Three phases, an exit at every one, and terms you can read before the call
We do not just write code; we build systems that change a financial result. This is how an engagement runs, what it costs, and what you own.
- to the first working deliverable
- 2 weeks
- of clients stay longer than 12 months
- 89%
- uptime achieved for a Fortune 500 platform
- 99.999%
- certified information security
- ISO 27001
From one use case to an AI operation
AI Pilot
One use case, your data, your stack. Week 1 business analysis and data access; week 2 architecture and a working slice; week 3 the pilot running in production conditions with evaluation; week 4 hardening, a go/no-go decision and a production plan with costs.
Exit: Exit with the code, the evaluation results and the plan. No further commitment.
Pilot to production
Integrations with the systems that run the operation, data governance and access controls, evaluation pipelines, observability and cost controls, runbooks, and training for the people whose work depends on it.
Exit: Every line of code and every architecture document handed over at go-live.
Run and scale
Monitoring, model and prompt updates, drift and cost management, and the next use cases on the same foundation. Pricing can track the outcome; in exchange we agree a scope, a measured baseline and written exit criteria.
Exit: 30 days notice. 89% of clients stay longer than 12 months by choice.
How AI gets built here
The practices that separate a demo from a system your team trusts.
Evaluation before scale
Every AI system ships with an evaluation set and pass criteria agreed with you. We measure accuracy, latency and cost before and after every change.
Human checkpoints by design
Agents run autonomously between defined checkpoints. Where judgment matters, a person is in the loop, and the interface makes that fast.
Observability and cost control
Traces, audit logs and spend per workflow from day one. You can see what the system did, why, and what it cost.
Model-agnostic, your infrastructure
Claude, OpenAI, Gemini or open models behind an abstraction you own, on AWS, Databricks, Azure, Oracle or on-premises.
Four ways to bring us in
Every engagement starts with a discovery call, an NDA and a proposal within days: a clear roadmap and our technical assessment, with real value and without obligations.
Software and technology audit
An external assessment of your systems, data and delivery process that ends in concrete strategies: automations, open-source components and custom solutions where they pay off.
Third-party solutions first
If an existing market solution is optimal, we recommend and integrate it. We do not push custom development where it is not needed.
Technology consulting and advisory
Architecture, AI readiness, security and cost decisions with an engineering leader who has made them before.
Becoming part of your team
A dedicated team or a boost to your in-house department, working in your tools, your cadence and your time zone overlap.
Compare the models
| Model | Best for | Scope | Pricing | Team | Notice |
|---|---|---|---|---|---|
| AI Pilot | Validating one AI use case fast | Fixed | Fixed fee | 4–5 cross-functional experts | None after week 4 |
| Fixed scope | Well-defined deliverables and budget | Fixed, change-controlled | Fixed price per milestone | Sized to scope | Per contract |
| Time and materials | Evolving requirements, product work | Flexible, prioritized backlog | Monthly, in arrears, net 15 | Scales up or down monthly | 30 days |
| Dedicated team | Long-running platforms and AI operations | Roadmap owned with you | Monthly per team member | Named engineers, PM, QA, BA | 3-month trial with 14 days notice, then 30 days |
Full commercial terms: refund and cancellation policy. Invoicing monthly in arrears, net 15; 10-day acceptance window; IP transfers on full payment.
Triple legal protection for your intellectual property
- Work-made-for-hire contract through our American company, backed by professional liability insurance.
- Enforceable agreements. All Erbis employees sign NDAs and IP assignment agreements under local legislation.
- Direct agreements. Optional NDAs and creative-rights assignments between Erbis engineers and your organization.
ISO 27001 controls we operate
- Secure development
- Risk management
- Incident management
- Physical access control
- Hardware security
- Network security
- Information security policy
- Supplier security
- Employee training
Two-week sprints, weekly demos
Agile and Scrum are the delivery framework. Work is split into two-week sprints with a demo at the end of each, weekly written updates, and a shared backlog you can reprioritize at any time. Changing requirements is welcome, even late in development.
Every team ships with a project manager, QA and business analysis. We study the commercial side of the problem, keep the workflow predictable and stay on after launch for support and the next increment.
How we work: questions we get asked
How fast can we start and see something working?
A discovery call and NDA in the first week, a proposal within days, and work starts on signing. For a 4-week AI pilot you see the system running in production conditions by week four. For development engagements we have something to show in as little as two weeks, then demos every sprint.
Who owns the code, the models and the data?
You do. Work is contracted as work made for hire through our US company, backed by professional liability insurance. Every employee signs NDA and IP assignment agreements, and you can add direct agreements with the engineers on your team. Intellectual property transfers on full payment for each invoice.
What are the commercial terms?
Monthly invoicing in arrears with net-15 payment. Master agreements carry 60 days notice. Dedicated teams start with a three-month trial with 14 days notice, then 30 days. Deliverables have a 10-day acceptance window. The full terms are published in our refund and cancellation policy.
How do you keep our data safe?
Erbis is ISO 27001 certified. Controls cover secure development, risk and incident management, physical and hardware security, network security, supplier security and employee training. For regulated work we design HIPAA, GDPR and PCI DSS requirements into the architecture and keep data inside your environment when required.
Do you work in our time zone?
Headquarters are in San Francisco with engineering in Kraków and Dnipro. Teams are staffed for overlap with US business hours, and every engagement has a named delivery lead you can reach directly.
What if the pilot shows custom AI is not the right move?
Then we say so, in writing, with the evidence. The pilot exists to learn what we do not know before anyone commits to outcomes. You keep the code, the evaluation results and the recommendation, including when the recommendation is an off-the-shelf tool.
If your team spends real time working around what your systems cannot do, we should talk.
A discovery call takes 45 minutes. We will tell you directly whether custom AI is the right move. If it is not, we will say so.
AWS partner · Databricks partner · ISO 27001 · You own the code from day one