Phased automation breaks a factory-wide ambition into testable investments. Instead of buying equipment first, management defines the business constraint, proves one representative process in a 90-day pilot, and uses FAT, SAT, and operational handover gates to decide whether to scale. This guide explains how manufacturers in Thailand can prioritize automation investments, issue a decision-ready RFP, and build reusable FA systems without turning a small start into a permanent prototype.
What you will learn
- How to build an automation roadmap around business evidence rather than a machine list
- How to prioritize safety, quality, capacity, labor, maintainability, and economics together
- How a 90-day pilot can support Go, Conditional Go, Repeat, or No-Go decisions
- What an FA-system RFP, FAT/SAT plan, backup package, and handover must contain
- How to read Thailand BOI information without assuming automatic eligibility
Unless stated otherwise, the 90-day structure, scoring weights, gates, targets, cycle times, and payback scenarios in this article are recommended project examples—not mandatory values imposed by law, a standard, or BOI. Confirm machine risks, customer requirements, applicable law, contracts, installed equipment, labor implications, tax treatment, and BOI eligibility for the actual project.
Why phased automation is a management discipline
An automation project can fail even when the robot and PLC run. If an engineer is required for every model change, upstream starvation remains unresolved, maintenance cannot recover the cell, quality evidence is missing, or benefits cannot be compared with a baseline, the investment has not created a sustainable production capability. A big-bang rollout repeats wrong assumptions across several lines and multiplies rework and downtime exposure.
Phased automation treats investment as an evidence chain:
- Define the business problem.
- Measure the current condition and identify the true constraint.
- Prove technology, safety, operations, and economics on one representative process.
- Convert the result into standards, training, maintenance, and backup.
- Use a gate to scale, correct, repeat, or stop.
Starting small is not the objective by itself. The objective is to create evidence that justifies—or prevents—the next commitment. A temporary demo that runs once does not resolve 24/7 production, product variation, recovery, night shift, maintenance, cybersecurity, or spares.
Start the automation roadmap with a business constraint
Fix the problem statement in one sentence
“Install a robot” is not a problem statement. A useful statement identifies the process, outcome, and constraint: “reduce night-shift manual lifting while stabilizing bottleneck capacity during the shipping peak,” or “reduce inspection variation while linking images and process conditions to the production lot.”
Measure a baseline for each candidate. Include enough operating patterns to represent model mix, lots, shifts, operators, machine state, material, rework, and downtime. Label estimates as estimates.
| Perspective | Example baseline | Evidence after automation |
|---|---|---|
| Safety | Exposure, lifting, contact, intervention frequency | Risk assessment, interlock tests, residual risk |
| Production | Cycle, stoppage, changeover, WIP | Timestamped events, good output, reason codes |
| Quality | Defects, escapes, reinspection, variation | MSA, boundary samples, false decisions, traceability |
| People | Work content, relief labor, night work, skill dependence | Standard work, role redesign, training, exception response |
| Maintenance | Failures, recovery, spares, external dependence | MTTR evidence, alarms, recovery drill, backup |
| Economics | Labor, loss, scrap, consumables, downtime | Incremental cash flow, sensitivity, actual variance |
Separate a bottleneck from a symptom
The busiest manual station is not automatically the best investment. If unstable supply upstream is the cause, a faster downstream machine only waits more efficiently. Map capacity, queues, stoppage causality, quality gates, hazardous work, and supply instability separately.
Use capital-investment planning to frame the decision envelope and production-line modification planning to account for legacy equipment, construction, interfaces, and shutdown windows.
A scorecard for automation investment priorities
Separate screening from detailed evaluation. First identify candidates that have an unacceptable safety condition, cannot meet a customer requirement, lack critical data, or have no process owner. Then compare viable candidates using the same dimensions.
| Dimension | Decision question | Example weight |
|---|---|---|
| Safety and ergonomics | Can the concept remove a hazard or high-load task by design? | 25 |
| Quality and traceability | Can it reduce variation, escapes, and unknown causes? | 20 |
| Capacity and delivery | Does it improve the true constraint and demand response? | 20 |
| Feasibility | Are the workpiece, space, data, utilities, and skills suitable? | 15 |
| Operations and maintenance | Can the site change over, recover, and maintain it? | 10 |
| Economics | Does the case survive several operating scenarios? | 10 |
These weights are examples. Safety is an independent gate; it must not be offset by a high economic score. If a 1–5 scale is used, keep the data source, owner, confidence, and open question behind every score.
Value divided by difficulty is not enough
A high-value, easy project may teach little about wider rollout. A first pilot should combine:
- Value: safety, quality, capacity, delivery, or maintenance outcome
- Representativeness: interfaces, product behavior, and exceptions shared by other lines
- Reversibility: a controlled path back to manual production if the pilot fails
- Observability: data to compare baseline and result
- Ownership: named process, maintenance, quality, and IT/OT owners
A short-looking payback can omit fixtures, changeovers, spares, software upgrades, networking, training, support, and shutdown work. Show minimum, base, and adverse scenarios rather than one precise number.
Four waves of phased factory automation
Wave 0: Measure and stabilize
Before adding equipment, stabilize standard work, quality conditions, reason codes, product masters, and equipment identities. Sensors and basic data collection may reveal that a fixture or method change should precede automation. Automating an unstable process simply moves variation faster.
Outputs include the baseline, process FMEA or risk assessment, a data dictionary, constraints, and the benefit calculation. “Do not automate yet” is a valid gate result.
Wave 1: A 90-day single-process pilot
Use one representative process. Test model changes, abnormal workpieces, sensor failure, communication loss, emergency stop, restart, and manual recovery—not only the normal cycle. Temporary equipment still needs an approved risk boundary.
Wave 2: Integrate the cell
Connect material flow, inspection, MES, quality records, and maintenance notification. A fast machine does not improve the line if starvation, blocking, re-entry, or buffer logic remains unresolved. Specify ownership and every exception crossing a system boundary.
Wave 3: Standardize and scale
Package control templates, HMI conventions, alarms, safety documentation, network patterns, backups, bills of material, FAT/SAT, and training. Reassess the risk and site differences before copying. Standardization means reusing decisions, verification, and maintenance—not merely buying the same model.
How to run a 90-day automation pilot
Days 1–15: Scope and baseline
- Fix the process, representative products, exclusions, and owner.
- Measure safety, quality, capacity, stoppage, and work content.
- Inspect real tolerances, warp, oil, dirt, orientation, and mixed parts.
- Approve OT access, data ownership, change authority, and shutdown windows.
- Agree Go, Conditional Go, Repeat, and No-Go conditions before building.
Days 16–35: Minimum cell and exceptions
Build the minimum robot, fixture, sensor, PLC/HMI, inspection, and record chain. Write the exception list before polishing the normal sequence: double pick, no part, jam, wrong model, re-entry, safety-door interruption, power return, network loss, and master mismatch. Assign safe recovery roles.
Days 36–60: Performance, safety, and quality proof
Choose a continuous-run duration and sample volume that fit the process. Measure abnormal detection, false stops, recovery, changeover, and good output as well as normal success. Vision testing should include representative defects, boundary samples, illumination variation, contamination, and camera movement. Robot testing should include reach, grip, drop, interference, manual operation, and teaching changes.
Days 61–75: Maintenance, cybersecurity, and recovery
Ask site maintenance to diagnose from approved drawings, alarms, I/O, and backups. Test named accounts, approved remote access, allowed communication, a backup before and after change, and restoration in a separate environment.
Days 76–90: Investment gate
Compare actual results with the baseline and update unresolved risk, engineering work, production design, shutdown needs, training, SLA, spares, and multi-year cost.
| Decision | Meaning | Next step |
|---|---|---|
| Go | Acceptance evidence is sufficient and residual risk is approved | Production design and Wave 2 |
| Conditional Go | Value is plausible with bounded open items | Correct with owner and deadline, then re-gate |
| Repeat | The hypothesis remains but evidence is insufficient | Narrow and rerun the proof |
| No-Go | Technology, operation, or economics is not viable | Stop, retain evidence, choose an alternative |
The purpose is not to complete every site in 90 days. It is to earn the next investment decision.
What an FA-system RFP must specify
A model-number list encourages price comparison while leaving risk unpriced. Write requirements as scenarios, constraints, interfaces, exceptions, evidence, and responsibility boundaries.
Functional requirements
- Identify product, lot, work order, and equipment execution.
- Define transitions among automatic, semi-automatic, manual, and maintenance modes.
- Detect, stop, reject, re-enter, and recover abnormalities with traceable events.
- Manage recipe and PLC/HMI/robot-program revisions and change history.
- Link inspection decisions, images, and measurements to production evidence.
- Continue safely or stop during upstream-system loss, then synchronize without logical duplicates.
Non-functional requirements
- Responsibility for risk assessment, safety functions, validation, and residual risk
- Test conditions for cycle, changeover, availability, recovery, and performance
- Network segmentation, accounts, logging, remote service, and vulnerability handling
- Source, license, credential handover, backup, and restore procedures
- Recommended spares, obsolescence, support hours, local response, and change rates
- Thai/English interfaces, training, operating standards, and maintenance standards
Require suppliers to classify every line as Standard, Configuration, Custom, Third-party, or Out of scope, with assumptions and customer-supplied items. Combine the evaluation method in automation-equipment vendor selection with small-manufacturer automation examples and compare process fit, local maintenance, and acceptance evidence—not supplier size alone.
Keep machine safety as a separate investment gate
ISO 12100:2010 addresses principles for machinery risk assessment and risk reduction. ISO 13849-1:2023 addresses the methodology and requirements for designing and integrating safety-related parts of control systems. For robotics, ISO 10218-1:2025 covers industrial robots, while ISO 10218-2:2025 covers industrial robot applications and cells.
Listing a standard in the RFP does not make a cell safe. Assess the application, hazards, people, material, speed, force, maintenance, and foreseeable misuse. Translate that assessment into inherently safer design, safeguarding, information, verification, and residual-risk control. Confirm applicable standards, Thai law, customer rules, and any third-party assessment with competent specialists.
Repeat the risk assessment when a wave changes. Conveyor interaction, coordinated startup, buffer access, remote control, and restart hazards may exist only after integration.
Turn FAT and SAT into investment evidence
FAT tests controlled supplier conditions; SAT tests the actual site’s power, network, product, people, environment, and upstream/downstream flow. A pilot pass is not a substitute for production acceptance.
| Test | FAT focus | SAT focus | Evidence |
|---|---|---|---|
| Safety | Logic, fault insertion, stop behavior | Real layout, access, intrusion, restart | Risk record and test results |
| Capacity | Agreed products and continuous run | Actual line constraint and interfaces | Timestamped output and reason codes |
| Quality | Good, defective, and boundary samples | Real material, light, humidity, contamination | MSA and decision history |
| Exceptions | Sensor, grip, jam, communications | Site recovery roles and duration | Alarm and recovery record |
| Change | Revision update and rollback | Approval workflow and actual backup | Version, hash, change record |
| Maintenance | Replacement, diagnosis, drawings | Site technician demonstration | Skill check and work standard |
| OT | Roles, logs, allowed flows | Real network and remote service | Configuration and audit logs |
Define the method, population, and exclusions. If a specification says “cycle time at or below 15 seconds,” define start and end events, model, sample, micro-stops, and changeover treatment. Fifteen seconds here is only an illustration—not a recommended target.
Make handover an internal capability
Automation creates lasting value when the site can produce, recover, and change safely without permanent integrator presence. Do not accept “manuals included” as the handover definition.
- Approved electrical, pneumatic, mechanical, network, I/O, and BOM revisions
- Source and executable versions for PLC, HMI, robot, vision, and gateway
- Recipes, user model, licenses, certificates, and time configuration
- Safety programs, validation evidence, residual risks, and change restrictions
- Alarm cause, diagnosis, recovery, and escalation guides
- Preventive maintenance, calibration, consumables, spares, and obsolescence data
- Backup, restore, replacement startup, and rollback procedures
- Role-based training and demonstrated competence
Verify training by demonstration: safe restart for operators, model change for setup staff, I/O diagnosis and restore for maintenance, and account disablement and change approval for administrators. Define what a support SLA restores, not only its response time.
Design OT security and backup in Wave 1
NIST SP 800-82 Rev. 3 provides OT-security guidance that considers performance, reliability, and safety. The ISA/IEC 62443 series offers a lifecycle and shared-responsibility framework for asset owners, suppliers, integrators, and service providers.
Do not let pilot exceptions become the production architecture: flat-network connections, shared administrator accounts, or permanently open remote desktop. Start asset inventory, allowed communications, named accounts, logs, external access, patch decisions, and safe incident states in Wave 1.
NIST SP 1339, OT Backup Quick Start Guide, published in June 2026, emphasizes integrating backup into change management, creating and testing backups regularly, and reviewing them in recovery exercises. A file is not proof of recovery. Include PLC/HMI/robot source, vision models, recipes, gateways, networks, licenses, certificates, dependencies, and recovery order.
Using Thailand BOI information responsibly
Thailand BOI’s public Smart and Sustainable Industry upgrade measure states a minimum efficiency-enhancement investment of THB 1 million, excluding land and working capital. The page describes machinery import-duty exemption and, for existing projects under stated conditions, a three-year corporate-income-tax exemption ordinarily capped at 50% of the efficiency-improvement investment. It also describes a 100% cap where at least 30% of the total value of machinery, automation systems, or robotics used or upgraded is linked to Thailand’s domestic automation industry.
These statements do not mean that any automation project automatically qualifies. Confirm the promoted activity, timing, existing/new-project treatment, eligible cost, domestic-link evidence, certificate requirements, and completion period directly with BOI or qualified advisers. Keep a business case without incentives and a separate incentive-qualified scenario.
BOI’s first-half 2026 release reported 1,299 investment applications totaling approximately THB 1.47 trillion. Machinery, automation, and robotics accounted for 82 applications and approximately THB 13.1 billion. These are applications, not proof of approval, realized investment, or the ROI of a specific factory.
Build the economics beyond “headcount saved”
Benefit categories include hazardous-work reduction, bottleneck output, avoided overtime or outsourcing, defects and investigation, changeover and recovery, traceability, and resilience to recruitment or skills shortages. Cost categories include machines, fixtures, safety, inspection, controls, network, engineering, SI, construction, shutdown, validation, licenses, support, spares, calibration, training, change, and ramp-up loss.
Labor capacity does not always become immediate payroll reduction. It may support growth, vacancy coverage, safer work, or redeployment. Show demand reduction, late ramp-up, performance shortfall, and higher support-cost sensitivities. Update the assumptions with measured evidence at every gate.
Common failure modes
Buying the cheapest machine before defining the portfolio
Controls, HMIs, data, parts, and support fragment. Define minimum standards in Wave 0 and approve exceptions.
Passing the pilot on the normal cycle
Production loses time in exceptions, changeover, re-entry, cleaning, and maintenance. Include fault insertion and recovery.
Using only labor reduction as the KPI
Measure safety, quality, constraint capacity, recovery, and changeability. Design the work that remains.
Accepting supplier lock-in
Agree source, versions, backups, licenses, credentials, data formats, change rates, and exit handover in the RFP.
Leaving prototype shortcuts in production
Resolve shared accounts, temporary wiring, unapproved remote access, and manual data correction before scale—or record a Conditional Go.
Copying a safe cell without reassessing risk
Layout, product, people, speed, and interfaces change hazards. Reuse documents, but repeat differential risk assessment and SAT.
FAQ: phased automation and the automation roadmap
What is phased automation?
It divides investment into measurement and stabilization, representative-process proof, cell integration, and standard rollout. Each wave produces evidence on safety, quality, capacity, operations, and economics before more capital is committed.
How far ahead should an automation roadmap look?
A practical structure uses a strategic direction aligned with equipment life, a candidate portfolio for roughly the next 12 months, and a detailed next 90-day proof. Those periods are examples and should follow the business-planning cycle.
How should automation investments be prioritized?
Evaluate safety and ergonomics, quality, the true constraint, feasibility, operations, maintenance, and economics. Add representativeness, reversibility, observability, and ownership. Keep safety as a non-offsettable gate.
Which process is best for a small-start automation pilot?
One with meaningful value, representative conditions, a safe fallback, and measurable baseline and results. It is not necessarily the most visible process or the one with the most people.
What should a 90-day pilot prove?
Normal performance, product variation, exceptions, recovery, changeover, machine safety, quality, OT connectivity, maintenance, backup, training, and the investment assumptions.
When should an RFP be issued?
After the baseline, scope, representative products, constraints, acceptance evidence, and responsibility boundary are defined. If uncertainty remains high, separate concept/proof services from production procurement.
What is the difference between FAT and SAT?
FAT verifies design and exceptions in controlled supplier conditions. SAT verifies the actual product, utilities, network, operators, environment, and line interfaces. Both retain evidence and open items.
Can BOI incentives be included in payback?
They can be a separate scenario, but never an automatic assumption. Reconfirm the latest measure, activity, application criteria, eligible costs, domestic link, and deadlines with BOI or qualified advisers.
What must be received at FA-system handover?
Approved drawings, BOM, source and executable versions, licenses, account handover, revisions, backup and restore, safety validation, maintenance and alarm standards, spares, training evidence, SLA, and change procedures.
Conclusion
Phased automation does more than reduce the size of the first purchase. It reduces uncertainty in testable units and directs capital using evidence. Fix the business constraint and baseline before equipment selection. Compare candidates using safety, quality, capacity, operations, and economics. In the 90-day pilot, test exceptions, recovery, maintenance, security, and handover as seriously as the normal cycle. Convert the result into FAT/SAT evidence and reusable standards before the next rollout.
If your team is still shaping the automation roadmap, investment priorities, RFP, or 90-day acceptance plan, TOMAS TECH can support site assessment, FA concept design, PLC/robot/inspection integration, MES/OT connectivity, FAT/SAT, and operational handover. Contact us.
