelve months. Building on those results before proposing a large investment makes internal agreement dramatically faster.

A small start is not a timid choice. It is an investment in raising the success rate of stage 4 and beyond.

Labour saving case study patterns: five common factory processes and which stage works

From here we look at process types and which stage delivers. No company names are used; these are described as the process types we actually work on in Thailand and Vietnam.

The indicative annual savings are calculated on the same basis as before: one direct operator = THB 200,000/year, hourly rate THB 83, 300 days × 2 shifts.

Feeding and loading

This is where labour saving pays off most readily. It is extremely common to find a machine running automatically while a person does nothing but load it, and that person is tied to the machine cycle and cannot do anything else. One person per shift is THB 200,000/year; two shifts is THB 400,000/year. Combined with the investment in a parts feeder or magazine feeding, the low end gives 400,000 ÷ 200,000 = 2.0 years and the high end 1,500,000 ÷ 400,000 = 3.75 years. In other words, this is a process where two-shift operation is what justifies the investment. Put a THB 1,500,000 feeding system on a single-shift line and it takes 1,500,000 ÷ 200,000 = 7.5 years, so you should stop at stage 1 magazine feeding instead.

The decisive question is whether part orientation is stable. Parts with burrs, gate remnants or deformation will jam a feeder. If that is a concern, use stage 1 first (alignment jigs, magazines) to reduce manning while stabilising part orientation, and only then move to stage 3.

100% visual inspection

In plants serving quality-demanding customers it is not unusual to find two people per shift, four across two shifts, tied to inspection. Four people is 4 × 200,000 = THB 800,000/year. The value of vision inspection here is large, but automating visual inspection is also the hardest area of all.

Three conditions must hold.

• The appearance of a defect can be defined. Inspection that rejects parts because “something looks off” cannot be automated. You first have to write down limit samples and judgement criteria.
• Lighting can be fixed. Where daylight through factory windows or seasonal changes in illumination affect the judgement, a light-shielded enclosure is mandatory.
• Do not demand zero escapes from the machine. In practice, a semi-automatic configuration in which the machine rejects everything suspicious and a person makes the final call will take inspectors from four down to one. That saving is 3 × 200,000 = THB 600,000/year.

Inter-process transport

Manual trolley transport is the classic case of man-hours that nobody recognises. The person moving material is usually not a designated handler but a direct operator carrying things when free, so it never appears on the manning table. That is precisely why the stocktake is needed. 0.5 person on one shift is THB 100,000/year; one person across two shifts is THB 400,000/year.

The caution here is that an AGV is not the first option. If the layout is straight and there is no change in height, gravity chutes or roller conveyors are far cheaper and do not break down. AGVs earn their keep when routes are long, destinations are numerous and the layout changes often. The decision criteria are set out in detail in AGV and AMR Implementation in Thailand: Approach and Costs.

Recording and reporting

The amounts look small, but this is the process with the best ratio of return to investment. If the equivalent of 0.5 person per shift (4 man-hours/shift) goes into paper records and transcription, two shifts make that the equivalent of 0.5 × 2 = 1 person = THB 200,000/year. Replacing it with automatic IoT collection costing THB 150,000 to 800,000 gives a payback range from 150,000 ÷ 200,000 = 0.75 years to 800,000 ÷ 200,000 = 4.0 years. Where a few sensors and simple collection software are enough, under one year is not unusual.

There is also a second-order benefit: once handwritten records disappear, the man-hour stocktake from the previous section can run continuously. Data for the next investment decision accumulates automatically, which makes automating records an unusually good entry point into labour-saving automation.

Changeover

Changeover time consumes whole man-hours without counting towards output — pure non-value-adding work. Here is the range, calculated properly.

• Low case: 0.5 hours saved per changeover × 2 people × 2 times/day × 300 days = 600 man-hours/year → 600 × 83 = THB 49,800/year
• High case: 0.75 hours saved per changeover × 3 people × 4 times/day × 300 days = 2,700 man-hours/year → 2,700 × 83 = THB 224,100/year

The investment is usually only THB 50,000 to 400,000 of jig and clamp modification, which makes this the area a high-mix low-volume plant should attack before any robot. High-mix low-volume production means frequent changeovers, so the result lands close to the high case.

The ROI formula and payback simulation for automation

Now let us put all of this into a form you can use in a capital request.

The formula (drop in your own numbers and it produces an answer)

“`

Annual saving =

(man-hours removed/shift x operating days x shifts x hourly rate)

+ (annual reduction in defect and rework cost)

+ (annual saving from shorter changeover)

• (additional annual labour: maintenance, changeover, monitoring)
• (increase in annual running cost: power, consumables, spares, software maintenance)

Simple payback period = initial investment / annual saving

“`

Most capital requests fail because the last two lines, the negative terms, are left out. Automation always increases maintenance man-hours and consumable cost. Allow for them from the beginning and nobody says “this is not what we were told” after start-up. Equally, leaving out the second and third lines (defect reduction, changeover reduction) understates the investment and gets it rejected. A correct request contains all five items.

Three-case simulation

Here are the representative cases for stages 1+2, stage 3 and stage 4 from the previous section, lined up as three investment cases. All use the same assumptions: one direct operator = THB 200,000/year, hourly rate THB 83, 300 days × 2 shifts, USD 1 = THB 32.

Checking the total row: investment is 230,000 + 1,200,000 + 3,800,000 = THB 5,230,000. The annual saving is 224,500 + 480,000 + 980,000 = THB 1,684,500. The combined payback is 5,230,000 ÷ 1,684,500 = 3.10 years.

Checking the THB 124,500 labour saving in Case A: THB 74,700 from stage 1 plus THB 49,800 of labour saving from stage 2 = THB 124,500. Adding the THB 100,000 quality saving gives THB 224,500/year.

How to read this table

Comparing the three cases produces a clear pattern.

• The larger the investment, the longer the payback. Investment rises from THB 230,000 to 1,200,000 to 3,800,000 (about 5.2x then about 3.2x), while payback moves from 1.02 to 2.50 to 3.88 years (about 2.5x then about 1.6x). Payback does not lengthen as fast as investment grows, but it does lengthen.
• Annual return per baht invested is highest for the smallest investment. Case A is 224,500 ÷ 230,000 = 98% a year, Case B is 480,000 ÷ 1,200,000 = 40% a year, and Case C is 980,000 ÷ 3,800,000 = 26% a year.
• Even so, the absolute saving is largest in Case C. Repeating Case A alone only accumulates THB 224,500 a year at a time.

The practical conclusion is therefore to generate quick cash with Case A and use it to fund a staged move into Cases B and C. Executing all three leaves a permanent fixed-cost reduction of THB 1,684,500 (about USD 52,640) a year from year five onwards.

Here is how that figure relates to the THB 1,792,800 a year of non-value-adding man-hours identified in the stocktake earlier. THB 1,684,500 includes the quality and defect saving (THB 520,000) and the additional cost (−THB 360,000), so it cannot simply be divided against the non-value-adding man-hours figure. Taking out the labour saving alone gives 124,500 + 400,000 + 1,000,000 = THB 1,524,500 a year (about USD 47,640). Of that, Case C’s THB 1,000,000 includes THB 600,000 for three night-shift direct operators, and that reduction sits outside the non-value-adding man-hours block.

So the portion that directly removes the THB 1,792,800 block identified in the stocktake is 124,500 + 400,000 + 400,000 = THB 924,500 a year (about USD 28,890), a ratio of 924,500 ÷ 1,792,800 = about 52%. The accurate picture across the three cases combined is therefore that over five years you remove roughly half of the non-value-adding man-hours, and on top of that you gain the quality-cost saving and the night-shift direct-labour reduction as separate items. Bear this in mind before you write the capital request: you cannot expect to remove the entire block identified in the stocktake.

What payback period is reasonable

As an internal hurdle among Japanese-owned manufacturers, the ranges we most often encounter on projects are as follows. These are not regulatory standards, only the tendency we see in the work that comes to us.

Five patterns of failed labour-saving automation

When clients in Thailand and Vietnam ask us to rescue a system that did not work out, the cause is almost always one of these five.

Pattern 1: high-mix low-volume production, and the robot cannot keep up

This is the most common failure. Robots are strong at repeating the same motion and weak at product changeover. If there are 20 variants and teaching plus jig change takes 30 minutes per variant, three changeovers a day create 90 minutes of setup, which cancels out the labour saved.

How to avoid it: run a Pareto analysis of volume by variant and automate only the top two or three variants, which typically account for 70 to 80% of the quantity. People make the rest. The moment you aim for “all variants on one machine”, the equipment gets complex, the investment balloons and changeover slows down. Not trying to do everything on one machine is the single biggest trick in high-mix low-volume automation.

Collaborative robots have an advantage in changeover flexibility, but they are not universal either. The conditions under which they apply are set out in Collaborative Robot Implementation: Costs and Approach.

Pattern 2: the machine cannot absorb part-to-part variation

A human adjusts by feel and assembles the part even when there are burrs or gate remnants. A machine cannot. Install an automatic machine where dimensional tolerance, orientation or surface condition varies, and minor stoppages become frequent, so somebody ends up permanently attending to them — more man-hours, not fewer.

How to avoid it: before you consider equipment, measure the accuracy of parts from the previous process. This costs nothing. Measure 100 pieces and see how many fall inside tolerance. If the spread is wide, fix the previous process first, or add a stage 1 alignment jig that absorbs the variation. Do not reverse the order.

Pattern 3: changeover time is not in the calculation

Capital requests very often omit changeover time from the cycle time calculation. The plan says “cycle time 30 seconds × 1,000 pieces = 500 minutes”, but in reality 90 minutes a day goes into changeover and the planned capacity is never reached. People are then kept on to make up the output, and the labour saving disappears.

How to avoid it: always write the cycle time in a capital request as an effective cycle time that includes changeover. As calculated earlier, shortening changeover is itself worth THB 50,000 to 220,000 a year, so the correct approach is to plan stage 1 jig modification and the automation investment together.

Pattern 4: automating a process that has no work standard

If a process exists only as “the way Mr A, our veteran, does it”, the requirement definition never finishes. The equipment maker needs to know in what sequence, at what position and with what force, and nobody has written it down. As a result the specification never settles, design proceeds anyway, and modification work piles up after start-up.

How to avoid it: doing stages 1 and 2 first fixes the standard physically. A jig creates “it can only be placed here” and a poka-yoke device creates “it can only be done in this order”. That is a stronger form of standardisation than documentation. It is not that you cannot automate because there is no standard; it is that there is no standard because stages 1 and 2 were skipped.

Pattern 5: no maintenance capability in house

The machine is installed, but when it breaks there is nobody inside the company who can fix it. Calling the vendor takes three days, and during those days the line reverts to manual labour. If that happens a few times a year, the labour saving evaporates.

How to avoid it: at the planning stage, deduct the cost of maintenance staff from the annual saving. That is exactly why Case C above includes “one maintenance and changeover technician = THB 200,000/year”. In addition, keep the machine control readable in house as far as possible. A black-boxed program creates a permanent external maintenance cost. This point is also discussed in Outsourcing PLC Program Development.

Conditions for unmanned night shifts and 100% inspection automation

These two carry the largest labour-saving impact. The conditions are demanding, so here they are as checklists.

Four conditions for an unmanned night shift

Taking out a whole night shift removes 3 direct operators × 200,000 = THB 600,000/year in one move. As Case C showed, that is the main driver justifying a THB 3,800,000 investment. But all four of the following must hold.

The most commonly overlooked item is the discharge side. Everyone thinks about material supply on the input side, but because nobody designs where eight hours of finished parts will go, someone ends up collecting them every two hours. That is not unattended operation.

Also, because night work carries a premium, removing one night-shift person is normally worth more than removing one day-shift person. This article conservatively uses the same figure (THB 200,000/year) for both; reflecting your own night-shift premium will shorten the payback further.

Three conditions for 100% inspection automation

Restating the points touched on earlier.

1. There is a catalogue of defects. Defective parts from the past year have been classified, with frequency and appearance recorded by type. Without this, a vision inspection supplier cannot design the judgement algorithm. With it, quotation accuracy improves dramatically.
2. Judgement criteria can be expressed numerically. Things like “scratch length 0.5 mm or more” or “foreign matter area 0.1 square mm or more” — expressible as dimension, area or brightness difference. “It does not look good” cannot be automated.
3. Over-rejection is tolerated. A machine will always reject some good parts. If you accept a 5% over-rejection rate with human re-judgement, the system can be introduced; demand zero over-rejection and escapes will occur instead. The workable design philosophy is that a person can recover an over-rejection, but nobody can recover an escape.

Where 100% inspection is automated and four inspectors (2 per shift × 2 shifts) can be reduced to one, the saving is 3 × 200,000 = THB 600,000/year, on top of which come reductions in claim handling and sorting for escaped defects. The THB 300,000/year of quality cost reduction booked in Case C is this element.

Incentive schemes you can use in Thailand and Vietnam, and their limits

The last piece of the investment decision is the incentive schemes. This is a widely misunderstood area, so the scope matters.

Thailand BOI Announcement No. 4/2569 (automation and robotics promotion)

Announcement No. 4/2569, published on 31 March 2026, promotes the adoption of automation and robotics. The main points are as follows.

Important caveat: this announcement is for automotive-related categories, not for all industries. Always confirm with the BOI or a professional adviser whether your company falls within the eligible categories (Tilleke & Gibbins).

Here is what eligibility would be worth for Case C. The following is an illustrative calculation of how such a scheme works; the actual reduction depends on each company’s taxable income.

Suppose a company with an annual corporate income tax charge of THB 1,000,000. With a 50% reduction over three years:

• Reduction per year: 1,000,000 × 50% = THB 500,000/year
• Three-year total: 500,000 × 3 years = THB 1,500,000

Applying that to the cash flow of Case C (investment THB 3,800,000, annual saving THB 980,000):

At the end of year 2 the cumulative figure is THB 2,960,000, leaving THB 840,000 against the THB 3,800,000 investment. Year 3 brings in THB 1,480,000, so 840,000 ÷ 1,480,000 = 0.57 years. Payback is therefore 2 + 0.57 = about 2.6 years, roughly 1.3 years shorter than the 3.88 years without the incentive.

The same equipment can differ by more than a year in payback depending on whether an incentive applies. That is why the investment plan and the BOI application should proceed in parallel. For the overall design of FA investment including BOI privileges, see also Factory Automation in Thailand: Implementation Sequence, Costs and BOI Incentives.

Thailand depa 200% deduction scheme (for SMEs)

The second scheme is the 200% deduction for SMEs operated by depa (the Digital Economy Promotion Agency) together with the Revenue Department.

Here is the effect if the cap is fully used. Against expenditure of THB 300,000, you can deduct 300,000 × 200% = THB 600,000, which is THB 300,000 more deduction than normal. The actual tax relief is that amount multiplied by the applicable corporate income tax rate. Rates differ by company classification, so as an illustration at 20%, 300,000 × 20% = THB 60,000 of tax saving. Always confirm the applicable rate with your tax adviser.

The amount is not large, but the value of this scheme is that its size fits measures in the THB 150,000 to 300,000 bracket, such as stage 2 poka-yoke devices or IoT for the recording and reporting process. For an SME considering a small start at stages 1 and 2, part of the investment effectively comes back. The condition is that the product is registered in the Thailand Digital Catalog, so check eligibility before purchasing (ThaiPR.NET / Mahanakorn Partners).

Points that apply to any scheme

• Apply before you place the equipment order. Many schemes exclude expenditure already ordered or paid, so start talking to whoever handles incentives at the earliest stage of the investment plan.
• Write the payback case without the incentive as well. Management will not approve a plan that collapses if the privilege is not granted. Treat incentives as upside.
• Schemes change. BOI Announcement 4/2569 accepts applications until the end of 2027, and the depa measure runs to 31 December 2027. Both have deadlines, so postponing the decision has a cost.

A 90-day roadmap to start labour-saving automation

Finally, here is the whole thing in a form you can start tomorrow. The goal is: within 90 days, measure the actual effect of stages 1 and 2, and complete the investment case for stage 3 and beyond.

Three rules to follow in this roadmap

1. Never skip days 1 to 15. Choosing measures without a stocktake means writing a capital request with no basis for the saving, so it is either rejected or approved but unmeasurable. These are the most important 15 days of the 90.
2. Keep the “measure the same way before and after” rule in days 46 to 75. Change the measurement method after start-up and the effect cannot be compared. Same categories, same time of day, same observer.
3. Put the measured values into the days 76 to 90 investment case. Knowing the gap between plan and actual at stages 1 and 2 improves the accuracy of estimates for stage 3 and beyond. If you achieved 80% of plan, assume 80% for stage 3 as well. Whether or not you hold that correction factor makes a large difference to your later success rate.

What changes in 90 days

Using the Case A numbers, that is an investment of THB 230,000 for an annual saving of THB 224,500. If start-up happens in the second half of the 90 days, roughly half a year of effect can be booked within the same fiscal year: 224,500 ÷ 2 = about THB 112,000. The amount is small, but the track record of being an organisation that can deliver labour saving as planned is the strongest material you will have when the next request runs into the millions of baht.

Frequently asked questions

What is labour-saving automation, and how does it differ from effort reduction and full automation?

Labour saving means reducing the headcount or man-hours a process requires. Effort reduction means reducing the burden on the operator. Automation means replacing human judgement and motion with a machine. If the work becomes easier but the operator still stands on the line, labour cost does not fall. Automation is also not a prerequisite for labour saving: in many cases digitising records or clarifying pass/fail criteria alone removes man-hours. In a capital request, describe it as “how many man-hours, from when, and how much money”.

How cheaply can labour-saving equipment be started?

As a guide from the projects we take on in Thailand, jigs and simple modification start from THB 30,000 to 150,000, and poka-yoke devices from THB 50,000 to 300,000. In the representative cases in this article, a THB 80,000 jig gives THB 74,700 a year (payback 1.07 years), and a THB 150,000 poka-yoke device gives THB 149,800 a year (payback 1.00 years). Build a track record of about one-year payback at the level of a few hundred thousand baht, then use it to fund a move to a semi-automatic machine (THB 300,000 to 1,500,000). That small-start sequence fails least often.

Can high-mix low-volume production be automated?

Yes, on one condition: do not aim for “all variants on one machine”. Run a Pareto analysis of volume by variant and automate only the top two or three variants that make up 70 to 80% of the quantity. People make the rest. In high-mix plants, changeover consumes a lot of man-hours, so shortening changeover with stage 1 measures (one-touch clamps, locating jigs) often gives a better return than installing equipment. In the calculations in this article, changeover reduction alone is worth THB 49,800 to 224,100 a year.

How do I calculate automation cost-effectiveness, and what payback period is acceptable?

The formula is: annual saving = labour saving + defect reduction + changeover reduction − additional labour − increased running cost. The payback period is initial investment divided by annual saving. Make sure both negative terms (maintenance staff, consumables and power) are included. On the evidence we see, a reasonable payback is 1 to 2 years for investments under THB 500,000, 2 to 3 years for THB 500,000 to 3,000,000, and 3 to 5 years for THB 3,000,000 to 10,000,000. Above THB 10,000,000, the realistic approach is to judge it together with a capacity increase, a new model or an incentive scheme.

In what kind of factory can an unmanned night shift work?

In a process where four conditions hold: (1) eight hours of material can be fed continuously, (2) eight hours of finished parts can be discharged and stored, (3) abnormalities trigger a safe stop and a notification, and (4) every piece is judged in line. The most frequently overlooked is (2), the discharge side: because nobody designs where finished parts go, a person has to collect them every two hours. If the four conditions are not fully met, the safe route is to run unattended for four hours first and extend the period in steps.

Should I go to an automation equipment manufacturer or to a system integrator?

If a single process can be completed by a single machine, an automation equipment manufacturer with a dedicated machine has the advantage. Where the work spans multiple processes, involves retrofitting existing equipment, or connects to host systems, a system integrator is a better fit. The dividing question is whether you are buying a machine or redesigning a process. In general, stages 1 to 3 in this article belong to integrators and jig shops, stage 4 usually combines an equipment maker with an integrator, and stage 5 is integrator-led.

Is it better to retrofit existing equipment or to buy new?

When capacity is idle — Thailand’s average utilisation was 57.47% in the second quarter of 2026 — retrofitting is worth considering first. Adding a loader, unloader or sensors to existing equipment often costs a fraction of new equipment and takes no extra floor space. However, if the control panel is old and parts are no longer supplied, or if there are no drawings, retrofit cost rises sharply, so check the condition of the control panel and the availability of drawings before deciding.

What do we do with the people freed up by labour-saving automation?

This is a management question rather than a technical one, but in practice it is the most important one. Thailand is short of roughly 500,000 workers across manufacturing, agriculture and construction, and most plants absorb the change not by letting people go but by filling vacancies they have been unable to recruit for. It is also standard practice to redirect part of the freed capacity into developing maintenance staff. The “one maintenance and changeover technician” booked in Case C is exactly this: move the person freed by labour saving into maintenance, and you contain outsourcing cost while raising equipment availability.

Summary

Labour-saving automation is not an equipment selection exercise. It is the activity of putting a price tag on human man-hours and deciding the order in which to remove them. The key points:

• Define labour saving, effort reduction and automation separately. If the work gets easier but nobody leaves the line, cost does not improve by a single baht.
• Thailand’s average capacity utilisation is 57.47%. Equipment is idle; what is short is people. So man-hour-reducing labour saving comes before capacity-adding automation.
• Wages keep rising. Thailand’s minimum wage went from THB 372 to THB 400 (about 7.5%), which for a 100-person plant is THB 840,000 a year of extra fixed cost. Vietnam is +7.2% in 2026 and a proposed +7.8% in 2027, about 15.6% over two years.
• Take stock of man-hours before you start. In the example of one line with ten people on two shifts, non-value-adding man-hours alone are THB 1,792,800 a year. Pricing that block is the starting point of every investment decision.
• Climb the five stages in order. Jigs (payback 1.07 years) → poka-yoke (1.00) → semi-automatic machine (2.50) → standalone automation (3.88) → line linking (6.67). Skip a stage and you will be missing either the standard, the part accuracy or the maintenance capability.
• Across all three cases: investment THB 5,230,000, annual saving THB 1,684,500, payback 3.10 years. Small investments have the higher efficiency; large ones have the larger absolute saving. That is why you proceed in stages.
• Incentives change payback by more than a year. For a Case C project eligible under BOI Announcement 4/2569 (automotive categories only; CIT halved for three years), the calculation shortens payback from 3.88 years to about 2.6 years. The depa 200% deduction (capped at THB 300,000) fits the size of stages 1 and 2.
• In 90 days you can reach measured results for stages 1 and 2 and a completed investment case for stage 3 onwards.

If you would like to apply this to your own line, start by building the man-hour stocktake table from this article. Filling in the hourly rate and the man-hours in the five categories is enough to produce the value of the reduction available to you. Looking at that result, we can also work through with you which stage of measure is likely to suit which process, and roughly what investment range to expect. TOMAS TECH is based in Bangkok and has done hands-on work for factories in Thailand and Vietnam, from a single set of jigs through to standalone automation and line linking. Show us the output of your stocktake and we will give you a frank view on priorities and indicative investment ranges. You are welcome to get in touch through our contact form, including at the exploratory stage.

References

• Bangkok minimum wage raised to THB 400 per day from 1 July (JETRO Business Brief, in Japanese)
• Thailand’s industrial output falls in June 2026 (Xinhua, 27 July 2026)
• Thailand Unveils New Incentives for Automotive and HEV/PHEV Manufacturing (Tilleke & Gibbins)
• 200% tax deduction for SMEs by depa and the Revenue Department (ThaiPR.NET)
• Thailand Approves New Tax Incentive to Accelerate SME Digital Transformation (Mahanakorn Partners)
• Vietnam’s New Minimum Wage from January 1, 2026 (Vietnam Briefing)
• Regional minimum wage up 7.2% from 2026 (VietnamPlus)
• Global Robot Demand in Factories Doubles Over 10 Years (IFR)
• Thailand extends work permits for migrant workers (Mizzima, 22 July 2026)
• Businesses warn of worker shortage (Bangkok Post)
• Thailand’s labour market and ageing (IDE-JETRO, in Japanese)
• Thai economy and the working-age population (Bangkok Shuho, in Japanese)