How Datacom's AI sandpit helps businesses fast-track safely

Jane Roughan - Senior Solution Architect PSG, Padrick Goodwin - Practice Lead PSG

• Datacom’s AI Sandpit provides a safe, governance-driven space to learn, experiment and mature AI capabilities at scale, helping organisations move from pilots to production.

• Operating on-premises or in hybrid deployments (open architecture with Red Hat OpenShift and Red Hat AI on Dell hardware), the AI Sandpit keeps data on-site for sovereignty and supports a range of AI workloads and models to control cost and risk.

• The AI Sandpit program combines structured education (masterclasses, hands-on labs and on-demand content) over 12 months to turn learning into production-ready capabilities.

Artificial intelligence (AI) will define the competitiveness of every organisation in the next five years. Indeed, when we asked 200 New Zealand business leaders what they saw as the biggest technology opportunities in 2026, more than half (51%) put AI adoption and implementation at the top of the list.

At the same time, many leaders are understandably hesitant to move beyond pilots and proof of concepts, because the risks associated with data, cost, compliance and reputation feel very real. 

What is missing in many organisations is a safe place to learn, experiment and build AI capability at scale. That’s why Datacom has spent the last year building exactly that – an “AI Sandpit” that lets teams explore powerful AI platforms and applications in a controlled environment, then graduate the best ideas into production with confidence. ​

Most organisations are under pressure from boards and CEOs to “do something with AI”, while their engineering and data teams are still working out where to start. An AI Sandpit bridges that gap.

The concept has existed in software development for years. Even nations, including New Zealand, the United Kingdom, Singapore and the United Arab Emirates, have employed sandpits to spur safe, responsible innovation in areas like fintech and blockchain development. 

• provide a secure, ring‑fenced space to experiment with realistic data, without exposing core systems or breaching sovereignty requirements.
• give teams hands‑on access to modern AI tooling so they can understand the real costs, performance characteristics and limitations before committing to large investments.
• support governance teams with a controlled environment where responsible AI policies, data handling rules and guardrails can be designed and tested in practice.

Without this structure, organisations either stall or push too quickly into public cloud services, encountering bill shock, unexpected data exposure or models that do not align with their risk appetite. 

We designed Datacom’s AI Sandpit as a containerised offering that combines hardware, software and structured education into a single package that organisations can deploy in their own data centres. 

• On‑premises or hybrid deployment: the AI Sandpit runs in the customer’s own environment, giving them full control over data location, network access and integration points.
• Deliberately non‑production by design: the environment is optimised for experimentation, with the option for larger configurations to be evolved into production clusters if needed, but with clear boundaries between “playground” and “live”.
• Flexible, open architecture: using Red Hat OpenShift and Red Hat AI on Dell hardware, customers can work with a broad mix of open-source frameworks, GPUs (NVIDIA or AMD) and even external model endpoints, rather than being locked into a single vendor stack.

This lets teams experiment with everything from fine‑tuning large language models (LLMs) to traditional machine learning (ML) and computer vision workloads, while keeping infrastructure ownership and cost control firmly in their hands. 

A core design principle of our AI Sandpit is that experimentation should not come at the expense of security, privacy or sustainability.

To achieve that, the platform can be completely isolated from the public internet, with traffic tightly controlled behind firewalls and segmented networks, dramatically reducing the risk of data leakage or unintended access. ​

Datacom’s AI Sandpit keeps sensitive data on‑premises – under customer control – supporting sovereignty requirements and alleviating concerns about sending regulated datasets offshore to public cloud providers. This also strengthens digital resilience by keeping data and workloads on-premises, helping organisations continue critical AI work and recover quickly from disruptions.

AI at the edge reduces reliance on energy-intensive cloud data centres by processing tasks closer to the user, cutting down on network traffic and associated emissions. An AI Sandpit allows organisations to experiment and optimise these edge deployments, ensuring models run efficiently on local infrastructure, minimising unnecessary compute cycles and power consumption. 

Technology alone will not make AI successful. The real differentiator is whether an organisation’s people understand how to frame AI opportunities, prepare data and work effectively with modern tooling. That is why we built a structured education layer directly into the AI Sandpit. ​

• A full‑day AI Sandpit Masterclass that brings together infrastructure specialists, data scientists and business stakeholders to work through the end‑to‑end lifecycle. This includes scoping use cases to sizing infrastructure and selecting models and understanding trade‑offs between training from scratch versus fine‑tuning.
• Hands‑on labs using familiar tools, such as Jupyter Notebooks and Python, where teams learn practical techniques for data cleansing, deduplication, feature engineering and secure handling of sensitive data.
• On‑demand training content that mirrors the in‑person course, so staff can revisit concepts and labs as they start new AI initiatives, reinforcing learning instead of losing it once the workshop ends.

We basically turn the AI Sandpit into a living education platform. Over a 12‑month program, organisations move from “we should do something with AI” to having cross‑functional teams that can scope, build and evaluate AI solutions with a clear understanding of cost, risk and value. 

The ultimate goal of our AI Sandpit is not endless experimentation, but accelerated, safer paths to production.  

We are already seeing strong interest from healthcare organisations looking to support research projects, and from universities that want to train the next generation of engineers and data scientists on real infrastructure rather than abstract simulations. The same model applies equally well to finance, insurance, public sector (central government and local government) and other data‑rich, highly regulated sectors.  

By the time teams graduate from the AI Sandpit program, they have a validated backlog of use cases, a realistic understanding of data readiness and evidence of what works technically and commercially.

Iterating in a controlled environment surfaces issues early – whether in data quality, model behaviour or integration patterns – before they can impact customers or operations.

At Datacom, our education‑focused teams work alongside our AI consulting and delivery specialists, so once a customer finishes the AI Sandpit program, there is a clean transition into designing and operating production‑grade platforms and applications. 

The AI Sandpit exists because of deep, practical partnerships with Dell and Red Hat, which, together, provide the backbone for a flexible, enterprise‑grade AI Sandpit that still feels approachable for teams taking their first serious steps into AI. 

Dell supplies the high‑quality server infrastructure that underpins our AI Sandpit offerings, giving customers reliable, well-performing hardware that can be deployed into their own data centres and, at the higher tier, scaled or clustered into production‑ready configurations if required.

Red Hat provides the open, cloud‑native software layer, with Red Hat OpenShift and Red Hat AI enabling containerised AI workloads, GPU scheduling and integration with a wide range of frameworks and accelerators, including NVIDIA and AMD, while avoiding lock‑in to a single AI vendor stack.

For leaders who want to move decisively on AI without compromising on governance, security or sustainability, a structured AI Sandpit approach provides the missing middle ground. Coupled with deliberate, high‑quality training, it gives your organisation the confidence to embrace AI at scale on your own terms, at your own pace and with your people fully equipped to make the most of the opportunity.  

Find out more about Datacom’s AI Sandpit and register your interest for one of our AI Sandpit introductory sessions.