Factory Automation Southeast Asia: Thailand Roadmap 2026

Across the region, the conversation about factory automation in Southeast Asia has quietly shifted from “should we do this?” to “in what order, and where do we start?” In Bangkok, the minimum wage rose to THB 400 per day for all business sectors from 1 July 2025 (JETRO, based on the National Wage Committee resolution of 17 June 2025), while the International Federation of Robotics (IFR) reports in *World Robotics 2025* that 542,000 industrial robots were installed worldwide in 2024, with Asia accounting for 74% of them. On paper, automation looks inevitable.

Stand on an actual shop floor, though, and the picture is messier: twenty-year-old machines sharing a line with current-generation PLCs, no way to hire an engineer who can commission a cell, and a capital request sitting untouched at regional HQ for six months. This article is not a generic explainer of what factory automation is. It sets out, using published data and current policy information, the order in which manufacturers with plants in Thailand should approach automation when budget and headcount are both constrained. The short version: buying a robot first is rarely the move. Getting to a state where equipment and process data actually exist is what determines the return on everything that follows.

Why factory automation in Thailand is no longer optional

Cheap labour no longer settles the argument

For a long time, the logic in Thai plants was sound: labour costs a fraction of what it does in Japan, Europe or the US, so adding people beats buying an expensive robot. That premise is eroding.

As reported by JETRO, Thailand’s National Wage Committee resolved on 17 June 2025 to apply a THB 400 daily minimum wage to all business sectors in Bangkok from 1 July 2025. The THB 400 rate had previously applied only to selected sectors and provinces; it has now been extended across every sector in the capital. Further increases have been reported as a government target, but they are not confirmed policy — so they should not be built into an investment case as fact. Treat them as a scenario in which the upward trend continues, nothing more. (This is a Thailand-specific wage framework and does not apply to Vietnam, Indonesia or other ASEAN markets.)

What matters more than the headline rate is the structure of total labour cost. Shop-floor labour is not just base pay. It is social security, overtime premiums, shuttle buses, dormitories and canteens, recruitment cost, training cost, and the retraining cost that follows every resignation. On lines running three shifts or chronic peak-season overtime, an increase in base pay is amplified through the overtime multiplier. The right unit of analysis is not “what is the hourly rate?” but “what does it cost us to produce 10,000 units in one shift?” Without that view, the automation business case cannot be evaluated correctly.

Labour availability bites harder than wage inflation

Mordor Intelligence’s Southeast Asia industrial and service robot market report names labour shortages in Singapore and Thailand as one of the region’s growth drivers. In Thai manufacturing, the problem is not only that wages are rising — it is that the right number of people with the right skills cannot be assembled at all.

The symptoms are easy to spot on any plant walk:

• Experienced inspectors retire or leave, and successor training cannot keep pace
• Night-shift vacancies go unfilled, so effective capacity is capped by the day shift
• Peak demand is covered with agency or subcontract labour, and quality variation increases
• Only one or two people can operate a particular machine, so the line stops when they are absent

None of these is solved by hiring more people. They are not headcount problems; they are repeatability-of-skill problems. The value factory automation genuinely delivers is less “labour cost reduction” than “reproducibility of work and judgement.” Whether a plant makes that mental switch changes the entire evaluation framework after go-live.

On top of this, external requirements keep rising: traceability demands from OEM customers and Western buyers, quality records at individual-unit level, visibility on on-time delivery. Paper and spreadsheets absorb ever more man-hours to satisfy those demands, whereas equipment that reports data automatically cuts the indirect effort of audits and regional HQ reporting directly. Put the other way round: evaluating automation purely as direct-labour reduction systematically undervalues it.

Industrial robots in numbers: the world, Southeast Asia and Thailand

Global installations have doubled in a decade

According to IFR’s *World Robotics 2025* (data year 2024), 542,000 industrial robots were installed worldwide in 2024 — the second-highest figure on record. The operational stock reached 4,664,000 units, up 9% year on year. IFR notes that annual installations have doubled over the past ten years.

By region, Asia accounted for 74% of new installations, followed by Europe at 16% (85,000 units) and the Americas at 9% (50,100 units). China alone installed 295,000 units, 54% of the world total and an all-time national record, with an operational stock above two million. Mature markets contracted: Japan installed 44,500 units (-4%, stock 450,500), Germany 26,982 units (-5%) and the United States 34,200 units (-9%).

IFR forecasts 575,000 installations in 2025 (+6%) and more than 700,000 per year by 2028.

(Source: IFR, *World Robotics 2025*)

The takeaway from this table is not “China is impressive.” It is that the robot procurement landscape itself has been reorganised around Asia. European and Japanese vendors and their distribution networks were once effectively the only option; in today’s Thai market, multiple realistic options exist, Chinese suppliers included. For a buyer, that is a tailwind on price negotiation — but, as discussed below, a dangerous one without a proper framework for judging maintenance, spare-part supply and technical support.

Robot density: the absence of a Thai figure is itself informative

IFR’s robot density release published in April 2026 (2024 data) puts the world average at 132 units per 10,000 manufacturing employees, Asia at 131, Western Europe at 267, North America at 204 and the EU27 at 231. By country, Korea leads by a wide margin at 1,220, followed by Singapore at 818, Germany at 449, Japan at 446 and the United States at 307.

(Source: IFR, “Robot density surges in Europe, Asia and the Americas”, published April 2026, 2024 data)

Thailand’s density is not included in that press release, so this article does not quote a figure for it. What the numbers do show is that Singapore — another ASEAN member — sits among the most robot-dense economies in the world at 818 units. In other words, “Southeast Asia is behind on automation” is not a generalisation that survives contact with the data. Maturity varies enormously between countries, and between companies inside the same country.

Where Thailand sits within the Southeast Asian market

Mordor Intelligence sizes the Southeast Asian industrial and service robot market at USD 1.21 billion in 2025, growing to USD 1.83 billion by 2031, a CAGR of 7.24% over the 2026–2031 forecast period. Within that, Thailand accounts for roughly 22% of regional revenue, underpinned by annual vehicle production of around 1.9 million units.

The same report states that some 4,500 new robots have been deployed in the Eastern Economic Corridor (EEC), and that measures such as eight-year corporate income tax exemptions and duty waivers have attracted USD 3.2 billion of automation-related foreign direct investment. (The EEC is a Thailand-specific zone; its incentives do not extend to neighbouring countries.)

Growth drivers identified in the report include Industry 4.0 subsidy programmes across ASEAN states (frameworks supporting up to 70% of eligible automation spending), labour shortages in Singapore and Thailand, China+1 relocation of electronics component production (USD 13 billion of FDI into Vietnam and Malaysia), autonomous mobile robots in e-commerce logistics, and service robots for healthcare and hospitality.

The restraints, meanwhile, will sound painfully familiar to anyone running a plant in the region:

• Low migrant-labour wages make capital equipment hard to justify, so ROI is thin (particularly in Vietnam and Indonesia)
• Fragmented factory utilities make system integration difficult
• A weak domestic component base means import duties and long lead times
• Robotics talent shortages delay commissioning

“ROI is hard to prove,” “integration is hard,” and “we can’t find the people.” Those three are the real barriers to factory automation in Thailand, and the second half of this article is devoted to dealing with each of them.

Installed base and ecosystem inside Thailand

On Thailand’s own track record, a commentary published for Japanese-affiliated manufacturers in Thailand reports that the country installed approximately 3,300 industrial robots in 2022, ranking 14th globally and second in Southeast Asia (behind Singapore). Adoption is spread across automotive, food and beverage, electronics, metalworking, pharmaceuticals, plastics and education.

For professional service robots, the same source reports 1,103 units operating in Thailand as of 2023, of which delivery accounted for 44.4% (roughly 490 units), hospitality 20.9% (roughly 230 units) and professional cleaning 19.9% (roughly 220 units). Monthly rental rates for service robots fell below THB 10,000 per unit in 2023, and three-year lease cycles have become the norm.

The same commentary lists the constraints on this market: insufficient market infrastructure and operational support, saturation of Bangkok’s restaurant segment, price competitiveness of Chinese suppliers, an immature domestic robotics base, and — most importantly — limited understanding of the optimal number of robots and of human–robot collaboration models. That last point deserves emphasis. How many robots to install is a design question, not a technical one, and the ability to decide that a given step is still faster performed by a person is often what separates a successful project from a stalled one.

2026 development: humanoid component production in the EEC

More recently, it was reported that in February 2026 the BOI approved investments totalling over THB 10 billion by five Chinese companies, establishing Thailand’s first production base for humanoid robot components in the EEC.

This will not change anything on your line next quarter, but it is worth reading as a signal. As Thailand’s robotics supply chain deepens, component lead times and local service capacity should improve over the medium term. It is also consistent with the direction of the new BOI incentive discussed below, which rewards sourcing linked to Thailand’s domestic automation machinery industry — in other words, policy and industry are both moving towards making local sourcing the more advantageous choice.

Five barriers that stop factory automation projects in Thailand

Generic implementation methodologies for factory automation are available in any vendor brochure. What trips up plants in Thailand is usually the part that isn’t in the methodology. Here are five recurring barriers.

Barrier 1: Legacy equipment and mixed PLC generations

Most established plants in Thailand host machines bought in different eras on the same site: equipment relocated from the parent company’s home plant in the 1990s, locally procured machines from the 2000s, and current-generation lines installed in the last five years — each with a different maker, a different PLC generation and a different communication protocol. Some machines have no PLC at all and still run on relay logic.

Drawing up a “connect the whole plant” master plan as your first step, in that environment, is a guaranteed way to stall. The realistic approach is to capture signals from outside the machine without modifying it: read the existing andon stack light, measure current draw, or add a photoelectric sensor, and start by simply recording the facts of running, stopped and micro-stoppage. Internal machine data can wait for the next phase, and it will still arrive in plenty of time.

Barrier 2: Shortage of robot system integrators and engineers

As Mordor Intelligence explicitly notes, robotics talent shortage is a leading cause of commissioning delays. In Thailand you can find traders and manufacturers who will sell you a robot arm, but the number of firms that can act as a genuine robot system integrator in Thailand — designing the gripper, the fixtures and the safety system, integrating with the PLC and taking the cell through to stable production — is limited.

The consequences show up predictably:

• The order is placed, but commissioning runs far longer than planned
• Every teaching change requires calling the integrator back, so changeover becomes rigid
• The integrator’s lead engineer leaves and no handover documentation exists
• A maintenance contract is in place, but on-site attendance takes several days

The only point at which you can fix this is at the point of order. Write into the specification what your own people will be able to do, and how far, after commissioning. This is covered in more detail in the partner selection section below.

Barrier 3: An ROI model built on labour savings alone

The classic reason an automation request dies at regional HQ is that the payback rationale rests entirely on direct labour reduction. In a high-wage home-country plant, labour savings alone can carry the payback period. Run the same calculation in Thailand and the payback stretches, and the conclusion becomes “people are still cheaper.”

Mordor Intelligence identifies exactly this dynamic as a regional restraint: low migrant-labour wages against heavy capital costs make ROI hard to demonstrate. Given that structure, an ROI model for Thailand needs to accumulate more than direct labour: reductions in defects, rework and scrap; reduced man-hours for inspection and record-keeping; additional output made possible by less downtime; lower indirect effort for traceability compliance; reduced turnover and retraining cost; less overtime and weekend working; and the tax effect of any applicable BOI incentive. (Measurement indicators and monetisation logic for each of these are tabulated later in the article.)

The catch is that most of these cannot be calculated without a current-state baseline. In a plant where nobody knows how many minutes of micro-stoppage occur or how many hours go into rework, the benefit simply cannot be estimated. Which is precisely why visibility comes before automation.

Barrier 4: Approval processes and timelines at regional HQ

Investment at a Thai site is usually governed by the capital expenditure standards and approval workflow of the parent company or regional HQ. Three kinds of mismatch recur.

First, a mismatch in evaluation criteria. HQ looks at payback period; the site is acting to avoid the risk of simply not being able to staff the line. These are not the same language. The effective response from the site is to monetise the opportunity cost of the understaffed scenario — orders that cannot be accepted, penalties for late delivery — and present that alongside the payback figure.

Second, a mismatch in timing. BOI incentives carry application deadlines; the announcement discussed below has an application deadline at the end of 2027. If the corporate budget cycle does not line up with the policy deadline, the incentive is lost. Fix the policy deadline first and work the approval schedule backwards from it.

Third, a mismatch in specification authority. HQ engineering often wants home-market standard specifications, while the Thai site needs to prioritise spare-part availability and whether a local integrator can support the equipment. The pragmatic move is to write “must be maintainable in Thailand” into the specification requirements early.

Barrier 5: Price competition without an evaluation framework

The commentary cited earlier lists the price competitiveness of Chinese suppliers as a feature of the Thai market. In practice, quotations for nominally equivalent specifications can differ substantially. That in itself is not a problem — it means buyers have more options.

The problem is comparing those options with no framework for what the price difference represents. At minimum, add the following dimensions to your quotation comparison sheet:

A low price is not in itself a red flag. Restating every quotation on a total cost of ownership basis is, however, non-negotiable. Judge on whether concrete answers come back to the questions above — not on the supplier’s nationality.

A phased factory automation roadmap: visibility → partial automation → line integration → AI

This is the core of the article. When a plant in Thailand sets out to automate, what should be tackled in what order? In practice, framing the decision in four phases makes it far easier to manage — and it is the difference between a genuine smart factory in Thailand and a collection of expensive isolated cells.

There is a reason for this sequence. The following sections set out what matters in each phase.

Phase 0: Visibility — get to a state you can measure

This is the phase most often skipped and the one that must never be skipped. The reason is simple: without a baseline you cannot measure the effect of automation.

Automation is a means, not an end, and evaluating a means requires a before-and-after comparison. Yet in many plants, utilisation is compiled by hand at month-end, stop reasons pour into an “other” bucket, and output figures depend on a daily report the operator wrote from memory. Install a robot into that environment and, afterwards, nobody can say more than “it feels better.” Reporting to regional HQ turns qualitative, and the next investment request fails.

Four data sets are the minimum for Phase 0:

1. Time the machine was running and time it was stopped (with timestamps)
2. Why it stopped (changeover, material wait, micro-stoppage, breakdown, quality adjustment, and so on)
3. Good and defective counts (ideally broken down by defect type)
4. Measured cycle time and its variation

With those four, OEE can be decomposed into availability, performance and quality, and the largest loss and its location become visible. Which process to robotise should be an output of that analysis, not an input to it.

Importantly, Phase 0 does not necessarily require a large investment. As noted above, even legacy machines can yield run/stop status from stack-light signals or current measurement. Nor does it need to cover every line at once. Starting with the one line believed to be the bottleneck, or a handful of machines, and letting the organisation experience data changing a decision, is what generates the momentum for the next phase.

At this stage it is also worth clarifying the division of roles between the equipment data platform and the production management system that handles planning, actuals and inventory. The differences and the selection logic are covered in detail in our guide to choosing a production management system for factories in Thailand.

Phase 1: Partial automation — start by choosing the right process

Once Phase 0 data has identified the target process, move to partial automation. The governing principle here is do not automate the whole line at once.

Process selection can be organised along the following dimensions:

Processes involving heavy loads, high temperatures or hazardous work rise in priority not only on cost-benefit grounds but on occupational safety grounds. Likewise, processes where night-shift staffing is chronically difficult carry an opportunity cost that rarely appears in a spreadsheet — and can therefore turn out to have unusually strong returns.

On material handling, the benefits compound if you look beyond inter-process transfer to the connection between the warehouse and the production floor. The practical side of warehouse automation and WMS deployment is covered in our article on logistics DX and warehouse automation in Southeast Asia.

The classic Phase 1 failure is a robot that runs beautifully while the processes upstream and downstream cannot keep up, leaving line throughput unchanged. As the theory of constraints would predict, speeding up anything other than the bottleneck does not speed up the system. Verify this at the investment decision stage; with Phase 0 data in hand, the check takes minutes.

Phase 2: Line integration — eliminate the handwritten daily report

Once partial automation is working at several points, the next step is to connect them. The objective is not only physical linkage of material flow. In many cases linking the information delivers the larger benefit.

Concretely, the target state looks like this:

• Production instructions pass automatically from the host system to the equipment, reducing setup errors on changeover
• Output is booked automatically from the equipment, removing manual entry of daily reports
• Material lot information is linked to finished products automatically, so traceability enquiries can be answered immediately
• When a defect occurs, the affected lot, time window and machine can be identified at once
• Inventory data matches physical reality, cutting the effort spent investigating stocktake discrepancies

The benefit of this phase shows up in indirect man-hours and quality cost rather than direct labour. Audit response, complaint handling, root-cause investigation and HQ reporting are rarely tracked as “man-hours” day to day, yet they consume a great deal of manager and engineer time.

If back-office document processing is also in scope, the approach is discussed in our piece on AI-OCR for back-office automation in Thai and ASEAN plants. Shop-floor automation and indirect-function automation tend to be run as separate projects, but evaluating them together produces a more honest ROI figure.

The restraint Mordor Intelligence describes — fragmented factory utilities making integration difficult — is exactly what surfaces in this phase. Deciding your policy on communication standards, network and data model back in Phase 0 dramatically changes the cost of Phase 2 integration. Conversely, if each machine keeps being procured as a local optimum, later integration can become effectively impossible.

Phase 3: AI — turning prediction into action

Only once data is accumulating continuously and processes are integrated does AI become a realistic option. The typical use cases are predictive maintenance, improving the accuracy of visual inspection, root-cause analysis of quality defects, and optimisation of demand forecasting and production planning.

As a benchmark, figures compiled by an industrial AI publication report that, in cases where predictive maintenance has been operated in a mature state for 12 months or more, unplanned downtime fell by 31–47%, maintenance costs by 12–24%, and OEE improved by 8–15 percentage points. The same source reports that 28% of discrete manufacturing sites with 50 or more machines have AI monitoring in production use.

These are vendor-compiled reported ranges for deployment outcomes, not TOMAS TECH results, and they carry an explicit condition: “mature operation of 12 months or more.” They do not mean equivalent results appear immediately after go-live. They should be read as a level reached only once data has accumulated and shop-floor practice has settled. Nothing here is a guarantee of outcome.

Read the other way, those figures are also evidence that skipping Phases 0 to 2 and buying AI alone will not work. Predictive maintenance functions only when vibration, temperature and current data have been collected over a period and failure events have been labelled against them. Sign an AI tool contract in a plant with no data and the first twelve months are spent accumulating data and nothing else.

For the broader picture of AI in manufacturing, including where autonomous agents that execute tasks on their own fit in, see our overview of AI agents in manufacturing.

Why IoT equipment monitoring and production management underpin automation

The previous chapter argued “data first.” This one goes a level deeper into what IoT equipment monitoring and the production management system each own, and how they connect to automation.

Division of roles across three layers

With all three layers present, automation stops being a set of isolated points and becomes a system. A common misconception is that a production management system will also capture equipment data, or that connecting machines via IoT removes the need for production management. In reality the two differ in purpose and update frequency and do not overlap. Real-time machine monitoring and actuals management robust enough for cost accounting are different requirements.

Avoiding “we automated but can’t prove the benefit”

The most uncomfortable position in an automation project is being unable to demonstrate afterwards whether it worked. Avoiding that requires fixing four things before the investment decision:

1. The metrics to be measured (OEE, first-pass yield, lead time, labour cost per unit, and so on)
2. The pre-implementation baseline (at least one to three months of actual measurement)
3. The measurement method (who collects it, from which system, at what frequency)
4. How to separate out confounding factors (how volume swings and mix changes will be normalised)

The fourth is routinely underestimated. If order volume moves at the same time as the automation goes live, the metrics will move with it. Unless normalised indicators — per unit, per 1,000 pieces — are agreed at the outset, benefit measurement degenerates into an argument nobody can win.

What to do about legacy equipment

Old machines are the first obstacle in any IoT equipment monitoring discussion in Thailand. The practical options are broadly these:

• Stack-light signal capture: determine run/stop from the state of the existing andon tower. No machine modification, and the lowest barrier to entry
• Current measurement: use clamp-type current sensors on spindles or motors to infer operating state from load
• Add-on sensors: photoelectric or proximity sensors to count production
• Direct acquisition from the PLC: where a communication port exists and the protocol is known. Yields by far the richest data
• Handle it at replacement: make data output a mandatory requirement in the specification when equipment is renewed

The key point is that not every machine has to use the same method. As long as the different acquisition methods are normalised into a common format at the upper layer, analysis and management work perfectly well. Insisting that all equipment be connected identically is how a project never starts.

Factory automation cost, ROI and how to treat BOI incentives

Investment amounts vary enormously with the process, the number of units and the specification, so this article does not quote price benchmarks. Even for a nominally identical “one robot,” the total including gripper, fixtures, guarding, frame, conveyance, installation, commissioning and training can differ by an order of magnitude between projects. What can usefully be shared is the cost structure you need in order to compare and challenge quotations.

Line items that must appear in every quotation

• Robot body and controller
• Design and fabrication of the end effector (gripper or tool)
• Workpiece locating fixtures, frames and peripheral mechanisms
• Safety fencing, light curtains, safety PLC and risk assessment work
• Signal interface work with existing equipment
• Utility work such as power supply and air piping
• Delivery, installation and transport
• Teaching, trial runs and support through to volume production
• Training for operators and maintenance staff
• Initial stock of spare parts
• Maintenance contract (annual fee, coverage hours, response conditions)
• Software development for host system integration (actuals collection, instruction receipt)
• Project management cost

Most of the variance between quotation totals comes not from the robot itself but from the presence or absence of these surrounding items. Safety compliance, host integration, training and spare parts are the ones most often missing from a cheap quotation. The reliable approach is for the buyer to build the comparison sheet, at a consistent level of granularity, and impose it on every bidder.

Benefit categories to include in the payback calculation

As discussed, a Thailand payback period built on labour savings alone tends to stretch. Accumulate the following benefit categories instead, each paired with a measurable indicator.

Note that the equipment availability line can only be booked as contribution margin where there is headroom to sell more output. Raising utilisation when orders are flat does not generate profit, so this figure has to be reconciled with the sales forecast. An investment plan that conflates the two will be picked apart in HQ review every time.

Staged investment as an option

Running visibility first, rather than committing to one large investment, has financial advantages as well as technical ones. The Phase 0 spend is comparatively small and the effect becomes visible within months, which builds an internal track record. Submitting the Phase 1 request on the back of that data lets the site put a numbers-based proposal to regional HQ instead of an intuition-based one.

It is also worth noting, as above, that in the Thai service robot segment monthly rental rates fell below THB 10,000 per unit and three-year lease cycles have become common. Procurement conditions for industrial robots are different, but the wider point stands: ownership is not the only procurement model worth examining. Actual terms vary by supplier and must be confirmed case by case.

Building BOI automation incentives into the investment plan

For automation investment in Thailand, the Board of Investment (BOI) promotion regime affects the decision directly. Because incentives change the effective cost of the project, they belong on the same table as cost and ROI. What follows is a summary of publicly available information, and applicability to any specific project must be confirmed with the BOI and a tax adviser. Announcements are revised, so working from the latest official text is a prerequisite. Note also that all of this is Thailand-specific and does not apply to plants in Vietnam, Indonesia or elsewhere in ASEAN.

The 2026 incentive: Announcement No. 4/2569 (automotive and HEV/PHEV)

According to commentary from the law firm Tilleke & Gibbins, BOI Announcement No. 4/2569 was published in the Royal Gazette on 31 March 2026. It is designed to raise production efficiency in the automotive industry by supporting automation and robot adoption.

(Source: Tilleke & Gibbins, “Thailand Unveils New Incentives for Automotive and HEV/PHEV Manufacturing”)

Four features stand out.

First, existing projects are eligible. Plants already in operation that make additional automation investments have a route in, which is not a trivial point in practice.

Second, the local content differential. Where 30% or more of the machinery value is linked to Thailand’s domestic automation machinery industry, the exemption rises from 50% to 100%. Whether that condition is considered during supplier selection materially changes the level of benefit. The EEC humanoid component investment mentioned earlier makes more sense read in this context.

Third, the THB 1 million minimum investment threshold. Because software, IT systems and data centre services can reportedly be counted towards it, the project can be structured differently from a pure hardware investment.

Fourth, the end-of-2027 application deadline. As noted, alignment with the corporate budget cycle has to be sorted out early.

Frameworks outside the automotive sector

BOI frameworks relevant to automation and robotics exist beyond automotive. Per published summaries, development of robotics and automation may fall under category 5.6 / group A1; standard BOI corporate income tax exemption runs to a maximum of eight years, and combining BOI with EEC promotion is reported to allow up to 15 years (with a 50% reduction after the exemption ends).

For existing BOI-promoted companies undertaking technology upgrading (Activity 10.1), an additional three-year corporate income tax exemption is available. On the AI side, a 200% deduction is allowed for AI workforce development expenses, and companies investing 1–3% of payroll in AI training may receive up to three additional years of exemption on a merit basis.

(Sources: Pertama Partners, Tilleke & Gibbins, BOI official materials)

One clarification: the 200% deduction refers to AI workforce development expenses. It should not be read as an equivalent measure for automation equipment itself. The scheme names are similar enough that this gets confused regularly.

Separately, Mordor Intelligence cites ASEAN Industry 4.0 subsidies supporting up to 70% of eligible automation spending as a regional growth driver. That is a collective description of schemes across several countries in the region, not a reference to any specific Thai programme. Country-level schemes must be checked individually.

Cautions when planning around incentives

BOI incentives help the business case, but plans come apart on the following points:

• Sequence of application and commencement: as a general rule, promoted investment is subject to timing requirements around application and approval. Placing the order first can make the investment ineligible
• Definition of investment value: exclusions such as land and working capital mean the internal capex figure and the BOI investment figure may not match
• Evidencing compliance: meeting conditions such as local content ratios requires contracts, invoices and origin documentation
• Deadline management: where a scheme has an application deadline, internal approval delays translate directly into lost incentives
• Reporting obligations: post-award reporting and audit effort should be budgeted as part of project cost

To repeat: this is a general summary of published information. Actual eligibility, procedures and required documents must be confirmed with the BOI and your tax or legal advisers before you decide.

Selecting a robot system integrator in Thailand and building the internal team

A technically sound plan goes nowhere without people to execute it. In a Thai automation project, partner selection and internal organisation matter as much as the capital figure.

Ten checkpoints for integrator selection

1. Track record in the same process: references for the same operation and workpiece type. Visit a live installation if possible
2. Number of engineers in Thailand: is sales local while engineering sits offshore?
3. Language capability: can training for Thai operators and maintenance staff be delivered in Thai?
4. Post-commissioning self-sufficiency: does the scope include training so your team can change teaching points and add variants?
5. Program ownership: do you own the robot and PLC programs, with the right to modify them?
6. Documentation: scope and language of circuit diagrams, I/O lists, program comments and operating procedures
7. Ownership of safety design: who performs the risk assessment and who carries responsibility?
8. Service terms: inspection frequency, target response time, parts stock, cost breakdown
9. Host integration experience: references for integrating with production management systems and IoT platforms, not just standalone machines
10. Project structure: is a dedicated project manager assigned, or is the PM juggling several jobs?

Items 4 and 5 have the greatest effect on long-term cost. If every changeover requires an external contractor, responsiveness to a widening product mix drops — and automation ends up reducing manufacturing flexibility rather than improving it.

Internal structure: decide who owns it

Most organisational stalling in automation projects comes down to unclear ownership. Manufacturing is fully occupied with today’s output, production engineering with equipment installation, IT with the core business systems — and nobody holds the cross-functional driving role.

A workable structure usually includes the following:

• Name one project owner: a part-time role is acceptable, but responsibility and authority must be written down
• Include shop-floor representatives: if the people who will operate the cell are absent from the design phase, the new way of working will not stick
• Involve maintenance early: maintenance lives with the equipment afterwards, and excluding them from specification decisions leaves you with hard-to-service assets
• Fix a single point of contact with regional HQ: repeated re-explanation to different stakeholders is what stretches approval timelines
• Assign an owner for the data: unless it is clear who reviews Phase 0 data and who reacts to an anomaly, the dashboard becomes wallpaper

The robotics talent shortage that Mordor Intelligence flags as a cause of commissioning delay is not purely an external labour-market issue. It is also a question of whether anyone inside the company is developing automation expertise. Rather than outsourcing the first project entirely, attaching your own junior engineers to it and accumulating experience is what determines the speed of the second and third.

If you operate multiple sites across ASEAN

Companies with plants in more than one country — Thailand and Vietnam, for instance — face the question of which site to automate first. The essential input is that conditions differ considerably between them.

For Vietnam, manufacturing labour costs are reported at roughly USD 2.99–4.10 per hour (broadly half of China’s), with labour costs up 11.2% between 2021 and 2023. More than 5,000 industrial robots were in operation as of 2023, up 40% on 2021, and the AI-in-manufacturing and robotics market is put at around USD 1.2 billion. At the same time, initial costs for an SME adopting AI and robotics are reported to exceed USD 500,000 in many cases, and 2026 has been characterised as a “smart factories or bust” decision point.

An important caveat here: the Thai policies and figures used in this article — the THB 400 minimum wage, BOI Announcement No. 4/2569 and EEC-related incentives — do not apply in Vietnam. Vietnamese incentive schemes must be verified with local authorities and advisers. Extrapolating policy across borders is a serious risk to any investment plan.

With that established, a practical prioritisation is to consider, in order: (1) the site where securing labour is hardest, (2) the site serving the most demanding quality customers, and (3) the site where current-state data already exists and benefits can therefore be measured. Sequencing that way improves the odds of a successful first project.

Frequently asked questions

How should we estimate factory automation cost for a plant in Thailand?

Because the total varies so much with process, unit count and specification, quoting a single benchmark would be misleading. The practical method is for the buyer to build a line-item list covering not just the robot but the gripper, fixtures, safety equipment, utility work, installation, teaching, training, spare parts, maintenance contract and host system integration — then request quotations from several suppliers at that same level of granularity. The cost chapter above contains the full checklist. Because BOI incentives can change the effective burden, run the tax review in parallel.

Should we start with industrial robots or IoT equipment monitoring?

In most cases we recommend leading with visibility through IoT equipment monitoring, for two reasons. First, deciding which process to robotise is impossible without measured data on downtime, defect rates and cycle time. Second, without a pre-implementation baseline you cannot prove the benefit numerically afterwards, which makes the next capital request much harder to pass. That said, where a process presents an urgent occupational safety issue, or where staffing is physically unobtainable, prioritising robotisation ahead of the data work can be the right call.

Can BOI automation incentives be used by non-automotive factories?

BOI Announcement No. 4/2569, published in the Royal Gazette on 31 March 2026, covers general motor vehicle manufacturing (category 3.6) and PHEV/HEV manufacturing (category 3.8). Outside automotive, development of robotics and automation may fall under category 5.6 / group A1, and technology upgrading by existing BOI companies (Activity 10.1) carries an additional three-year corporate income tax exemption. Which category your business actually falls into, and which incentives you can access, is a case-by-case determination that must be confirmed with the BOI and a tax adviser.

What should we check when choosing a robot system integrator in Thailand?

At minimum: references in the same process, whether engineers are resident in Thailand, whether shop-floor training can be delivered in Thai, whether the scope includes training so your team can change teaching points after commissioning, whether you own and may modify the robot and PLC programs, the scope of documentation, who is responsible for safety design, response times and parts stock for maintenance, experience integrating with host systems, and whether a dedicated project manager is assigned. “Can our own people touch it after handover?” is the item with the biggest long-term cost impact. The partner selection chapter above covers each point.

Will automation cause resistance or resignations on the shop floor?

How the objective is communicated changes the reception dramatically — “headcount reduction” lands very differently from “removing hazardous and physically punishing work and stabilising quality.” In practice it helps to present, at project kick-off, a plan for redeploying affected staff into higher-value processes or into equipment maintenance and data roles. Involving actual operators from the design stage also raises adoption and creates ownership. Equipment installed without consulting the floor carries a real risk of simply not being used.

Is IoT equipment monitoring possible in a plant full of legacy machines?

Yes. Even for machines without a PLC, non-invasive approaches are available: determining run/stop from the state of the stack light, measuring load with clamp-type current sensors, or adding a photoelectric sensor to count output. The important thing is not to insist on a single acquisition method across all equipment. Different machines can use different methods, provided everything is normalised into a common format at the upper layer — analysis and management still work.

When should we start considering AI-based predictive maintenance?

Realistically, once sensor data is accumulating continuously and failure and stoppage events are recorded and linked to it. Industrial AI media report that mature deployments of 12 months or more have seen unplanned downtime fall 31–47%, maintenance costs fall 12–24% and OEE improve by 8–15 percentage points — but these are vendor-published reported ranges, not TOMAS TECH results, and they are not a guarantee of outcome. The “mature operation” condition is the key detail: the plan has to allow time for data accumulation and for the new practice to become routine.

Should Chinese-made robots be on our shortlist?

More procurement options is, in itself, a favourable change for a buyer. The decision should rest not on nationality but on concrete items: service structure, spare-part supply, scope of documentation, program ownership, training programmes, references in comparable processes and conformity to safety standards. The evaluation table earlier in this article is designed to be dropped straight into a quotation comparison sheet. Restating everything on a total cost of ownership basis is the only way to judge a price difference properly.

Any tips for getting automation capex approved by regional HQ?

Three. First, build the benefit case from more than labour savings — include quality cost, indirect man-hours and opportunity cost. Second, present a measured pre-implementation baseline and define the benefit measurement method in advance. Third, schedule backwards from policy deadlines such as the BOI application cut-off, and state explicitly that internal approval delays translate into lost incentives. All three are far easier to prepare once Phase 0 visibility is complete.

Talk to us about automation in Thailand

TOMAS TECH is based in Bangkok and supports manufacturers across Thailand and ASEAN with the PEGASUS production management system, IoT equipment monitoring and AI adoption. As set out above, automation is a means rather than an end, and its return depends heavily on how early you can reach a state where the data actually exists. We are happy to start from the earliest questions: how much equipment data can realistically be captured today, and which process should be tackled first so that benefits can genuinely be measured. Questions of eligibility for policy incentives such as BOI require confirmation with specialists, and we can discuss how to run that process too. We do not promise specific outcomes; we do offer realistic options grounded in your current situation.

References

• IFR | Global robot demand in factories doubles over 10 years: https://ifr.org/ifr-press-releases/news/global-robot-demand-in-factories-doubles-over-10-years
• IFR | Robot density surges in Europe, Asia and the Americas: https://ifr.org/ifr-press-releases/news/robot-density-surges-in-europe-asia-and-americas
• Mordor Intelligence | Southeast Asia Industrial and Service Robot Market: https://www.mordorintelligence.com/industry-reports/southeast-asia-industrial-and-service-robot-market
• Tilleke & Gibbins | Thailand Unveils New Incentives for Automotive and HEV/PHEV Manufacturing: https://www.tilleke.com/insights/thailand-unveils-new-incentives-for-automotive-and-hev-phev-manufacturing/3/
• Pertama Partners | Thailand BOI Manufacturing: https://www.pertamapartners.com/funding/thailand-boi-manufacturing
• BOI | Thailand Investment Promotion Policy for Automation and Robotics Industries (PDF): https://www.boi.go.th/upload/content/BOI%20Thailand%20Investment%20Promotion%20Policy%20for%20Automation%20Robotics%20Industries%20Japanese_619b411183526.pdf
• JETRO | Business news: THB 400 minimum wage applied in Bangkok: https://www.jetro.go.jp/biznews/2025/07/b21007a1ac8f7fca.html
• Tokyo SME Support Center, Bangkok Office | Column on Thailand’s robot market: https://thai.tokyo-sme.com/business_info/column/2025/07/vol-30/
• Bangkok Shuho | Humanoid robot component production base in the EEC: https://bangkokshuho.com/thainews-818/
• f7i.ai | Industrial AI Statistics 2026: https://f7i.ai/blog/industrial-ai-statistics-2026-the-hard-data-behind-manufacturings-transformation
• ITPC Ho Chi Minh City | Smart factories or bust: decision time in 2026: https://itpc.hochiminhcity.gov.vn/web/en/-/smart-factories-or-bust-decision-time-in-2026
• Research and Markets | Vietnam AI in Manufacturing & Robotics Market: https://www.researchandmarkets.com/reports/6207519/vietnam-ai-in-manufacturing-robotics-market

Related reading

• Choosing a Production Management System for Factories in Thailand 2026
• AI Agents in Manufacturing 2026
• Logistics DX in Southeast Asia 2026: Warehouse Automation and WMS in Thailand
• AI-OCR for Back-Office Automation in Thai and ASEAN Factories 2026