Startups Building ERP With AI Put SaaS Costs on Trial
Key Takeaways
- Audit SaaS usage against actual workflows before renewing broad platforms.
- Treat internal AI builds as owned software, with maintenance and security costs included.
- SaaS vendors should defend ecosystems and workflow reliability, not commodity features.
Why it matters
- ProductProduct leaders need to know which workflows customers can now rebuild cheaply.
- InvestorsInvestors should test whether SaaS revenue is protected by ecosystem depth or only feature breadth.
AI coding is making custom internal systems plausible again, which turns SaaS pricing into a product strategy problem.
Some of the most interesting product launches never get a landing page. They happen when a startup quietly replaces a vendor, ships an internal tool, and tells finance the recurring bill can stop. That is the sharp edge of the new ERP build vs buy debate: AI is not just helping startups buy software smarter, it is making some of them ask why they are buying at all.
The build button is getting cheaper
Seoul Economic Daily reported that more startups are building business software internally instead of subscribing to SaaS products, because frontier AI models have reduced the time required to make those systems. The examples are not toy dashboards. According to the paper, startups are building ERP tools for accounting and human resources, plus groupware such as electronic approval systems.
The money story is concrete enough to make a CFO lean forward. Seoul Economic Daily said one startup had paid about 12 million won, $8,200, a year for a large conglomerate workplace platform while using only electronic approval and expense settlement functions, then eliminated that cost with its own program. The same report said another startup is developing ERP features with the goal of phasing out subscription fees of about 5 million won a month. That is not a market forecast, but it is a product warning flare.
The old buy argument has a new weak spot
According to Seoul Economic Daily, the mechanism is vibe coding, where AI writes code from natural language instructions, combined with reverse engineering of existing software structures. That combination matters because the classic SaaS sales pitch has always been, please do not spend scarce engineering time rebuilding internal plumbing. If AI shrinks the cost of that plumbing, the pitch has to move from convenience to durable advantage.
This is where the second-order effect shows up. A startup that only needs approvals and expense settlement does not experience the full value of a broad workplace suite, it experiences the invoice. This pricing page is a Choose Your Own Adventure where every ending is expensive if the customer uses only two rooms in the mansion. Seoul Economic Daily also noted that existing software firms face pressure when trying to tailor AI features for individual companies while keeping user-based pricing intact.
Incumbents are not standing still
TechTarget’s interview with NetSuite founder Evan Goldberg offers the incumbent counterargument. TechTarget reported Goldberg’s view that AI makes ERP more accessible to startups and other SMBs, similar to how cloud software changed access after NetLedger was founded in 1998. He also said AI can help produce an implementation that fits quickly, easily, and inexpensively.
The most interesting line was economic, not technical. Goldberg told TechTarget that NetSuite AI costs “zero over your existing NetSuite subscription.” That is the obvious defensive move for SaaS vendors: make AI part of the paid product, reduce implementation pain, and remove the most tempting reason to rebuild. If the internal build saves money only before maintenance, permissions, security, and process exceptions arrive, the vendor still has a game to win.
The moat moves from features to ecosystems
Seoul Economic Daily cited a venture-capital official who argued that SaaS products with strong internal and external ecosystems, with Slack named as an example, may be better positioned if AI-assisted rebuilding spreads. That is the competitive map hiding inside the ERP story. Feature parity is becoming a weaker moat when a small team can approximate workflows faster than before.
The stronger moat is the stuff customers hate migrating: ecosystem gravity, workflow history, organizational habits, integrations, identity rules, and the quiet trust that month-end will not become a bug hunt. For startup operators, the build decision now needs two ledgers, one for subscription savings and one for ownership costs. For SaaS builders, the answer is not another badge that says AI-powered; it is packaging that makes the ecosystem, data model, and workflow reliability impossible to casually clone.
What to watch next
GitHub Enterprise Cloud Docs show why this trend has room to compound. GitHub says Copilot in the cloud can use a dotfiles repository for preferred instructions, reusable skills, and specialized agents across repositories, and that personal custom instructions can apply to every conversation in Copilot in the cloud. That kind of tooling pushes internal software creation from artisanal side project toward repeatable operating habit.
The next logical move is clear. Buyers should test whether an internal build really replaces a workflow, not just a subscription line item. Vendors should assume the low end of their feature set is now under permanent price pressure, then defend the parts customers cannot rebuild over a long weekend with an AI coding assistant. The winners will be the teams that know which software is a moat and which software is just rent.
Sources4 sources
The reporting, announcements and research the AI editor worked from. Links open the original publisher.
- Startups Build Their Own Software Instead of Paying for SaaS - Seoul Economic Dailyen.sedaily.com
- Startups Build Their Own ERP With AI, Cutting SaaS Subscription Costs - Seoul Economic Dailyen.sedaily.com
- NetSuite founder Evan Goldberg on AI, the end of ERP as we know ittechtarget.com
- Personalizing your cloud sessionsdocs.github.com
