What does it cost to build a SaaS, AI agent, or multi-surface app?

"How much will this cost?" gets answered in two ways: with a number that's wrong, or a non-answer that's useless. Here's the third option — three real brackets, with the line items most studios leave out of the proposal.

Every cost estimate is conditional. The variables that move the price by 3× are: surfaces (one or many), regulation (none or strict), integrations (zero or many), and team you bring (none, some, full). What follows is a 2026-dollar baseline, assuming a competent product studio with senior people on the work.

Bracket 1 — Lean MVP. $40k–$120k. 4–8 weeks.

One surface (web or mobile, not both). One persona. Two or three core flows. Off-the-shelf auth, off-the-shelf payments, off-the-shelf hosting. No AI features. The goal is "is this a thing people use?" not "can we scale it."

  • Engineering: 2 people, 4–8 weeks → $30–80k
  • Design: 1 person, half-time → $8–18k
  • Infra (year 1): $200–800/month → $3–10k
  • Tools, monitoring, analytics: $1–3k

What this won't include: compliance, multi-tenant architecture, audit, white-label, BYO-model AI, complex permissions, integrations beyond one or two webhooks. If your buyer asks for any of those, you're not in this bracket.

Bracket 2 — Production-ready build. $150k–$320k. 8–14 weeks.

One or two surfaces, properly. SSO, role-based permissions, audit logging, real telemetry, real test coverage. Ready for paying customers, not just users. AI features are realistic here if scoped — a single retrieval surface, not an autonomous agent.

  • Engineering: 3 people, 8–14 weeks → $110–230k
  • Design + product: 1.5 people across the engagement → $25–50k
  • Infra (year 1): $400–2k/month → $5–25k
  • Compliance prep (SOC 2 lite, security baseline): $5–15k
  • One or two integrations: $5–20k

Trap line item: the "we'll add SSO later" promise. Adding SSO and tenancy after the fact costs as much as building it from the start. Don't defer it if your roadmap has any enterprise customer in it.

Bracket 3 — Multi-surface ship. $380k–$700k+. 14–22 weeks.

Three or more surfaces (e.g. desktop + iOS + AI), shared design system, shared data model, telemetry across all surfaces, audit-ready for regulated buyers. The lower end of this bracket is what a 14-week, 4-person multi-surface build typically looks like; the upper end is when more than four surfaces are in scope, or when a regulated rollout adds compliance review cycles.

  • Engineering: 4 people, 14–22 weeks → $260–500k
  • Design lead: 1 person across → $40–80k
  • Design system tooling, token audit infrastructure: $10–25k
  • Infra (year 1): $1.5–8k/month → $20–100k
  • Compliance, security review, penetration test: $15–40k
  • Native platform fees, certs, app store reviews: $3–10k

The line items most proposals hide.

  1. Eval and observability for AI features. Building "the AI part" is 30%. The eval harness, observability, and refusal logic are the other 70%. Real budget: $20–80k extra on top of the AI feature itself.
  2. Telemetry instrumentation. "We'll add analytics" is a one-week task that takes four. Budget for it. Telemetry is part of the deliverable in our engagements; we don't ship without it.
  3. Migration from your existing stack. If you're replacing something, you're paying for migration, not just for the new build. Add 15–25% if there's data to move.
  4. Onboarding flow. Often skipped in MVPs, then built emergency in production. Budget 5–10% of total.
  5. Year-2 cost. Operations, model spend, support, and incremental work. Plan for 30–50% of the build cost as Year-2 run-rate.

What changes the price by 2×.

  • Adding a regulated industry (healthcare, financial, legal): +30–60%.
  • Adding a third surface: +25–40%.
  • Adding real-time / sub-second latency requirements: +20–35%.
  • Custom AI inference vs. API: +$80–250k engineering, depending on scope.

The shortest version.

An MVP that proves the thing: $40–120k. A production-ready build a real customer will pay for: $150–320k. A multi-surface, regulation-ready ship: $380–700k+. The variance inside each bracket is mostly about regulation and surfaces; everything else is rounding error.

If a studio quotes lower than the bottom of these brackets, they're either junior, off-shore-only, or they're skipping a bracket of work that you'll pay for in production support later. None of those is a deal.

Oviompt's engagement pricing is in the same brackets, in writing. File an intent with your scope and we'll reply with a more precise number in 48 hours.