AI software development: apps and products with a language model doing real work inside them
AI software development at Goldenroz means building a complete product, web or mobile, in which a language model reads, writes, decides or acts as part of the main workflow. We build the model layer and everything around it: accounts, database, permissions, billing, fallbacks and cost tracking. Projects start at $25,000 at a fixed price per phase after a free scoping call. A working prototype typically takes one to two weeks, and production roughly four to eight weeks depending on integrations.
From $25,000, fixed price per phase
Updated 2026-11-14. By Shai Goldenberg
What usually goes wrong
- The demo works, the product does notA prompt in a notebook answers ten test questions well. With real users it meets empty inputs, long documents, slow responses and people trying to break it.
- Model bills nobody can explainToken usage grows with every feature and every retry. Without cost recorded per user and per run, you cannot price the product or find the feature that loses money.
- One provider outage takes the app downA single model call with no timeout, no retry and no fallback turns a provider incident into your incident.
- The model can see more than the user mayIf the model queries data with a service key, a well-phrased request can return another customer's records. Permissions have to apply to the model's tools, not only to the screens.
What is included
- The application itselfWeb app in TypeScript and React, and iOS and Android through React Native when needed, with authentication, roles and a PostgreSQL database.
- Model layer with fallbackClaude, OpenAI or another model chosen per task, with timeouts, retries, a fallback chain and a fixed reply when every call fails.
- Tools and data accessThe model reaches your data through defined tools or MCP servers with per-user permissions, read-only by default. See AI integration.
- Usage metering and billingToken and cost tracking per run and per customer, connected to Stripe plans, credits or usage-based pricing.
- Evaluation set and audit logA set of real examples you rerun after each change, and a log of every model call, tool call and result.
AI app development is mostly ordinary engineering around one unreliable part
In AI app development, also called AI application development, the model call is a small share of the code. The rest is what any product needs: sign-in, data model, permissions, payments, release pipeline. The difference is that one component gives a different answer each time and sometimes gives none. The design job is to decide what the model is trusted with and to keep everything else as plain code.
We write that split down first. Anything that can be a rule stays a rule. The model handles language: reading a message, drafting a reply, extracting fields, choosing the next step from a short list. Actions that cost money or cannot be undone wait for a person's approval.
What to check in an AI software development company
Ask any AI software development company or AI app development company to show a system that runs in production, and how it behaves when the model fails. Shai, our CTO, built and runs a WhatsApp ordering bot on Claude with a model fallback chain and deterministic fallback replies, a support assistant with a knowledge base, and durable agent loops with token and cost tracking. He also built a research copilot product in about three days. More about his work is on Shai's page.
We build on existing models. We do not train foundation models or do classic machine learning research, and we do not claim experience there. If your product depends on a model trained on your own data from scratch, a research team is the right hire.
How the work runs
Scoping call
Before the projectWe map the process, the systems involved and the cost of a wrong action. You get a fixed price for the first phase.
Prototype
About 1 to 2 weeksA working agent on one narrow task, run against real examples in a test environment. You see where it fails before you pay for production.
Production build
Roughly 4 to 8 weeksIntegrations, permissions, fallbacks, approval steps, logging and the evaluation set. The range depends on how many systems the agent connects to.
Launch with limits
At launchThe agent starts on a share of the volume, or with every action approved by a person. Limits widen as the logs show it holds.
What it costs
Every phase has a fixed price agreed before work starts. Model usage is paid to the model provider from your own account.
- Focused agent or agentic workflowFixed price per phaseFrom $25,000
- Larger AI platformsSeveral agents, many integrations or a full product$100,000 and up
- Scoping callEnds with a written scope and a price$0
- Model usageBilled to your own accountPaid to the provider
Who it is not for
- Training a foundation model or classic machine learning research.
- Products where the model moves money or deletes data with no human approval.
- Budgets under $25,000.
Who does the work
Shai leads the technology side: business software, SaaS and mobile apps, payments and escrow systems, Postgres security, AI agents built with Claude Code and MCP, and IT and infrastructure projects with senior infrastructure engineers.
Questions buyers ask
How much does AI software development cost?
Goldenroz projects start at $25,000 at a fixed price per phase. Larger AI platforms with several agents, many integrations or a full product run $100,000 and up. Model usage is paid to the provider from your own account. See pricing.
How long does AI product development take?
A working prototype of a focused feature typically takes one to two weeks. A production release takes roughly four to eight weeks, depending on how many systems the app connects to. A full product with billing and mobile apps is built over several phases.
Which models do you use for AI development?
We choose per task among Claude, OpenAI and other hosted models, and build a fallback chain so one provider's outage does not stop the product. The provider accounts are yours.
Can you add custom AI solutions to an existing app?
Yes. Adding an assistant, extraction or drafting feature to an existing product is a common first phase. We review how your current code handles permissions first, because the model must respect the same rules.
Who owns the prompts, code and data?
You do. The code and prompts live in your repository from the first commit, and data stays in your database and your model provider account. You also receive the evaluation set and a runbook.
Guides on this topic
- Claude Code hooks run your code at fixed points and can block actions. Subagents isolate work in their own context. See a working PreToolUse guard hook.
- Twelve generative AI use cases grouped by what the model does, with the data each one needs, the main risk and a scoring method to order them in a roadmap.
- What an AI agent costs to build and to run: the five cost drivers, project shapes by effort, and a worked running-cost example. See where your project lands.
AI by industry
- AI agent development
- AI for accounting
- AI code review and AI coding training
- AI strategy and readiness
- Claude consulting and implementation
- AI for construction
- AI for education
- AI for enterprise
- AI for finance and banking
- Generative AI and LLM development
- AI for healthcare
- AI for insurance
- AI for law firms
- AI for logistics
- AI for manufacturing
- MCP and Claude integration
- AI for real estate
- AI for retail and ecommerce
- AI for small and mid-size business
- AI
Talk to the person who will do the work
One hour, no sales team. You talk to Shai Goldenberg, the CTO, about your systems or your product, and you leave with a clear next step and a price range.