Getting Started

The first step is an introductory call — usually 30 to 45 minutes — where we discuss your business objectives, current technical landscape, and what success looks like for your organisation. This is a two-way conversation: we need to understand your constraints and ambitions, and you need to assess whether our expertise aligns with your needs. There is no preparation required for this call beyond having a general sense of what you want AI to accomplish. After the call, we provide a brief summary of our recommendations and a proposed next step, typically a paid discovery engagement.

No formal preparation is needed, but having answers to a few questions will make the conversation more productive. Consider your primary business objective — what problem are you trying to solve or what capability do you want to build. Think about your current data landscape — where does the relevant data live and in what formats. Understand your infrastructure posture — are you cloud-native, on-premises, or hybrid. Know your compliance landscape — which regulatory frameworks apply to your organisation. And have a sense of timeline and urgency — is this exploratory or do you have a deadline driving the initiative.

The assessment phase is a structured two to three week engagement. We interview key stakeholders across business, technology, data, and compliance teams. We audit your existing data assets, infrastructure, and AI capabilities. We evaluate vendor options and technology choices against your specific requirements. We identify quick wins that can demonstrate value within weeks alongside longer-term strategic initiatives. The deliverable is a comprehensive report with a prioritised roadmap, reference architecture, resource plan, and investment estimate. This becomes the foundation for all subsequent work.

The best first projects share three characteristics: they address a real business pain point, they have accessible and well-understood data, and they can show measurable results within eight to twelve weeks. Common starting points include internal knowledge search using RAG over your document repositories, automated document processing for contracts or compliance filings, customer support augmentation that helps agents find answers faster, and code review or documentation assistants for engineering teams. We deliberately avoid moonshot projects for the first engagement because building organisational confidence in AI requires visible, quick success.

No. Many of our clients engage us precisely because they want to build AI capability but do not yet have a dedicated team. We can operate as your AI engineering function during the initial phase, delivering production systems while simultaneously helping you define the roles, skills, and team structure you will need long-term. As your internal capabilities grow, we transition from implementation to advisory. That said, you will need at least one internal technical champion — someone who understands your data and systems and can facilitate access and decision-making throughout the engagement.

For cloud-based AI projects, you need an active cloud account with GPU instance quotas — we can help request quota increases if needed. You will also need a version-controlled code repository, CI/CD pipeline capability, and access to the data sources we will be working with. For on-premises deployments, you need rack space, power, and cooling capacity for the GPU hardware plus network connectivity to your data sources. If you are starting from scratch, we can architect and provision the entire stack. Most clients have some infrastructure in place and we build on what exists rather than replacing it.

Knowledge transfer is built into every engagement, not bolted on at the end. Your team members participate in architecture decisions and code reviews throughout the project. We pair-program with your engineers on complex components. We produce runbooks and operational documentation that reflect your actual deployment, not generic templates. At project completion, we conduct structured handover sessions covering architecture, operations, troubleshooting, and maintenance procedures. For larger engagements, we offer a transition period where your team operates the system with our team available for escalation support.

From a typical first engagement, you should expect a working, production-grade AI system addressing a specific business need, measurable performance metrics demonstrating business value, comprehensive documentation and runbooks for ongoing operations, a trained internal team capable of maintaining and extending the system, a clear roadmap for subsequent AI initiatives based on lessons learned, and an honest assessment of what worked and what to adjust going forward. We measure success by whether the system is still running and delivering value six months after we hand it over, not by how impressive the demo looked on day one.

Transparency is our default. If we encounter unexpected technical challenges, data quality issues, or scope changes, we raise them immediately with specific options for resolution. Every engagement includes defined checkpoints where we review progress against objectives and can adjust scope, timeline, or approach. If during discovery we determine that AI is not the right solution for your problem, we will tell you directly rather than building something unnecessary. We have occasionally recommended that clients invest in data quality or process improvement before starting an AI project, because doing so dramatically improves the eventual outcome.

Absolutely, and we encourage this approach. A pilot engagement typically runs four to six weeks with a deliberately narrow scope — one use case, one data source, one user group. The goal is to validate the technical feasibility, measure real business impact, and build organisational confidence before scaling. Pilots are priced as standalone engagements with no obligation to continue. If the pilot succeeds, we use the learnings to scope the production engagement with much higher confidence in timeline and budget. If it reveals that the approach needs adjustment, you have that insight at minimal investment.