Short answer
Buy SaaS for work you do the same way as everyone else, such as accounting, email and payroll; build custom software for the process that sets your business apart. AI made small tools cheaper to build but not cheaper to own, and over five years cost follows headcount: a subscription is usually cheaper for a small, stable team, while per-user pricing can overtake building for a growing one. Most Thai businesses end up using both.
Should you buy SaaS or build custom software?
Buy SaaS for work you do the same way as everyone else: accounting, email, payroll, chat. Build custom software for the process that sets your business apart, or where every product on the market forces you to work around it. That rule has not changed.
What AI changed is where the line sits. Small internal tools, dashboards and automations that once needed a project now take days, so more of them are worth building. What AI did not change is the cost of owning software for years: someone still has to maintain it, secure it and answer the phone when it breaks.
Cost follows headcount. For a small, stable team a subscription is usually cheaper over five years; for a team that is growing, per-user pricing can overtake the cost of building. Most Thai businesses end up using both: SaaS for the core, and custom pieces built around it.
What changed in 2025 and 2026?
Building got cheaper, especially for small tools. In Retool's 2026 build-versus-buy survey of 817 builders, 35% said they had already replaced at least one SaaS tool with something they built, and 78% expected to build more in 2026. The categories replaced most were workflow automations, internal admin tools and reporting dashboards. The sample is mostly Retool customers, people already inclined to build, so read it as a direction rather than a market share.
Investors took the threat seriously. In early February 2026, after Anthropic launched Claude Cowork, software stocks fell sharply in a selloff the financial press called the SaaSpocalypse. The fear was specific: if AI agents do work that people used to do inside a SaaS product, fewer people need a seat, and per-seat pricing is how most SaaS companies earn their money.
Vendors had seen it coming. Salesforce launched its Agentforce AI agents in 2024 priced per conversation rather than per seat, and more products now charge for AI features by usage on top of the subscription, which makes a SaaS bill harder to budget than it used to be.
What did the headlines get wrong?
The best-known example of a company dropping SaaS for AI turned out to be something else. In December 2024, Klarna said it was shutting down Salesforce and Workday as part of an AI overhaul. On closer reporting, HR moved to Deel, another SaaS product, and the CRM work was spread across other SaaS tools plus some in-house systems. Klarna consolidated its software; it did not replace it with AI.
The speed-up from AI coding is also smaller and less certain than the marketing suggests. In METR's randomised study of experienced open-source developers in early 2025, tasks took 19% longer with AI tools, even though the developers believed AI had made them 20% faster. METR's 2026 follow-up with newer tools showed a much smaller effect, a 4% slowdown with a wide margin of error, and METR itself now says developers are probably sped up today but that its data cannot yet measure by how much.
The honest reading: AI makes writing code faster for many tasks, but writing code was never most of the cost of a business system. Working out what the system should do, migrating data, testing, and running it for years still take the time they always took.
What does SaaS give you, and what does it really cost?
The benefits are real. You can start today, the vendor handles hosting, security patches, backups and uptime, new features arrive without you paying to build them, and the product has already been tested by thousands of other customers. For standard work, that is hard to beat at any price.
The subscription price is only the visible part of the cost. Budget for these as well:
- Every new hire adds a seat, so the bill grows with headcount rather than with value
- Annual price rises, and add-on charges for AI features priced by usage
- Higher tiers you are pushed into for one feature you need, such as an API or audit log
- Setup, data migration and training, often through a paid implementation partner
- Integration with your other systems, which the subscription rarely includes
- Staff time working around what the product cannot do
- Exchange-rate swings when the price is in US dollars
- The cost of leaving: exporting data and retraining people if the vendor raises prices or shuts down
What does custom software give you, and what does it really cost?
The benefits are just as real. The system fits your process exactly, the cost does not rise with every hire, you own the code and decide where the data lives, and you can connect it to anything. When the process is what sets your business apart, that fit is the whole point.
The build quote is also only the visible part. Over the life of the system, budget for these as well:
- Requirements and design work before any code is written
- Hosting, monitoring and backups every month
- Maintenance and security updates, commonly 15% to 20% of the build cost each year
- Every new feature, which you pay for rather than receiving in a vendor update
- Keeping up with changes in Thai tax rules and in the services it connects to
- Documentation and handover, so the system survives a change of developer
- The time of whoever in your company owns the system and makes decisions about it
Over five years, which costs less: SaaS or custom software?
It depends on headcount: in this example, ten users pay less on SaaS, while a team growing from 20 to 40 users pays less for a custom system from year 3. The figures below are an illustration built on stated assumptions, not a quote from any vendor, so replace them with your own. SaaS: USD 50 per user per month, 5% price rise each year, 35 baht to the dollar, and 100,000 baht for setup and migration; the growing team has 20 users in year one, 30 in year two and 40 from year three. Custom: a mid-size system at 1.2 million baht, 20% of that a year for maintenance, and 5,000 baht a month for hosting.
The result turns on headcount. With ten users, SaaS costs about half as much as building. With a team growing from 20 to 40 users, the subscription ends up costing about 1.4 million baht more than the custom system. In both cases the custom figure excludes new features, and the SaaS figure excludes workarounds and add-ons, so check which of those your situation actually carries.
- SaaS, 10 users
- SaaS, 20 growing to 40 users
- Custom software
| SaaS, 10 users | SaaS, 20 growing to 40 users | Custom software | |
|---|---|---|---|
| Year 1 | About 310,000 baht | About 520,000 baht | About 1,500,000 baht |
| Each later year | About 220,000 to 260,000 baht | About 660,000 to 1,020,000 baht | About 300,000 baht |
| Five-year total | About 1.3 million baht | About 4.1 million baht | About 2.7 million baht |
| What drives it | Seat price and yearly rises | Headcount growth | Build cost and maintenance |
What is different for a Thai business?
Foreign SaaS is usually priced in US dollars per user, so the baht cost moves with the exchange rate and grows with every hire. For a team that is growing quickly, that per-seat line is often the number that tips the decision toward building, or toward a product priced per company rather than per person.
Data location matters under PDPA. Sending personal data to servers abroad is allowed only under the conditions the law sets for cross-border transfers, so check where a SaaS product stores your customers' data before you sign, particularly for health, financial or HR records.
Thai tax and Thai habits are where general-purpose products fit worst: VAT and tax invoice formats, withholding tax certificates, e-Tax Invoice, and customers who expect to reach you on LINE rather than by email. Many Thai SaaS products already handle the tax side well. Where none handles your exact combination, that gap is usually the part worth building.
Every trade-off side by side
Read the rows that matter most to your business first; no option wins them all. The third column is where most businesses end up, and AI made it stronger, because the small custom pieces that connect SaaS products and automate the work between them are exactly what got cheaper to build.
| SaaS subscription | Custom software | SaaS core, custom around it | |
|---|---|---|---|
| Time to start | Today | Weeks to months | Days for SaaS, weeks for the custom parts |
| Up-front cost | Low: setup and migration | High: the build | Low to medium |
| Ongoing cost | Per seat, every month | Hosting and maintenance | Subscription plus upkeep of the custom parts |
| How cost grows | With every user you add | With every feature you add | Mostly with users, less than full SaaS if seats are trimmed |
| Budget predictability | Price rises and AI usage fees can surprise you | Predictable once built, but change requests add up | Two budgets to watch |
| Fits how you work | As far as the vendor allows | Completely | Standard where standard is fine, custom where it matters |
| New features | Arrive with vendor updates, whether you need them or not | Only what you pay to build | Vendor updates for the core, your own where they matter |
| Ownership and control | You rent access; the vendor decides the roadmap | You own the code and the roadmap | You own the parts that differentiate you |
| Data location and PDPA | Wherever the vendor hosts it | Wherever you choose, including Thailand | Choose a core that stores data where you need it |
| Thai tax and LINE | Depends on the product | Built to fit | Choose a Thai-capable core, build the gaps |
| Security and uptime | Vendor's responsibility, shared across customers | Yours, or whoever you hire | Vendor for the core, you for the custom parts |
| Integration | Only what the vendor's API allows | Anything you build | Built around the vendors' APIs |
| AI features | Vendor's AI, limited to data inside the product | Your own AI, across all your data | Vendor AI plus your own agents across products |
| Leaving later | Depends on how completely you can export your data | Easy if you own the code and it is documented | Swap the core, keep your custom parts |
| Main risk | Price rises, lock-in, data abroad | Nobody left who can maintain it | Integrations breaking when a vendor changes its API |
If you build or combine, where does SyncEdge fit?
We can take on any part of it: scoping, building, connecting your system to the SaaS products you already use, and supporting it after launch. Whether to buy or build is a separate decision, and the next section sets out how to make it. We sit on both sides: we build custom software, and we run a SaaS product of our own, SyncEdge Clinic, so we have no stake in one answer.
If you have developers with time to spare who have built this kind of system before, doing it in-house is a good route. The table shows where the two approaches differ.
| You do it in-house | With SyncEdge | |
|---|---|---|
| Scoping | Your team sets it and can change it as it goes, which also makes it easy to grow | A scoping phase first, producing the full scope, timeline and cost before development starts |
| Price certainty | The cost is salaries and time spent, known for certain once the work is done | One fixed price agreed before development, no hourly billing; any change of scope is quoted separately and approved in writing first |
| Who builds it | Your developers, if they have the time and have done this kind of work | Our team, with a demo of working software every week instead of a status report |
| Integrations with your SaaS | Your team learns each product's API | We connect your system to the SaaS, accounting or ERP products you use, through their APIs |
| Code and data ownership | Yours from the start | All code, documentation and IP are yours on final payment; data hosted in Thailand |
| Maintenance after launch | The same team carries on; if the people who built it leave, the knowledge can leave with them | Handover documentation your own team can maintain it from, or an optional monthly retainer covering critical fixes and security patches |
When should you buy, and when should you build?
Buy when the work is standard, the vendor is established, the per-seat cost stays reasonable at the headcount you expect in three years, and you can export all your data if you leave.
Build when the process itself is your advantage, when no product fits without costly workarounds, when per-seat pricing grows faster than the value you get, or when data has to stay on infrastructure you control. Build small first: one workflow, one dashboard, one integration, and only grow it once it earns its keep.
Combine the two when a good SaaS product covers most of the job. Connect it to your other systems, automate the steps people currently re-key by hand, and add an AI agent that works on your data through the products' APIs. That usually gets most of the benefit of custom software at a fraction of the ownership cost.
Before you buy or build
- Is this process what makes your business different, or do you run it like everyone else?
- What will the subscription cost at the headcount you expect in three years, in baht?
- Have you added setup, integration, add-ons and price rises to the SaaS figure?
- Have you added maintenance, hosting and future features to the custom figure?
- Can you export all your data, in a usable format, if you leave the vendor?
- Where is the data stored, and is that allowed for this data under PDPA?
- If you build, who maintains it after launch, and could someone else take over?
- Could a small custom piece around an existing product solve the problem instead?
Frequently asked questions
Is SaaS dead because of AI?
No. AI put pressure on per-seat pricing and made small internal tools cheaper to build, which is why software stocks fell in early 2026. But companies still buy SaaS for standard work like accounting, email and payroll, because running that software themselves costs more than subscribing. Even Klarna's widely reported move away from Salesforce and Workday mostly meant switching to other SaaS products.
Can AI build our business software cheaply now?
It can make the coding part faster, especially for small tools, dashboards and automations. It does not remove the work of deciding what the system should do, migrating data, testing, and maintaining it for years, which is where most of the cost of a business system sits. Treat AI as a cheaper way to build, not a free one.
Is custom software more expensive than SaaS?
At the start, almost always, because you pay to build before you use it. Over five years it depends mostly on headcount. In our worked example, ten SaaS users cost about 1.3 million baht against 2.7 million for a custom system, but a team growing to 40 users pays about 4.1 million for the subscription. Price both over the same period with your own figures.
How much does custom software cost to maintain each year?
A common planning figure is 15% to 20% of the build cost each year, covering security updates, bug fixes, small changes and keeping up with the services it connects to. Hosting comes on top, and new features are priced separately. Agree the maintenance scope in writing before the build starts.
Should we start with SaaS and build later?
Often, yes. Starting with a subscription shows you what you actually need before you pay to build it. The condition is being able to leave: make sure you can export all your data in a usable format, so moving later is a migration rather than a rescue.
What is the biggest risk of building our own software?
That nobody can maintain it later. The classic case is the tool only one employee understood; with AI-generated code it can be a tool no employee understands. Make sure the code is readable, tested and documented, and that you own it outright, before you depend on it.