How much does a custom web or AI app cost in 2026?
Realistic budgets for a business app, SaaS product or AI integration in 2026. Price ranges, timelines, hidden costs and how to scope a V1 without losing six months.
The wrong question is: "how much does an app cost?"
The useful question is: which first version has to exist before the product starts creating value?
In 2026, published ranges for custom web apps run from a few thousand euros to several hundred thousand. Genee gives a French market range from €5,000 to €500,000 for a custom web application, with most business app projects between €20,000 and €80,000 for a first functional version. INYSTER also places the common range for French SMBs between €15,000 and €80,000 for a first production version.
That is not a contradiction. It means the word "app" is useless until the scope is signed.
What quotes often hide
Market ranges, not a commercial promise. The real price depends on signed scope, integrations and risk level.
€15-40k
Business app MVP
3 to 6 core features
€15-40k
SaaS MVP
auth, payments, product base
€40-80k
Production V1
reliability, monitoring, tests
€13-60k
Scoped AI project
diagnosis, approach, implementation
Average price of a custom web app in 2026
For a business web app, a serious first version often lands between €15,000 and €40,000 when the scope is focused: a few core features, one or two user roles, a dashboard, one or two third-party integrations. Genee uses the same order of magnitude for a business MVP.
A broader app with a client portal, complex roles, billing, sensitive data, notifications, audit trail, monitoring, tests and documentation can quickly move toward €40,000 to €80,000, sometimes more. The same brief can produce two very different quotes if one provider prices only screens and the other includes architecture, tests, security and handover.
Price is not just a sum of pages. It mainly depends on five things:
- business logic complexity;
- number of user roles;
- external integrations;
- expected security level;
- number of decisions still open when the quote is written.
The more uncertainty remains, the bigger the quote becomes. Or it looks low, and the cost comes back later as change requests, technical debt or a rebuild.
Price of a custom SaaS product in 2026
A SaaS costs more than a standard business app because it usually carries a complete product base: user accounts, subscriptions, payment, permissions, transactional emails, onboarding, metrics, support, sometimes multi-tenancy.
Pixel Paris gives a useful range: €15,000 to €40,000 for a SaaS MVP, then €40,000 to €80,000 for a production V1 with stronger reliability, tests, monitoring and performance. The published timeline for a SaaS MVP is typically 4 to 8 weeks, while a full V1 usually takes several months.
At 21Day, we do not promise to ship "the whole SaaS" in 21 days. We ship a version that can be used, shown, tested, sold or rejected. That is different.
A good 21-day SaaS scope looks like this:
| V1 scope | Keep | Push to later |
|---|---|---|
| Authentication | email, simple roles, password reset | enterprise SSO, granular permissions |
| Payment | Stripe checkout or simple subscription | complex pricing, advanced coupons |
| Product | one central usable workflow | ten secondary modules |
| Admin | read, correct, useful exports | complete back office |
| Quality | logs, errors, basic security | full product analytics |
This is not lowering the bar. This is how you get a product in hand before the market, idea or internal priority changes.
A project, a question, a doubt.
Drop us a line. We answer within 24 hours, no script.
Price of an AI product or AI integration
AI rarely adds cost because it is "AI". It adds cost when it touches data, permissions, answer quality, traceability, security, internal tools or critical business workflows.
The French Ministry of Economy details the IA Booster France 2030 programme: a Data AI diagnosis at €13,000 excluding VAT with 50% coverage, an AI approach selection mission at €13,000 excluding VAT, and implementation support up to €60,000 excluding VAT, subsidised at 50% for eligible companies.
Those numbers matter because they send a public signal: a useful enterprise AI project is not just a chatbot wired quickly. You need to scope the use case, data, deployment and responsibility.
A simple AI product can stay within a controlled budget if the use case is precise:
- internal assistant on a limited knowledge base;
- human-reviewed draft generation;
- extraction from recurring documents;
- qualification, sorting or summarisation workflow;
- business interface calling an AI model within a bounded process.
The budget rises when you add strong constraints: sensitive data, long history, compliance, costly wrong answers, multiple integrations, detailed permissions, answer monitoring, prompt audit, specific hosting.
Why quotes blow up
Most quotes do not blow up because of code. They blow up because of assumptions.
A client says: "we need a user space." For one developer, that can mean three screens. For another, it means signup, login, password reset, two-factor authentication, roles, permissions, history, logs, exports, account deletion, privacy compliance, emails, support and admin.
Same sentence. Two different products.
The second trap is false savings. A low quote can exclude tests, documentation, backups, monitoring, data migration or maintenance. INYSTER notes that a €20,000 quote without tests or documentation is not comparable to a €35,000 quote that includes CI/CD, monitoring, tests and training.
The third trap is confusing MVP with a neglected product. DBB Software frames it well: MVP cost mostly depends on scope and complexity, and the goal is validation, not building a complete product.
What can really ship in 21 days
In 21 days, you can ship a useful version if the scope is clear.
You can ship:
- a business app replacing a critical spreadsheet;
- a simple client portal;
- an AI assistant on bounded data;
- an operational dashboard;
- a first SaaS version with one central workflow;
- a business automation with a control interface.
You should not promise:
- a full ERP;
- a mature multi-sided marketplace;
- a SaaS with ten modules;
- a native iOS + Android app with a complete back office;
- a critical AI system without security scoping.
The 21Day method is simple: we do not reduce quality, we reduce scope. The code still has to be clean, maintainable, documented and handed over. What sits outside the scope goes to V2, not into a fragile corner of the product.
The checklist that lowers the price
Which business decision should the product make possible?
Who uses it on day 22, and to perform which action?
Where is the data, who can access it, what should stay out?
Which tools are mandatory in V1?
Which risks must be handled before launch?
Who maintains, measures and evolves the product?
How to frame your budget before asking for a quote
Before asking for a price, prepare one page. Not a forty-page specification. One clear page.
It should answer six questions:
- Which precise problem must disappear?
- Who uses the product first?
- Which action must be possible on delivery day?
- Which data is required?
- Which tools must be connected in V1?
- What can wait thirty days?
If you cannot answer, the first budget item is not development. It is scoping. Good scoping can lower the total cost because it removes what does not need to be built now.
Our reading grid at 21Day
When we price a project, we do not start from a feature list. We start from the moment where the product becomes useful.
We look for the smallest version that meets these criteria:
- it can go live;
- it can be used by real people;
- it handles the main project risk;
- it leaves the code with the client;
- it can be picked up by another team;
- it does not depend on a hidden provider.
If that version fits in 21 days, we can scope it, price it and ship it. If it does not fit, we prefer saying it on the first call rather than selling a timeline that will not hold.
For a deeper breakdown by project type, see our web development agency, AI agency, SaaS agency pages or go straight to the 21Day offer.
Useful sources and benchmarks
- Genee: Custom web app cost in 2026
- INYSTER: Custom web app cost in 2026
- Pixel Paris: Custom SaaS cost in 2026
- DBB Software: MVP development cost 2026
- French Ministry of Economy: AI support and training for businesses
Conclusion
A custom web or AI app does not become expensive because it has many screens. It becomes expensive when it carries too many undecided decisions.
The right budget is not the one that buys the most features. It is the one that buys the right first version: usable, maintainable, secure, shipped fast, and clear enough to decide the next step.
If you need a product in hand, not a six-month quote, we can scope it in one call.
We don't sell code. We ship a product.
Three weeks from the first call to live. Your code, 30-day warranty.