Automated Forklift (AGF) 2026 | What Stops It Is Pallets and Drop Points, Not Travel

Automated forklift projects usually stall at the approval stage not because the vehicles are slow, but because the fleet size does not come down. Whether the pick failure rate is 12% or 1.5% splits the average cycle into 7.44 minutes versus 6.18 minutes, and it splits the required fleet into 5 units versus 4 units. The payback periods are 16.5 years, 12.3 years and 8.6 years. Apply BOI incentives and 8.6 years becomes 4.3 years, and under the right conditions 2.5 years. What creates that gap is not the vehicle. It is the pallets and the drop points.

What stops an automated forklift is not travel

An automated forklift is also called an automated guided forklift, or AGF. We use AGF from here on.

When you visit a demonstration, the vehicles run well. They turn in narrow aisles, stop when a person crosses, and start again. Almost nobody comes back from a demo doubting travel performance.

What actually happens after installation, however, is not a travel problem. The forks do not enter the pallet pockets. They enter, but one side is too shallow. At the drop point, a pallet is sticking out halfway, so there is no room to place the load. Loose stretch film hangs down and the sensor misreads it as an obstacle. Each time, the vehicle fails safe and stops, someone is called, a person corrects it by hand, and operation resumes.

Each one of those stops turns directly into fleet size. The reason is simple: the required number of automated forklifts is decided by “vehicle-minutes needed per day divided by the minutes one unit can work.” More failures means a longer average cycle per trip, which means more required vehicle time, which means one more unit. A standard model is THB 3,900,000 per unit, and THB 4,300,000 once you add the charging equipment.

In this article we set up one model of a Japanese-owned factory in Thailand with an attached warehouse, and follow this structure through to the end in numbers. Here is the conclusion up front. If you cut the THB 3,000,000 of drop point and pallet work, you end up buying a fifth unit that costs THB 4,300,000. And on top of that, your annual savings stay THB 655,166 lower every year.

Note that the calculations below are built as “one baseline and three designs.” All three designs are measured against the same baseline, so their effects cannot be added together. We come back to this point at the end.

Assumptions | 750 pallets per day, 2 shifts, 264 days, Japanese-owned factory in Thailand

We set only one set of assumptions. Every number that appears later comes out of this table.

How the figure of 750 moves is set

Pallet transport of 750 moves per day is the total of unloading at receiving, supply from the warehouse to the processes, inter-process moves, and shipping of finished goods. This count differs completely from factory to factory, and in practice we recommend that you measure it on site first. The first thing you need when studying an automated forklift is not a vehicle catalogue, but the actual count of how many pallets move from where to where each day.

And if you can reduce the number of moves itself, that is the cheapest improvement of all. Changes such as revising the process layout so that inter-process transport disappears, or reviewing order lot sizes so that the number of receiving events drops, lower the required fleet size without buying a single vehicle. We cover this kind of review separately in In-plant logistics improvement. This article assumes that “the 750 moves cannot be reduced” and focuses on how to carry those 750 moves.

Basis for setting effective operator labour cost at 420,000 baht

According to SalaryExpert, the average annual salary of a forklift operator in Thailand is 375,699 baht, which converts to 181 baht per hour. There is a range: entry level (1 to 3 years of experience) is 287,906 baht and senior level (8 years or more) is 452,645 baht. In this article we multiply the average of 375,699 baht by 1.12 for statutory contributions and benefits, and round to 420,000 THB per year per person as the effective labour cost.

For reference, the statutory minimum wage in Thailand is 337 to 400 baht per day (400 baht in Bangkok and Phuket). Forklift operators are certified personnel who have completed the legally required training described later, so they sit in a different pay band from minimum wage workers. If you calculate this line at minimum wage, you underestimate the cost of manned operation and therefore estimate the effect of automation as smaller than it really is.

Basis for setting AGF effective operating time at 1,200 minutes

We treat 2-shift operation as 24 hours per day, then deduct 4 hours for charging and inspection, leaving 20 hours, that is 1,200 minutes per day, as the effective operating time per unit. This changes depending on the battery type and the charging practice, but if you assume the vehicle runs a full 24 hours you will always end up short on fleet size, so the deduction should always be included.

On the other hand, actual working time for a manned operator is set at 480 minutes x 80% utilisation = 384 minutes per day. This is an effective value that includes breaks, morning meetings, paperwork and waiting.

The gap between these two numbers is the primary effect of removing the driver. A person gives 384 minutes, a vehicle gives 1,200 minutes. However, if you calculate fleet size from that gap alone, you will always be wrong. The reasons are handled from the next section onward.

Baseline | THB 5,130,500 per year for manned operation

First we convert the current manned forklift operation into money. This is the only baseline in this article.

Calculating the required headcount

Required working time = 750 moves x 4.5 minutes = 3,375 minutes per day

Actual working time per operator = 480 minutes x 80% = 384 minutes per day

Required headcount = 3,375 / 384 = 8.79, so 9 people

The 9 people break down into 5 on the day shift and 4 on the night shift. Five manned forklifts are in operation.

Annual cost

THB 5,130,500 per year. Of that, THB 3,780,000 is labour cost. This is where an automated forklift has the largest room to cut, which also means that if you design in a way that keeps the people, you will get almost no effect. The three designs described later all assume that 1 to 2 manned staff remain.

The THB 260,000 for handling-related damage and contact incidents includes pallet damage, crushed product corners, contact with racking and walls, and the cost of recovering collapsed loads. This item goes down with automation, but it does not reach zero. How much it goes down differs by design, and that matters later.

When we show this table on site, people often say “vehicle running costs are smaller than I thought.” That is correct: the THB 1,080,000 of running cost for 5 units is not even as much as 3 operators at THB 420,000 each. The cost of forklift operation attaches to people, not to vehicles. That is exactly why the decision to go driverless is decided by how many people you can remove, not by the vehicle price.

Fleet size is decided by failure rate, not distance

This is the heart of the article.

The average cycle formula

The cycle per move of an automated forklift is not decided by the time when things go well. The recovery time when things fail is added on in proportion to the failure rate.

“`

Average cycle = (1 – failure rate) x base cycle + failure rate x (base cycle + 12 minutes)

Required vehicle time = 750 moves x average cycle

Required units = required vehicle time / 1,200 minutes (rounded up)

“`

In this model, a move where picking or dropping fails costs +12 minutes. That covers the safe stop of the vehicle, the call from the host system, a person walking to the site and correcting it by hand, and restarting the vehicle. 12 minutes may look long, but once you include travel time when the failure happens at night or on another floor, it is within a realistic range.

The base cycle is set at 5.4 minutes for a high-end model with fast travel, and 6.0 minutes for a standard model. This 0.6-minute difference is the part that stands out most when you compare catalogues.

The three designs

Design A and Design C keep the failure rate at 12% and change only the vehicle. Design B keeps the standard vehicle and lowers the failure rate to 1.5%.

*Figure 1: Average cycle for the three designs (6.84 min / 7.44 min / 6.18 min) and the resulting required fleet size (5 units / 5 units / 4 units)*

The boundary for fitting into 4 units is 6.4 minutes

Whether you fit into 4 units depends on whether required vehicle time is at or below 1,200 minutes x 4 units = 4,800 minutes. Converted into average cycle, the boundary is 4,800 / 750 = 6.4 minutes.

Design A exceeds the boundary by only 0.44 minutes, even though it uses a high-end model with fast travel. Seen in required vehicle time, that is an overrun of 330 minutes. You end up buying a fifth unit for the sake of those 330 minutes.

Counting the vehicle time that failures consume

Let us break this down one more level. Out of 750 moves per day, the number of failures is as follows.

In a factory with a 12% failure rate, 1,080 minutes of vehicle time is spent every day recovering from failures. That is 0.90 of a vehicle. Put another way, a site with a 12% failure rate is buying close to a whole vehicle just to redo picks.

Let us verify. Standard model base cycle 6.0 minutes x 750 moves = 4,500 minutes. Add the 1,080 minutes from a 12% failure rate and you get 5,580 minutes (Design C); add the 135 minutes from a 1.5% failure rate and you get 4,635 minutes (Design B). These match the table. For the high-end model, 5.4 minutes x 750 moves = 4,050 minutes, plus 1,080 minutes, gives 5,130 minutes (Design A).

A 16.9% reduction turns 5 units into 4

Lowering the failure rate from 12% to 1.5% moves the average cycle from 7.44 minutes to 6.18 minutes. The reduction is 1 – 6.18/7.44 = 16.9%.

Only 16.9%. Gaining 16.9% through travel speed is not easy, and as Design A shows, switching to a high-end model still left the fleet at 5 units. Yet the 16.9% gained by lowering the pick failure rate turns the required fleet from 5 units into 4 units.

This is the point that is most often misunderstood in automated forklift investment decisions. What the catalogue lists is travel speed, turning radius and lift speed, that is, “performance when things go well.” What decides fleet size is the number of times things do not go well.

Comparing the three designs | 16.5, 12.3 and 8.6 years

Unit price assumptions

First, the unit prices used across all three designs.

For reference, AGF prices inside Japan are said to be around JPY 15,000,000 to 20,000,000 or more per unit. Exchange rates, transport, installation scope and local service conditions are completely different, so this cannot be compared directly with a delivered price in Thailand, but it is worth keeping the sense that “an AGF can be an order of magnitude more expensive than an AGV.” We cover how to think about vehicle unit prices and how the cost structure differs from AGVs in detail in AGV price and implementation cost.

Also, the two-tier structure of peripheral equipment, “fixed 1,000,000 THB plus 400,000 THB per unit,” tends to be overlooked. When the fleet grows by one unit, it is not only the vehicle price that increases, but also 400,000 baht of charging equipment.

Initial investment

*Figure 2: Initial investment for the three designs, stacked into vehicles, peripheral equipment and civil work. Design B, which carries 3,000,000 of civil work, still totals less than Design C with zero civil work*

Design B, which carries THB 3,000,000 of civil work, is THB 1,300,000 cheaper in initial investment than Design C with zero civil work. This reversal is the structure we most want to convey in this article.

Annual running cost

Design A and Design C differ only in vehicle price; their operating conditions are identical (12% failure rate, 5 units), so their annual running costs are also identical.

Please look at three of these lines.

Manned staff retained. In Designs A and C with a 12% failure rate, 90 failures occur per day. Because people are also needed during the night shift, you have no choice but to keep 1 person per shift, 2 in total. Design B with a 1.5% failure rate has 11.25 failures per day, so 1 person on the day shift can handle both recovery and special handling. The failure rate decides not only the number of vehicles, but also the number of people you keep.

Handling-related damage and incidents. The baseline of THB 260,000 becomes THB 150,000 in Designs A and C, and THB 60,000 in Design B. Only Design B falls sharply, because pallets are standardised to plastic, drop points get positioning guides, and load presentation becomes uniform. Much of the damage arises from handling failures themselves, so lowering the failure rate lowers damage at the same time.

Plastic pallet replacement. Only Design B carries THB 276,000 per year. This assumes that THB 1,380,000 worth of plastic pallets is replaced over 5 years. An estimate that takes the benefit of the civil work but leaves out the replacement cost will always break down in year 5. Even with an item that works against Design B included, Design B still wins. That is how to read this table.

Annual savings and simple payback

All annual savings are differences from the same baseline of THB 5,130,500.

Everything on one sheet

Paying THB 7,500,000 for the high-end model adds not one baht of savings

Compare Design A and Design C. The annual savings are exactly the same, THB 1,822,167.

The reason is right there in the table. The high-end model has a base cycle that is 0.6 minutes faster, and the average cycle also shortens by 0.60 minutes, from 7.44 to 6.84 minutes. But because it is still above the 6.4-minute boundary, the required fleet stays at 5 units, the retained headcount stays at 2, and maintenance stays at 5 units’ worth. Not one operating condition changes, so not one baht of savings changes.

As a result, the difference in initial investment, 30,000,000 – 22,500,000 = THB 7,500,000, lands entirely as a worse payback period. The 4.2-year gap between 16.5 years and 12.3 years is the result of dividing that THB 7,500,000 by the same savings amount (7,500,000 / 1,822,167 = 4.1 years; subtracting the rounded figures in the table gives 4.2 years).

The comparison between Design A and Design B is even more extreme. Design A costs 1.42 times more in initial investment than Design B (30,000,000 / 21,200,000 = 1.4151), yet its annual savings are only 0.74 times as large (1,822,167 / 2,477,333 = 0.7355). You pay 1.42 times and get back 0.74 times. That is what the decision to “buy the faster vehicle” actually contains.

To be clear, this is not a claim that the high-end model is a bad product. The point is that on a site where the failure rate has not come down, there is nowhere for the high-end model’s performance to land. Only once picking is stable does the difference in travel performance show up in fleet size.

Cut THB 3,000,000 of civil work and you buy a fifth unit at THB 4,300,000

Let us line up Design C and Design B purely as differences. The two use the same standard model, and the only difference is whether the drop point and pallet work was carried out.

The moment you cut the civil work, the fifth unit is confirmed

Cut the THB 3,000,000 of civil work and the failure rate stays at 12%. The average cycle becomes 7.44 minutes, exceeding the 6.4-minute boundary by 1.04 minutes. The required fleet is 5 units. The amount needed for that fifth unit is vehicle 3,900,000 plus charging equipment 400,000 = THB 4,300,000.

To save THB 3,000,000, you pay THB 4,300,000. That is a net loss of THB 1,300,000, which matches the difference in initial investment, 22,500,000 – 21,200,000 = 1,300,000.

And you miss out on THB 655,166 every year

The loss is not limited to the initial investment. Design C continues in a state where annual savings are THB 655,166 per year lower than Design B (2,477,333 – 1,822,167 = 655,166). The breakdown is 420,000 for one manned operator, 420,000 for one unit of AGF maintenance, 90,000 for the difference in damage and incidents, and then the 276,000 of pallet replacement added back.

Let us verify. 420,000 + 420,000 + 90,000 = 930,000. Add the training difference, 2,333 – 1,167 = 1,166, and you get 931,166. Subtract the pallet replacement of 276,000 and you get 655,166. This matches the difference in the table.

In other words, a site that skips the civil work pays THB 1,300,000 more up front and then misses out on THB 655,166 of effect every year after that. The moment you write “we cut 3 million baht of civil work costs” in the approval document, both of those are confirmed.

The civil work is only 14.2% of the initial investment

One more view, as a ratio. The THB 3,000,000 of drop point and pallet work is 14.2% of Design B’s initial investment of THB 21,200,000 (3,000,000 / 21,200,000 = 0.1415).

When you look down a quotation from top to bottom, this is the line that looks easiest to cut. Vehicle prices do not fall much even with negotiation, and peripheral equipment is hard to touch because it says “safety measures.” Only the civil work looks like something you can cut at your own discretion.

However, this 14.2% is the only item booked as a cost in order to move the failure rate from 12% to 1.5%. Cut it and the effectiveness of the remaining 85.8% drops.

Four causes of pick failure and how to fix them

So what does it concretely mean to lower the failure rate from 12% to 1.5%? Let us open up the contents of the THB 3,000,000.

1,380,000 + 1,140,000 + 480,000 = 3,000,000. These three lines correspond to the four causes below.

Cause 1: Pallets are not uniform

On a site where wooden pallets are mixed in, the height and width of the fork pockets differ from pallet to pallet. Corners are chipped, stringers are cracked, nails are lifted, boards are warped. When a person is driving, they can match the fork height by eye and make fine adjustments. An automated forklift does not have that fine adjustment. If the sensor cannot recognise the pocket, that move is a failure.

The countermeasure is to standardise the transported items onto plastic pallets. In this model we budget 1,200 pallets x 1,150 baht = THB 1,380,000. The figure of 1,200 pallets comes from the total volume of pallets sitting in processes, in the warehouse and on trucks in order to sustain 750 moves per day. Always count the required number from actual data. If there are not enough, wooden pallets get mixed back in somewhere, and the moment they do, the failure rate jumps.

Plastic pallets are consumables. In this model we assume replacement over 5 years and book 1,380,000 / 5 years = 276,000 baht per year into running cost. If you implement without looking at this replacement cost, the failure rate returns in year 5, in the form of “pallets have deteriorated, so failures increased.”

Cause 2: Drop point positions are not fixed

Drop points in manned operation usually work on the basis of “roughly here.” Even where lines are painted, in practice loads end up 30 cm off, touching the neighbouring load, sticking a little into the aisle. People see that and avoid it, or nudge it over, and carry on.

An automated forklift inserts its forks at a defined coordinate. If there is no pallet at the coordinate, it fails; if the neighbouring load is too close, the forks cannot enter and it fails; if the load is sticking out, it cannot place and it fails.

The countermeasure is to build physical positioning guides at the 38 drop points. L-shaped stoppers, guides that catch the pallet corner, and three-dimensional constraints where a floor line alone is not enough. The purpose is not to “paint a line” but to “limit the placeable position to exactly one.” Even if a person places the load casually, the guide fixes the position physically. Without going this far, the failure rate does not come down.

Cause 3: The floor is not flat

The floor is easy to overlook. Joint steps, drainage slopes, sinking near truck doorways, traces of past repairs. Unevenness that a manned forklift only feels as a bump becomes body tilt for an automated forklift, shifts the fork height, and misses the pocket.

Furthermore, on vehicles that estimate their own position with SLAM or laser reflectors, floor unevenness enlarges the positioning error itself. The vehicle can travel, but it misses on the last few centimetres of positioning. Many cases of the symptom “it can drive but it cannot pick the load” have their cause in the floor.

The countermeasure is flatness repair. In this model we budget the fabrication and installation of positioning guides, floor flatness repair and line marking together at THB 1,140,000. There is usually no need to repour the whole floor; repairing only the routes the AGF travels and the stopping positions is normally enough. However, which areas should be repaired cannot be decided until the routes are decided, so this should proceed at the same time as layout design. Our thinking on aisle width, routes and intersections is summarised in AGV layout design.

Cause 4: Load presentation is not uniform

The fourth cause is not included in the civil work cost, because it is a matter of rules.

Cardboard boxes overhanging the pallet, stretch film hanging loose, stacks of uneven height, tilted loads. As long as people are carrying them, these pass as “well, it can be moved,” but with an automated forklift they cause sensor misreads and safety stops. If the end of the wrap hangs down, it is detected as an obstacle and the vehicle stops. This too costs 12 minutes each time.

The countermeasure is to set a load presentation standard and have the receiving side and the production side follow it. No overhang, proper handling of the end of the wrap, maximum height, level top surface. The cost is close to zero, but this is the hardest item to execute. Internal production processes can be changed by instruction, but changing the load presentation that arrives from suppliers requires negotiation.

For exactly that reason, the load presentation standard should be decided before the vehicles are ordered and communicated to the parties concerned. If you only start saying “we cannot pick this load presentation” after the vehicles arrive, the site will stop the automated forklifts and carry the loads manually instead. At that point it is no longer a discussion about failure rate, but about utilisation.

Mapping the four causes to countermeasures

The THB 480,000 for location signage and location master data setup is the “what is where” layer that sits on top of these four. It covers both physical signage and the master data on the host system side. We touch on this in the host system integration section.

Choosing a guidance method | Choose on “ease of change”

The main guidance methods for automated forklifts are magnetic guidance (magnetic tape), laser reflector guidance and SLAM. Real machines combine several sensors, and it is common for them to carry LiDAR, stereo cameras, an IMU and ultrasonic sensors together, using sensor fusion for position correction and obstacle avoidance.

Discussions about method selection tend to drift toward “which one is the most accurate,” but in practice a different axis matters.

Why choose on “ease of change” rather than “accuracy”

Factory layouts change. A new model comes in, lines are rearranged, warehouse slotting is changed, the customer changes. A layout that does not move at all for 5 years hardly ever exists in practice.

Magnetic guidance is simple in principle and stable, but every time you change a route you peel up the floor tape and lay it again. Even if the work itself is small, each time you stop the area, call a contractor, and add more management of tape wear and peeling. It is a method that is “stable and cheap, but incurs cost at every change.”

SLAM lets you change routes in software, but the map depends on the surrounding shapes. Temporary loads pile up, stacks of pallets change from day to day, and the view changes depending on whether a shutter is open or closed. In such environments, map updating and exception handling design become necessary. It is a method that is “strong against change, but requires the environment to be built up.”

Laser reflector guidance sits in between; it is stable in buildings where reflectors can be installed permanently. However, positioning becomes unstable when loads or people come between the reflector and the vehicle, so whether you can keep a clear line of sight to the reflectors at all times is a selection condition.

Inputs for the selection decision

Before deciding on a method, we suggest confirming the following three points.

1. Whether any layout changes are planned in the next 3 years. If so, compare methods including the cost of re-laying tape
2. How much the environment in the travel area changes from day to day. On sites with many temporary storage areas, decide the operating rules for SLAM map updates as well
3. What will guarantee the accuracy of the stopping position. The guidance method provides the accuracy of “travelling to that point,” while the accuracy of “inserting the forks” is carried by the positioning guides and sensors described in the previous section

The third point is the most important. Moving up to a higher-grade guidance method does not lower the pick failure rate. Travel accuracy and handling accuracy are separate matters, and what drives the calculations in this article is the latter. Spending time on the guidance method debate while postponing the discussion of drop points and pallets is the wrong order.

People and AGFs sharing the floor | ISO 3691-4 and Thailand’s 12-hour legal training

Installing automated forklifts does not mean people disappear. Even in this model we keep 2 manned staff in Designs A and C and 1 in Design B. One manned forklift also remains. The design should assume that people and AGFs use the same floor.

Safety standard: ISO 3691-4

The safety standard for driverless industrial trucks is ISO 3691-4. The current edition is ISO 3691-4:2023, published in June 2023, which specifies personnel detection fields, operating modes and the required performance levels of braking systems. The 2023 revision added definitions of “active detection field” and “operational stop.”

A second edition is reported to be under review, but this article does not assert whether it has been published or when it will be published. If you write an edition number into a procurement specification, please confirm the latest edition at the time of ordering with ISO and with the manufacturer. If you sign a contract with an outdated edition number in the specification, you will end up in disputes over conformity interpretation after delivery.

The practical points to confirm are as follows.

Thailand’s legal training: the 12-hour course

The Thai side requirements should also be covered. The Ministerial Regulation of the Thai Ministry of Labour on “Occupational Safety, Health and Environment Management Standards for Machinery, Cranes and Boilers B.E.2564 (2021)” was published in the Royal Gazette, Volume 138, Issue 52A, on 6 August 2021, and its Part 4 covers forklifts. The training requirement for forklift operators is stated as a minimum of 12 hours of classroom and practical training, and it came into force on 17 January 2025.

In this model we book this 12-hour course at 3,500 baht per person, renewed every 3 years.

In monetary terms this is a small item. However, being small in amount and being safe to ignore are two different things. As long as one manned forklift remains, its operator continues to be subject to the legal training requirement. If you conclude that “we automated, so training is no longer needed,” you will be cited in an audit.

Also, the scope of application varies with your own equipment configuration and work content. Which requirements apply to your company should be confirmed with the competent labour authority and your safety officer.

What to decide in the design of mixed operation

On top of the standard and the law, here are the items that should be decided as site operating rules.

• The boundary between AGF-only areas and mixed areas. If you make the whole site a mixed area, speed limits apply everywhere and the average cycle lengthens
• The procedure for people entering an AGF route. Who stops the vehicle, how, and who restarts it
• Who is called when a failure occurs. What area the 2 people in Designs A and C, or the 1 person in Design B, cover, and within how many minutes
• Priority rules at points where manned forklifts and AGFs cross
• Coverage during the night shift. Design B assumes 1 person on the day shift, so a separate procedure is needed for failures that occur at night

The last item is the practical warning point if you choose Design B. A 1.5% failure rate means 11.25 failures per day, and it does not become zero during the night shift. If you cut to 1 person without deciding who handles recovery at night, only the figures on paper will be achieved. This model assumes that night-time failures are either limited to transport that can be processed by the day shift the following morning, or kept within the scope that existing night-shift staff can cover as an additional duty. The conditions for this to hold differ by factory, so we recommend that you verify it against your own night-shift structure.

With BOI, payback moves from 8.6 years to 4.3 years to 2.5 years

The simple payback of 8.6 years above applies no BOI incentives at all.

What the scheme provides

The BOI “Smart and Sustainable Industry” measure (industrial upgrading measure) grants exemption from import duty on machinery plus 3 years of corporate income tax (CIT) exemption. The cap on the exemption is normally 50% of the investment amount. However, if 30% or more of the value of the machinery being upgraded qualifies as linkage with or support for Thailand’s domestic automation industry, the cap becomes 100%.

In the first half of 2026, this scheme received 132 applications worth about THB 17.2 billion (about USD 507.6 million). On 15 January 2026 the BOI announced a new package of investment promotion measures, and many of the measures are stated to be open for application until the last business day of 2027. Because the scheme has an application deadline, the timing of your study itself affects the payback period.

For the details of the scheme, eligibility and the order of application, please confirm with the BOI and the competent authority at the time of ordering. The calculations below are only one illustration applied to Design B.

Assumption: CIT exemption is capped by the tax you would have paid

The most common error when calculating the effect of a CIT exemption is to treat the cap amount as the benefit amount. A CIT exemption reduces tax payable to zero, so you cannot gain more than the tax you were going to pay in the first place. The benefit is the smaller of “the cap amount” and “the tax payable over 3 years.”

In this calculation we set the company’s effective CIT payment at 5,000,000 baht per year, or 15,000,000 baht over 3 years.

In the 50% cap case, the cap amount of 10,600,000 is smaller than the 15,000,000 of tax payable, so the cap binds. In the 100% cap case, the tax payable of 15,000,000 is smaller than the cap amount of 21,200,000, so this time the tax amount binds. Even if you obtain the 100% cap, the full investment amount does not come back.

*Figure 3: Comparison of payback periods. Without civil work 16.5 and 12.3 years, with civil work 8.6 years, with the BOI 50% cap 4.3 years, and with the 100% cap 2.5 years*

The difference between 50% and 100% is 15,000,000 – 10,600,000 = THB 4,400,000, or 4.3 – 2.5 = 1.8 years in payback. Same machines, same savings, same civil work. The only thing creating the gap is “where you bought from.”

Please note, however, that this 2.5 years is a simple comparison. The CIT exemption is realised over 3 years, so at the 2.5-year payback point you have not yet received the full THB 15,000,000. If you put this in an approval document, it is safer to attach a separate cash flow table that lays out the year in which each tax effect arises.

Count whether you reach the 30% test before you order

The condition for the 100% cap is that 30% or more of the value of the machinery being upgraded qualifies as linkage with or support for Thailand’s domestic automation industry. Let us actually count this for Design B.

This build-up is an illustration, and what can be counted is subject to the BOI’s case-by-case judgement. The actual assessment should always be confirmed in advance. Note also that whether the THB 3,000,000 of drop point and pallet work can be counted in the “investment amount” that forms the base of the cap, and how far the “machinery value” that forms the base of the 30% test extends, are likewise matters for the BOI to judge case by case.

Machinery value of Design B = 15,600,000 (vehicles) + 2,600,000 (peripheral equipment) = THB 18,200,000

30% line = 18,200,000 x 0.3 = THB 5,460,000

You are short by THB 1,240,000.

How to fill that THB 1,240,000 becomes a question of procurement structure: raising the share of local installation and local modification of the vehicles themselves, or raising the local procurement share of peripheral equipment. And this cannot be moved once you have already selected the supplier. After you sign a contract to import the vehicles as a complete set from overseas, saying “we do not reach 30%, so we want to increase the local portion” comes too late, because the price structure is already fixed.

In other words, the BOI 100% cap is not a question of how the application form is written; it is a question of procurement design. The 1.8-year gap between 2.5 years and 4.3 years of payback is decided by how much local content you can build in at the quotation stage.

How to write this in the approval document

Here are three points to watch when writing an approval document that incorporates BOI.

1. Always show the figures without BOI alongside. If you write only 2.5 years without showing 8.6 years, the whole plan collapses if approval is not granted
2. State the assumption for CIT payable. The 5,000,000 baht per year in this calculation is an assumption, and if the actual tax payable is smaller, the benefit will be smaller too
3. State whether eligibility and timing have been confirmed. Eligibility, the order of application and the scope of eligible machinery should be confirmed with the BOI and the competent authority at the time of ordering

Host system integration | Instruction granularity and exception handling

The fixed THB 1,000,000 of peripheral equipment includes the wireless environment, host system integration and safety measures. It is smaller in amount than the vehicles, yet this is the side that governs the failure rate once you go live.

Instruction granularity | Who decides “from where,” “to where” and “what”

Instructions to an automated forklift come down to three pieces of information. Where to pick from, where to place, and what to carry. The question is who decides these three.

The annual savings in this model hold on the assumption that the host system generates transport instructions automatically. If you run it with a person issuing instructions each time, that person’s time remains, and retaining only 1 to 2 manned staff will not be enough.

Location master data drives the failure rate

The THB 480,000 for “location signage and location master data setup” listed in the previous section is the foundation for this automatic generation.

Location master data means giving each of the 38 drop points a unique address, so that the host system and the vehicles speak about locations in the same terms. When this is vague, failures like the following occur.

• The host system instructs “place at A-12,” but the actual A-12 is occupied (mismatch between inventory records and physical stock)
• The address is too broad, so 3 pallets can be placed at one address (the coordinate is not uniquely determined)
• A person placed the load at a different address (there is no signage, so nobody knows)

All of these push up the failure rate regardless of vehicle performance. Even if you build physical positioning guides, the instruction itself will be wrong if the addresses on the system side are not organised.

Exception handling | How to shorten the 12 minutes

In this model we book 12 minutes per failure. The breakdown was the safe stop of the vehicle, the call from the host system, a person walking to the site and correcting it by hand, and restarting the vehicle.

Lowering the failure rate comes first, but it is also worth designing in parallel to shorten the 12 minutes themselves. Even in Design B, 11.25 failures per day and 135 minutes of vehicle time are lost to failures.

• Detection and notification: at the moment of failure, whose smartphone receives which vehicle at which location
• Authority to recover: who is allowed to fix it by hand, and who is allowed to press the restart button. If this is vague, waiting time grows
• Rerouting: when one unit is stopped, do the remaining vehicles wait, or do they detour and continue with other transport
• Recording: automatically record how many failures occurred at which location. Without this, you cannot tell what to fix in order to lower the failure rate

The last one, recording, is especially important. A 12% failure rate is a plant-wide average; in reality it concentrates at particular locations and particular pallets. If you can capture failure counts by address, you can identify which drop points to fix first. Whether or not you include this recording function in the requirements for host system integration changes the speed of improvement one year after go-live.

How to proceed with implementation (4 stages)

Finally, a word about sequence. If you choose Design B, the order in which to start is fixed.

Stage 1: Measure

There are three things to count.

1. Pallet moves per day. From where to where, how many times. This corresponds to the 750 moves in this model
2. Current handling failure causes. Even with manned operation, “the forks do not go in easily” and “the drop point is occupied” happen daily. Recording these gives you an idea of the failure rate you would face after switching to AGFs
3. The number of drop points and the condition of each. This corresponds to the 38 locations in this model. Is there a guide, is the floor flat, is there an address

For the measurement period, please take a span that includes both busy and slack times. If the fleet is short on a day when receiving is concentrated, that day will fall back to manned operation.

Stage 2: Fix

Carry out the drop point and pallet work first. Before the vehicles.

There are two reasons. One is that, as calculated in this article, without the civil work the fleet grows by one unit. The other is that you can measure the effect of the civil work on its own. If you bring it in at the same time as the vehicles, you will not be able to tell whether the failure rate dropped because of the civil work or because of the vehicles.

If you complete the civil work first, handling time and damage improve even under manned operation. That is a track record from before the vehicles arrive, and it can also be used in the approval document.

Stage 3: Install

Install 4 vehicles, the peripheral equipment and the host system integration. What should be checked at this stage is not the travel performance of the vehicles, but whether the actual failure rate stays within the assumption.

For a certain period after go-live, we recommend recording failure counts by location and repeatedly eliminating the top 3. Failures are not evenly distributed. There will always be spots that Stage 2 did not fully fix, and failures concentrate there.

If you are applying for BOI, the scope of eligible machinery, the order of application versus ordering, and whether local procurement can be counted should be confirmed with the BOI and the competent authority before you select the supplier. Once you are in Stage 3, the procurement structure can no longer be changed.

Stage 4: Expand

Once the 4 units are stable, widen the transport scope. Work left to manned handling as special handling, routes excluded because of low frequency, transport to another building.

The thing to watch here is that widening the scope changes the average cycle. Adding routes with long travel distances lengthens the base cycle, and adding areas with high failure rates lengthens the average cycle. As a result, you will have to redo the fleet size calculation. Whether to add one more unit is once again a THB 4,300,000 decision.

Before expanding, we recommend calculating the average cycle of the transport you are adding. You can use the formula in this article as it is.

Frequently asked questions (FAQ)

How much does an automated forklift (AGF) cost?

In the Thailand delivery model in this article, a standard model is THB 3,900,000 per unit and a high-end model is THB 5,400,000 per unit (both including installation). On top of that come peripheral equipment at a fixed THB 1,000,000 (wireless network, host system integration, safety measures) and charging equipment at THB 400,000 per unit. For reference, AGF prices inside Japan are said to be around JPY 15,000,000 to 20,000,000 or more per unit, but exchange rates, transport, installation scope and local service conditions differ, so a direct comparison is not possible. What you should judge is not the price per unit, but how many units you will need. Five standard units is THB 19,500,000 and 4 units is THB 15,600,000, a difference of THB 3,900,000. That is not a gap you can close by negotiating unit prices.

What is the difference between an AGV and an AGF?

An AGF is a forklift-type AGV (automated guided vehicle). The biggest difference is that it has a positioning action of inserting forks into a pallet pocket. AGVs that carry loads on a deck or tow them have a wide load transfer surface and can tolerate some positional deviation. An AGF has to pass its tines through the limited opening of a pocket, so pallet quality, drop point positional accuracy and floor flatness directly determine success or failure. The failure rate discussed in this article is a problem specific to that insertion action. Things that were not an issue with AGVs show up as a difference in fleet size with AGFs.

How many automated forklifts do I need?

It is decided by required vehicle time divided by effective operating time per unit. In this model, that is 750 moves x average cycle / 1,200 minutes per unit (rounded up). Because the average cycle changes with the failure rate, the answer changes even for the same factory. With a 12% failure rate and a standard model, 7.44 minutes gives 5,580 minutes and therefore 5 units; with a 1.5% failure rate, 6.18 minutes gives 4,635 minutes and therefore 4 units. The boundary for fitting into 4 units is an average cycle of 6.4 minutes (1,200 minutes x 4 units / 750 moves). Even a fast high-end model, if the failure rate stays at 12%, comes to 6.84 minutes, exceeding the boundary by 0.44 minutes and landing at 5 units. If you want fewer units, start by measuring the failure rate.

Are there subsidies or incentives for automated forklifts in Thailand?

There is the BOI “Smart and Sustainable Industry” measure (industrial upgrading measure). It provides exemption from import duty on machinery plus 3 years of CIT exemption, with a cap of 50% of the investment amount. If 30% or more of the value of the machinery being upgraded qualifies as linkage with or support for Thailand’s domestic automation industry, the cap becomes 100%. For Design B in this article (initial investment THB 21,200,000), the calculation gives a net THB 10,600,000 and a payback of 4.3 years under the 50% cap, and a net THB 6,200,000 and a payback of 2.5 years under the 100% cap. In the first half of 2026 this scheme received 132 applications worth about THB 17.2 billion, and under the package announced on 15 January 2026 many measures are stated to be open for application until the last business day of 2027. Eligibility and the order of application should be confirmed with the BOI and the competent authority at the time of ordering.

Will an automated forklift work with our existing wooden pallets?

It will work, but the failure rate will not come down. When wooden pallets are mixed in, the height and width of the pockets differ from pallet to pallet, and chips, cracks and warping are added on top. A difference that a person can adjust for by eye becomes a failure for an automated forklift. In this model we booked THB 1,380,000 for standardising onto 1,200 plastic pallets and THB 276,000 per year for replacement over 5 years. If you skip this investment, you stay at a 12% failure rate and end up buying a fifth vehicle at THB 4,300,000. Pallets are consumables, so plan including the replacement cost. If replacement is neglected, the failure rate returns to its original level after a few years.

How many forklift operators can we reduce by installing automated forklifts?

In this model, against a baseline of 9 people (5 on the day shift and 4 on the night shift), the 12% failure rate design keeps 2 people (1 per shift) and the 1.5% failure rate design keeps 1 person (day shift only). What decides the retained headcount is not the number of vehicles, but the failure rate. If 12% of 750 moves per day fail, that is 90 failures; at 1.5% it is 11.25. Handling 90 failures including the night shift requires a person on each shift, while 11.25 can be covered by 1 person on the day shift. Note also that as long as one manned forklift remains, its operator continues to be subject to Thailand’s legal training (12-hour course, 3,500 baht per person, renewed every 3 years). Fewer people does not remove the training obligation.

Summary

Payback on an automated forklift is decided by the pick and drop failure rate, not by the travel performance of the vehicle.

1. The baseline is THB 5,130,500 per year. That is 9 operators at 3,780,000, 5 manned forklifts at 1,080,000, legal training at 10,500 and handling-related damage and incidents at 260,000. The cost attaches to people, not vehicles.
2. The failure rate decides the average cycle, and the average cycle decides fleet size. At a 12% failure rate, 90 failures and 1,080 minutes per day (0.90 of a vehicle) are consumed by recovery. At 1.5%, it is 11.25 failures and 135 minutes (0.11 of a vehicle).
3. The boundary for fitting into 4 units is an average cycle of 6.4 minutes. Design A at 6.84 minutes and Design C at 7.44 minutes exceed it and need 5 units; Design B at 6.18 minutes fits into 4 units.
4. Payback is 16.5, 12.3 and 8.6 years. The high-end model (A) costs 1.42 times more in initial investment than the standard model with civil work (B), yet delivers only 0.74 times the annual savings. Designs A and C have identical annual savings (1,822,167), so the THB 7,500,000 difference lands directly as a worse payback period.
5. Cut THB 3,000,000 of civil work and you buy a fifth unit at THB 4,300,000. On top of a net loss of THB 1,300,000, you miss out on THB 655,166 every year. The civil work is only 14.2% of the initial investment.
6. With BOI, 8.6 years becomes 4.3 years, and 2.5 years depending on the conditions. However, Design B is THB 1,240,000 short of the 30% line required for the 100% cap, and that is a point to close before the procurement structure is fixed.
7. The order of work is measure, fix, install, expand. The drop point and pallet work comes before the vehicle order.

One baseline, three designs. The three are alternatives measured against the same baseline, and their effects cannot be added together. The moment you choose Design B, the THB 1,822,167 of Design C no longer exists. It is already all inside Design B’s THB 2,477,333. If this double counting appears in a quotation, the figures will not match actual results one year after go-live.

And a word about how to treat schemes and standards. BOI eligibility and the order of application, the scope of Thailand’s legal training, and the edition number of ISO 3691-4. These three should be confirmed with the BOI and the competent authority at the time of ordering and before application. The figures in this article are one example of what happens when the conditions line up.

TOMAS TECH is based in Bangkok, Thailand, and provides integration of production management, factory IT/OT and FA systems for Japanese-owned manufacturers. On automated forklifts, if you can show us your daily pallet transport volume and the condition of your drop points, we can produce a fleet size and payback estimate in the same format as this article, using your own numbers. It is also fine if you are still at the stage before selecting vehicles and simply want to check whether “our drop points and pallets can support this at all.” Please feel free to get in touch through our contact page.

Sources referenced

• Forklift Operator Salary in Thailand (SalaryExpert)
• Minimum Wage in Thailand (Thai Law Online)
• Ministerial Regulation on Occupational Safety, Health and Environment Management of Machinery, Cranes and Boilers B.E.2564 (GB Planet)
• Forklift operator training courses compliant with the new law (NICHIYU Thailand)
• ISO 3691-4 (ISO)
• What changed in ISO 3691-4:2023 (Pilz)
• Smart and Sustainable Industry (Thailand BOI)
• Thailand secures $43.6bn 1H 2026 investment surge (Thailand Business News)
• Thailand’s BOI in 2026 (Mahanakorn Partners)
• AGF (automated guided forklift) prices and guidance methods (Hakuou)
• Southeast Asia Warehouse Automation Market Report (MarkNtel Advisors)
• Sensor configuration of an Automated Guided Forklift (JET Manufacturing)