Every few months a founder sends us the same question in a new form: "AI can write software now, so why are we still paying a SaaS vendor every month for every user?" It is a fair question. Subscription bills keep rising, and many companies have spent five years paying for software they will never own.
The honest answer depends on two numbers: how many people use the system, and how much of it is truly specific to your business. This article puts real figures on both. It uses public list prices, published implementation ranges, current developer rates, and what controlled studies actually measured about AI coding. It compares three options over three and five years, for a 10-user, a 50-user and a 150-user company. There is also a calculator so you can run your own numbers.
The short answer
- Under about 50 users, buy the ERP. A subscription ERP is cheaper over five years than building one, even with AI. It is two to three times cheaper than a US-built custom system.
- Above about 50 users, per-seat pricing starts to hurt. A custom build can match or beat premium SaaS on paper, especially with an offshore team. The risk moves from the invoice to delivery.
- AI saves less than you think on a full system (about 10 to 20 percent of effort). It saves more than you think on small, focused apps.
- The best value for most growing companies is a hybrid: a standard ERP core plus custom apps that replace niche per-seat SaaS tools.
What AI has really changed in software cost
Claims that AI makes software "ten times cheaper" come from demos. Controlled studies tell a more careful story:
| Study | What it measured | Result |
|---|---|---|
| Google randomized controlled trial (2024) | Engineers doing a complex enterprise coding task with AI | About 21% faster |
| METR randomized controlled trial (2025) | Experienced open-source developers working on their own large codebases | About 19% slower, though they believed they were faster |
| Thoughtworks field reports | Delivery teams using AI assistants | Coding about 30% faster, total cycle time 10 to 15% better |
The gap between "coding 30 percent faster" and "project 10 to 15 percent faster" matters. Writing code is roughly half of a software project. The rest is understanding the process, agreeing scope, designing data, testing, migrating data, training people and supporting them after go-live. AI helps a little with some of that and not at all with the rest. We use a 15 percent saving on total build effort as the default below. The calculator lets you test 0 to 40 percent.
Where AI does change the economics is small software. A customer portal, an approval workflow, a pricing tool or a management dashboard used to be a two-month project. With a modern stack and AI-assisted development, it is now often a two-to-four-week project. That is why the middle path later in this article is where the real savings are.
The assumptions behind the numbers
We used public list prices (US, billed yearly, checked November 2026). For costs vendors do not publish, we used the middle of ranges that implementers report:
| Input | Value used | Basis |
|---|---|---|
| Odoo (Custom plan) | $61 per user per month | Odoo pricing |
| Microsoft Dynamics 365 Business Central | $80 Essentials, $110 Premium per user per month | Microsoft pricing |
| Oracle NetSuite | About $999 per month base + about $150 per user per month | No public price list; reported range $129 to $199 per user |
| Implementation (10 / 50 / 150 users) | $30,000 / $80,000 / $200,000 | Reported ranges of $15,000 to $55,000 for small and $50,000 to $250,000+ for mid-market projects |
| SaaS price rises | 5% a year | Typical renewal increases of 5 to 10% |
| Partner support after go-live | 10% of implementation per year | Admin, small changes, version upgrades |
| Custom build effort (10 / 50 / 150 users) | 1,500 / 3,500 / 7,000 hours before AI | Lean, standard and broad ERP scope |
| AI saving | 15% of total effort | See the studies above |
| Developer rates | $125/hour US agency, $40/hour offshore | Reported market bands: US $100 to $150, offshore $25 to $50 |
| Custom maintenance | 18% of build cost per year, from year 2 | Industry baseline 15 to 25% |
| Hosting for custom | $100 + $10 per user per month | Cloud servers, backups, monitoring |
Two things are deliberately left out of both sides. One is your own staff's time on the project, which is similar either way. The other is Kuwait or US regional price differences, covered in their own sections below. Change any assumption in the calculator.
3-year and 5-year cost by company size
All figures are total cost of ownership in US dollars, rounded to the nearest thousand.
Small company: 10 users, lean scope
| Option | Year-1 licence or build | 3-year total | 5-year total |
|---|---|---|---|
| Odoo (Custom plan) | $7k licence | $62k | $85k |
| Business Central Essentials | $10k licence | $69k | $98k |
| NetSuite | $30k licence | $134k | $211k |
| Custom build, offshore team ($40/h) | $51k build | $77k | $100k |
| Custom build, US agency ($125/h) | $159k build | $224k | $286k |
At 10 users there is no contest for a full ERP. The licence is small, so most of the subscription cost is implementation, and a custom build has to pay for that same work and more. Even the cheapest custom option only ties Business Central after five years, and you would carry all the delivery risk to get there.
Mid-sized company: 50 users, standard scope
| Option | Year-1 licence or build | 3-year total | 5-year total |
|---|---|---|---|
| Odoo (Custom plan) | $37k licence | $219k | $322k |
| Business Central Essentials | $48k licence | $255k | $385k |
| NetSuite | $102k licence | $426k | $684k |
| Custom build, offshore team ($40/h) | $119k build | $183k | $241k |
| Custom build, US agency ($125/h) | $372k build | $527k | $676k |
At 50 users things change. Licences now grow with every hire, while custom cost grows with scope, not headcount. On paper an offshore custom build is the cheapest option over five years, about 25 percent below Odoo and 65 percent below NetSuite. A US-built custom system costs about the same as NetSuite.
"On paper" is doing a lot of work in that sentence. A 3,500-hour build that runs 30 percent over, which is common, adds about $36,000. You also inherit everything a vendor used to handle for you: accounting rules, payroll and tax changes, security patches, backups, and the risk that the two people who understand the code leave.
Larger company: 150 users, broad scope
| Option | Year-1 licence or build | 3-year total | 5-year total |
|---|---|---|---|
| Odoo (Custom plan) | $110k licence | $606k | $907k |
| Business Central Premium | $198k licence | $884k | $1.39M |
| NetSuite | $282k licence | $1.15M | $1.86M |
| Custom build, offshore team ($40/h) | $238k build | $381k | $505k |
| Custom build, US agency ($125/h) | $744k build | $1.07M | $1.38M |
At 150 users, per-seat pricing is the largest number on the page. NetSuite's licences alone pass a million dollars within four years. Here a custom or heavily owned system becomes a real strategic option, not just a cost exercise. That is why many larger companies run a custom core around the processes that differentiate them. It is also why open-core ERPs, whose per-user price is lower, are winning mid-market deals from the premium suites.
Try your own numbers
The calculator uses the same formulas as the tables. Leave add-on tools at none to reproduce the tables above. Choose $30 to reproduce the middle-path table below. Switch to KWD for Kuwaiti dinar totals.
Interactive calculator
SaaS vs custom software cost calculator
Change the inputs to see a 3- and 5-year total cost of ownership. Figures are in US dollars at list prices and market rates, not a quote.
3 Years
- SaaS ERP$219,000
Implementation, licences and add-ons rising 5% a year, partner support
- Fully custom build$183,000
Build, 18% yearly upkeep from year 2, hosting · Lowest on these inputs
- SaaS ERP + custom apps$219,000
Same ERP; small custom apps replace the add-on tools
5 Years
- SaaS ERP$322,000
Implementation, licences and add-ons rising 5% a year, partner support
- Fully custom build$241,000
Build, 18% yearly upkeep from year 2, hosting · Lowest on these inputs
- SaaS ERP + custom apps$322,000
Same ERP; small custom apps replace the add-on tools
Custom build cost after AI saving: $119,000. SaaS licences in year 1: $37,000.
Custom estimates overrun more often than subscriptions do. Treat a custom total within 20% of the SaaS total as a tie.
The middle path: keep the ERP, replace the SaaS add-ons
Most companies do not run only an ERP. They pay for the ERP and then for three or four per-seat tools around it: a customer portal, a field service app, a CPQ or pricing tool, an HR or approvals app, a reporting layer. These typically add $20 to $40 per user per month.
These tools are where AI-assisted custom development is strongest. They are small, specific to your process, and easy to connect to the ERP through its API. They also avoid the compliance-heavy parts of the system. Here is the same ERP (Odoo, as the lower-cost core) with $30 per user per month of add-on SaaS, compared with replacing those add-ons with custom apps (250 / 600 / 1,200 hours before AI):
| Company size | ERP + add-on SaaS, 5 years | ERP + custom apps, offshore | ERP + custom apps, US agency |
|---|---|---|---|
| 10 users | $105k | $100k | $131k |
| 50 users | $422k | $357k (−15%) | $432k |
| 150 users | $1.21M | $977k (−19%) | $1.13M |
This is the build-versus-buy answer for most growing companies. Buy the commodity core and build the edge. You keep vendor-maintained accounting, payroll and compliance. You stop paying per seat for workflows you could own. And the apps you build fit the way your people actually work. Retool's 2026 build-versus-buy research reports that 35 percent of enterprises have already replaced at least one SaaS tool with software they built themselves.
Hidden costs on both sides
SaaS costs that rarely make it into the business case
- Seat creep. Read-only users, seasonal staff and managers who log in twice a month still need licences on many plans.
- Renewal increases of 5 to 10 percent a year compound. A $50,000 licence at 7 percent becomes $65,500 in year five.
- Plan upgrades. One needed feature, such as manufacturing, multi-company or advanced approvals, can move every user to a higher tier.
- Customisation on top of SaaS. Most mid-sized implementations still need custom modules, and those need maintenance at every version upgrade.
- Exit cost. Data exports are rarely clean, and workflows built inside a vendor's tools do not move with you.
Custom software costs that rarely make it into the estimate
- Maintenance of 15 to 25 percent of the build every year, forever. Over five years this often equals the first build.
- Compliance you now own. Payroll rules, tax changes, e-invoicing and audit trails. A vendor spreads this cost across thousands of customers. You pay it alone.
- Key-person risk. If the system lives in two developers' heads, you have a dependency, not an asset. Insist on documentation, tests and code in a repository you control.
- Estimate overruns. Large custom builds overrun more often than subscriptions do. Hold 15 to 20 percent contingency.
- Security and hosting: patches, backups, monitoring and penetration testing.
Build, buy or hybrid: a decision checklist
Buy (SaaS or open-core ERP) when:
- The process is commodity: accounting, payroll, tax, standard inventory, standard HR.
- You have fewer than about 50 users, or a fast-changing team size.
- You do not have, and do not want, a technical owner for the system.
Build (custom software) when:
- The software is how you compete: your pricing, your scheduling, your customer experience.
- No product fits without heavy customisation that you would maintain anyway.
- You have many users at a high per-seat price, a stable scope, and a team or partner committed for years, not months.
Go hybrid when:
- A standard ERP covers 70 to 80 percent of your needs.
- You pay for several per-seat add-ons around it.
- Your differentiation lives in a few workflows, portals or dashboards rather than in the ledger.
Notes for companies in Kuwait
- Arabic and English, properly. Make sure the system supports right-to-left Arabic in the interface, invoices and reports, not only translated labels. This matters for both SaaS and custom.
- KWD has three decimals. Kuwaiti dinar amounts use fils (1 KWD = 1,000 fils), so rounding, tax and payment-gateway amounts must handle three decimal places. Test this early in any custom build.
- Local integrations decide the real cost: KNET and payment gateways, local banks, and e-commerce and delivery platforms. Our guide to ERP integrations in Kuwait covers what each involves.
- Regional pricing. Some vendors price by country, so check the Kuwait price, not the US list price. As a reference, $322,000 over five years is about 99,000 KWD at 0.307 KWD per dollar. The calculator converts automatically.
- Partner continuity. Choose a team that will still support you in five years and can work in both languages. Our Kuwait software company and Odoo partner Kuwait pages explain how we work there.
Notes for companies in the USA
- Per-seat pricing scales with headcount. Model your hiring plan, not today's user count, before signing a three-year agreement.
- Negotiate renewal caps. A written cap on yearly increases (for example 3 to 5 percent) is worth more than a first-year discount.
- Own your code and data. In any custom contract, make sure the source code, the repository and the cloud accounts are in your company's name from day one.
- Rates vary by four times. US agencies at $100 to $150 an hour and strong offshore teams at $25 to $50 an hour deliver very different totals. Use a delivery model with US-hours overlap, written scope and milestone demos. That is how you keep the saving without losing control. See our software development company for US businesses page.
What we would do in your position
If you have 10 to 50 users, start with a standard ERP and keep phase one close to standard. Our guide on how to plan an ERP budget shows how. List every per-seat tool you pay for around it. Then ask which of them are really your process, and build those.
If you have 100 or more users, a premium suite and a growing list of add-ons, run the calculator above with your real numbers. Then compare a five-year owned or open-core plan with your current renewal.
Either way, decide on a five-year total, not on the first-year price. If you would like a second opinion on your numbers, our custom software development and ERP implementation teams do this kind of review regularly. You can also message us directly below.
Frequently asked questions
- Is custom software cheaper than SaaS?
- Sometimes, but not for a full ERP at small scale. In our model a 10-user company pays about $85,000 to $98,000 over five years for a mainstream SaaS ERP. A fully custom ERP costs about $100,000 with an offshore team and about $286,000 with a US agency. Custom starts to win on cost when there are many users, a stable scope, and per-seat fees that grow with every hire.
- Has AI made custom software cheap enough to replace SaaS?
- AI has made custom software cheaper, but not by the amount the headlines suggest. Controlled studies range from about 21 percent faster on coding tasks (Google, 2024) to 19 percent slower for experienced developers on their own codebases (METR, 2025). Coding is only about half of a software project. A realistic saving on total build effort today is 10 to 20 percent. The bigger change is that small, focused internal apps are now quick to build.
- What is the total cost of ownership of a SaaS ERP over 5 years?
- For 50 users, our model puts five-year cost at about $322,000 for Odoo, $385,000 for Microsoft Dynamics 365 Business Central Essentials and $684,000 for NetSuite. It assumes list prices, $80,000 implementation, 5 percent yearly price rises and partner support. Licences are usually the largest line after year one.
- How much does it cost to maintain custom software each year?
- Plan for 15 to 25 percent of the original build cost every year. Quiet internal tools can sit near 10 to 15 percent. Products that keep changing, or that handle payroll, tax or payments, sit at the top of the range or above. Over five years, maintenance often costs as much as the first build.
- When should a company build instead of buy?
- Build when the software is how you compete, when no product fits your process without heavy customisation, or when per-seat fees for many users make subscriptions more expensive than owning. Buy for accounting, payroll, tax and other commodity processes where the vendor carries compliance updates.
- What is the hybrid approach to SaaS and custom software?
- Keep a standard ERP for accounting, inventory, HR and payroll. Then build small custom apps for the workflows that make your business different, such as a customer portal, field service app, pricing tool or dashboard. These apps often replace two or three per-seat SaaS add-ons and connect to the ERP through its API.
Keep reading
Related guides
- How Long Does an ERP Implementation Really Take? Timelines by Company SizeSmall companies go live in 2 to 4 months, mid-sized in 4 to 9, large multi-company rollouts in 9 to 18. Phase-by-phase timelines and what makes them slip.
- Odoo Implementation Cost in Kuwait and the GCC: What You Actually Pay For (2026)What Odoo costs in Kuwait, the UAE and Saudi Arabia: how licences are priced per user, what drives implementation cost, hidden costs and how to budget.
- Case Study: A 3D Product Animation Series for a US Hyperbaric Chamber CompanyHow our team produced five 3D product animations for a US hyperbaric chamber company, one film per model, and what medical equipment makers can learn.
Talk to us directly
Prefer a quick conversation?
Message us on WhatsApp or call. You'll reach the team, not a call centre.

