AI Production Launch
Your AI demo works. Production is a different story.
In up to 6 weeks we put your AI features into production, in your product or your operations, together with everything that makes them safe to run: evaluations on your data, data isolation, cost limits, monitoring and answers for your customers’ security reviews. A fixed price from €20,000 to €35,000, set after a 5-day audit for €3,500 that’s credited in full.
The AI Production Launch is a fixed-price engagement from Smoother Development AB: in up to 6 weeks we put one to three AI features or workflows into production together with a production kit (evaluations on your data, data isolation and permissions, cost limits, monitoring, a security-review pack and a handover your team can run). The price is set after a 5-day AI Production Audit (€3,500, credited in full) and ranges from €20,000 to €35,000 depending on scope.
What the price pays for
With AI coding tools, a working demo takes days. What takes weeks is everything that makes it safe to put in front of customers, and that’s the part most teams postpone. The launch delivers the features and this production kit with them.
- Evaluations on your data
- 150–300 real cases from your product or process, scored automatically. Change the model or the prompt and you see within minutes whether quality went up or down.
- You get: Evaluation set and test harness in your CI
- Data isolation and permissions
- The AI only sees what the user is allowed to see, and in a multi-tenant product it never crosses from one customer to another. Tested against prompt injection and data leaks.
- You get: Permission-aware retrieval and a security test suite
- Cost and speed under control
- Cost per request and per customer measured before launch, with caching, a cheaper model for simple requests and hard limits, so growing usage doesn’t eat your margin.
- You get: Cost model and budget alerts
- Monitoring
- Every request traced, with views of quality, cost and latency, alerts, and a fallback when the model or the provider fails.
- You get: Dashboards, alerts and a fallback path
- A safe release
- Behind a feature flag, rolled out to a defined group first, with human review where a wrong answer is expensive, and a way back.
- You get: Feature flag, rollout plan and rollback
- Answers for security reviews
- Data-flow diagram, models and subprocessors, EU hosting, a data processing agreement and a short note on how the EU AI Act applies, written so you can reuse it in your customers’ security questionnaires.
- You get: Security-review pack
- A handover your team can run
- Your engineers review our pull requests from week one. You get a runbook, architecture notes and a handover session, and the kit is reusable for your next AI feature.
- You get: Runbook, architecture notes and a handover session
Who it’s for
Software companies
B2B software companies with roughly 20–250 people that want an AI feature in their product, especially if you sell to larger companies or regulated industries whose security reviews now ask about AI. We work in your repository and through your code review, so your team keeps shipping the roadmap.
- An assistant that answers from each customer’s own data in your product
- Extracting and classifying documents inside your workflows
- Summaries, drafts and search over data your users already have
- An agent that carries out steps in your product, with approval
Companies without an AI team
Companies with roughly 50–250 employees in the Nordics that use ChatGPT or Copilot but haven’t connected AI to their own systems and data.
- Answering customer and supplier questions from your product data and documents
- Classifying and routing incoming e-mails, orders and invoices into your ERP
- Drafting quotes and tenders from your past projects
- An internal assistant over your policies, contracts and project documents
It isn’t for an AI strategy deck or a chatbot on your marketing site. And if your team would rather build it themselves, the audit report is written so they can.
AI Production Audit
€3,500, 5 business days
- Day 1 A 2–3 hour workshop with the product or process owner and an engineer from your team.
- Days 2–4 We review the systems, data and code involved, check data quality and access, go through data flows for security and GDPR, and test candidate models on a sample of your real cases.
- Day 5 A written report: the recommended feature, architecture, evaluation plan and baseline, cost per request, written acceptance criteria, risks, and a fixed-price launch proposal.
One honest outcome is "don't build it". If we don't think it will pay off, the report says so.
The €3,500 is credited in full if you start the launch within 60 days.
AI Production Launch
€20,000–35,000, fixed after the audit, up to 6 weeks
- First 2 weeks Access to your repository and cloud, the evaluation set built from real cases, and a first end-to-end slice. Sprint-1 demo against the acceptance criteria.
- Middle Integration into your product or systems, permissions and data isolation, cost and latency limits, tracing. Mid-point demo with evaluation results.
- Last 2 weeks Security tests, monitoring and alerts, the security-review pack, staged rollout to the defined user group, and the handover.
You get
- The features live for a defined user group, behind a feature flag
- The evaluation set and harness running in your CI, with results against the baseline
- Dashboards and alerts for quality, cost and latency
- The security-review pack, for your customers and your own compliance
- Runbook, architecture notes, and all code and infrastructure in your repository
Price by scope, never above €35,000
The audit fixes the price before you sign. Every size includes the full production kit; what changes is how much AI functionality goes into production.
An AI feature is one thing the AI does for a group of users, such as answering from documents, extracting data from invoices or drafting replies. Need more than three? The audit proposes a fixed-price plan, often a launch followed by a Delivery Pod.
Build it yourselves or with us?
You can build it yourselves, and plenty of teams should. This is what changes when we do it together.
Price and payment
- AI Production Audit 5 business days, credited against the launch
- €3,500
- AI Production Launch fixed after the audit, up to 6 weeks
- €20,000–35,000
- AI Run, after launch per month, month to month
- €1,500
- Payment 30% at signing, 40% at the mid-point demo, 30% at production acceptance
- 30 / 40 / 30
Prices exclude VAT. Invoiced by Smoother Development AB, Sweden.
First-sprint guarantee
We agree the acceptance criteria in writing during the audit. If the sprint-1 demo doesn't meet them, you can stop the launch, and you don't pay for that sprint.
Book the AI Production AuditWho works on your launch
A senior engineer who leads the build, a second engineer full-time, shared product and UX support, and a founder (Daniel Cherman or Vladyslav Sokolovskyi) accountable for delivery. Engineers work from our team in Ukraine; the contract, the data processing agreement and the accountability sit with our Swedish company.
Your data
- Hosting in an EU region, preferably in your own cloud account
- A data processing agreement with the EU Standard Contractual Clauses
- Least-privilege access for every engineer
- Model APIs that don't train on your data, or a self-hosted model when data must stay in your environment
After launch
Keep your AI features healthy with AI Run (€1,500 a month): evaluations re-run when models change, a monthly quality and cost report, model and prompt upgrades, and up to 15 hours of fixes. Build the next features with a Delivery Pod, or run everything yourselves: you own the code.
Pricing →AI we've built
CyberResilient
AI-driven NIS2 maturity assessment, developed with support from MSB and NCC-SE.
Read the caseUpliftIQ
Decision intelligence with role-adapted AI agents and RAG over business data.
Read the caseMultilingual product catalog
2,500+ products in 4 languages with automated PDF catalogs.
Read the caseRelated reading
What a 5-day AI production audit covers (and what you get at the end)
Day by day: what happens in a 5-day AI production audit, what we need from you, what the report contains, what it costs and when it says don't build.
BlogFrom AI proof of concept to production: a 6-week playbook
How to take one AI workflow from proof of concept to production: evaluation set, data access, integration, guardrails, monitoring, cost control and handover.
BlogThe real cost of AI integration: what to budget for
What AI integration really costs: build, model tokens, infrastructure, evaluation and maintenance, with current LLM API prices and a worked example.
Questions about the launch
We have engineers and AI coding tools. Why pay for this? +
Because the demo isn’t the expensive part. Most of the work goes into evaluations, data isolation, cost control, monitoring and the security-review pack, which in-house teams tend to postpone. If your team has the time and wants to build it themselves, the audit report is still useful: it’s written so they can follow it.
What if our AI feature is simple? +
Then you pay for a smaller size: one feature with one data source is €20,000 and takes about 4 weeks. The production kit is included at every size, because that’s the part that’s hard to get right.
What if our data isn't ready? +
That's what the audit is for. We check data quality and access in days 2–4, and the report tells you what has to change first. Sometimes the right next step is a data clean-up, not a launch.
Which AI models do you use? +
We're model-agnostic. We use models from OpenAI, Anthropic and Google through EU-region or enterprise endpoints such as Azure OpenAI and AWS Bedrock, or open models hosted in your environment. We choose per workflow based on accuracy on your evaluation set, cost and data requirements.
Can AWS co-fund the launch? +
Some projects qualify for AWS co-funding programmes. We check eligibility during the audit, but we never price the launch on the assumption that it will be funded.
Who owns the code? +
You do, once the related invoice is paid. All code and infrastructure live in your repository and cloud account from the start.
How much of our time does it take? +
About 1–2 hours a week from a product or process owner, an engineer from your team who reviews our pull requests, and access to the systems and data the feature uses. The audit workshop on day 1 takes 2–3 hours.
What if the workflow doesn't fit 6 weeks? +
We'll say so in the audit report and propose a smaller first step, or a longer fixed-price plan. We don't start a launch we don't believe fits.
Why is your engineering team in Ukraine, and what if there's a disruption? +
Our engineers have worked with Nordic and European clients through power and connectivity disruption since 2022, with backup power and connections. Every launch has a second engineer who knows the codebase, and code is pushed to your repository daily. The Security page has the details.
Book the AI Production Audit
Tell us which feature or workflow you want in production and what it touches. A founder replies, usually within one business day, to agree the audit scope and dates.
Book a 30-minute call
Pick a time in Daniel's calendar. You'll get an invite with a video link, and we'll come prepared with questions about your goal.
Pick a time- 1 A founder reads your message and usually replies within one business day.
- 2 We agree a time for a 30-minute call to understand your goal, systems and constraints.
- 3 You get a written next step: a fixed-price proposal, a paid audit, or an honest "we're not the right fit".
Prefer e-mail? [email protected]
Or send us a message