For many organisations, the conversation about AI still begins with tools. Which model are we using How powerful is it? How quickly can it generate outputs? In reality, most people still experience AI through a chat interface asking questions, generating text, or solving small, discrete problems. Even for more advanced users, AI is often something you “use” rather than something that fundamentally reshapes how work happens.

At SARD, we’ve found that this way of thinking only takes you so far. The real shift for us hasn’t come from adopting AI tools in isolation, but from building Marion – our AI orchestration layer. That starts from a completely different question. Not what can AI do? but what are we trying to achieve? And how should human and artificial intelligence work together to get us there?

From Tools to Intent

One of the most common misconceptions about AI is that productivity gains alone are transformative. In practice, speeding up individual tasks like writing emails, summarising documents, analysing data, doesn’t change how work actually flows through an organisation.

Most of the effort in any piece of work lies not in doing the task itself, but in working out what needs to be done, why it matters, and how it connects to everything else. That’s where intent becomes critical.

Marion was built around this idea. Instead of interacting with AI in isolation, you define what you’re trying to achieve, and Marion works from there, connecting your intentions to the wider context of your work, your organisation and your priorities.

In my day-to-day use, Marion does not just respond to prompts, she actively identifies what matters. She reviews my emails, surfaces opportunities, links conversations to people and organisations, and creates “intentions” which are structured representations of what needs to happen next. This means that instead of asking AI ad hoc questions, everything becomes anchored in purpose.

What AI Orchestration Really Means

Marion isn’t a single tool or model. She is an orchestration layer that brings together different systems, data sources and workflows into a single, coherent view. At her core, she connects four things: intent, process, people and technology.

Without orchestration, AI quickly becomes fragmented. Different teams use different tools in different ways, data sits in silos, and context gets lost. Even the most advanced systems such as CRMs, email tools and AI assistants will only ever show you part of the picture. What Marion does differently is aggregate and connect that context. She pulls together information from emails, calendars, conversations, CRM systems, transcripts and more, and links it to people, projects and outcomes.

Most crucially, she understands how all of these things relate to each other. A conversation is more than text, it’s tied to a person, an organisation, a project sitting within the CRM and often, subsequently, a commercial opportunity. Actions are not simply tasks, but sit within a wider “intentions tree” that reflects what you’re actually trying to achieve.

In that sense, Marion behaves less like a tool and more like an executive assistant. One that can synthesise huge amounts of information, retain context over time, and guide your attention towards what actually matters.

Changing How Work Happens at SARD

The impact of this approach is already visible in how we work at SARD. One of the biggest shifts is in how we deal with complexity. In most organisations, information is spread across emails, messaging platforms, documents and systems, and people are left to piece it together themselves. There’s a natural limit to how much any of us can take in and retain.

Marion removes that burden by aggregating and synthesising that information automatically. She surfaces what matters, links related pieces together, and makes sure nothing important falls through the gaps.

You see this really clearly in software development. Large projects like, our workforce planning integration engine, SARD One, require constant awareness of what we’re trying to build, ho different components fit together, and what needs to happen next.

With Marion, that context is always there. High-level goals sit at the top of an intentions tree, with sub-tasks, dependencies and related work structured underneath. When you’re working on something specific, it already understands how that fits into the bigger picture, so you’re not constantly having to reload that context over and over.

The result is not just speed; it’s also a shift in how work feels. Less time spent reconstructing information or managing systems, and more time spent thinking, deciding and creating. And it’s not just development. The same pattern applies across consulting, customer support and sales and anywhere that people are spending time pulling information together rather than using it. Rather than replacing people, AI strips away the bits we’re worst at and leaves us to focus on the parts we’re best at.

Trust, Guardrails and the Future of AI Adoption

None of this works without trust and in my experience, trust comes from constraints, not from removing them. Marion is designed with guardrails built in. She knows where to access data, how to use it, and where human oversight is needed. Her purpose is not to replacement judgement; she is actually there to support it.

This matters because the biggest challenge with AI isn’t capability, it’s confidence. The technology is moving quickly, but most organisations aren’t set up to absorb it in a controlled way. Without structure, adoption becomes inconsistent, and people lose trust in the outputs.