Scaling AI Beyond the Pilot – Thoughts from Manchester Digital Panel

Every day I help organisations implement AI in the real world - from proving concepts and piloting ideas, through to scaling solutions that deliver measurable outcomes.

The other week I joined a panel – AI Integration: What Happens After the Pilot? – at MD Future 26 by Manchester Digital. It felt fitting to be having that conversation in Manchester, where Robiquity was founded ten years ago and remains headquartered today. In this time, Manchester has become one of the UK's best hubs for AI and automation talent, and events like this are a reminder of how much of that innovation is being shaped by people, not just platforms.

What Implementation Looks Like on the Ground

My fellow panellists and I discussed how technology creates potential. But it is people, processes and operating models turn that potential into value.

On the ground, AI implementation looks a lot like any other successful transformation programme. You start with a problem, prove the value through a pilot, and then scale what works.

On a project I led for our client, Evri, we started with the challenge of manually monitoring parcel delivery compliance at a scale humans simply couldn't achieve. In this instance, AI was the best solution. The model we built scans some 90 million delivery pictures each month, creating full visibility for the entire network and generating an additional £1 million in revenue since its launch 8 months ago.

What made this project with Evri so successful was treating AI as a product, not a project. We had clear business ownership, agile ways of working and a strong focus on adoption as we scaled. Solving a problem was the goal, not the adoption of AI.

Taking AI from Proof of Concept into Production

I think it's helpful to separate proof of concept, pilot and scaled production. They are often treated as the same thing.

·       A proof of concept proves the technology can work

·       A pilot proves it can work with real users and real data

·       Production proves it can deliver value reliably at scale

Each stage brings distinct challenges, and the work between them must be planned from the start. Moving from idea to pilot means tackling a real business problem with real data and real users. Taking a pilot to everyday use is harder: people must trust it, ownership must be clear, processes and controls need to be in place, and value measured. Many organisations stall here. Proving the tech works isn’t enough; scaling demands, changes in people, processes and ways of working. The winners stay focused on the problem and the outcomes and are willing to learn and adapt.

Organisations are too often satisfied with a good pilot being their end game rather than the first step. Those who get it right stick to the problem they want to solve and track their business results rather than just the AI model.

Agentic AI: the Human Opportunity

The exciting part of our projects isn't using the agentic AI technology itself. It's what it enables people to focus on. Unlike traditional AI models that merely generate outputs for humans, agentic AI can autonomously pursue goals, plan actions, use tools and adapt based on feedback.

During the panel, panellists described the current landscape with a mix of excitement and uncertainty. While some employees are embracing AI, others are concerned about how it might impact their roles, making change management and workforce engagement critical to success.

This is what I find so exciting about agentic AI. Agentic AI is not just about providing better solutions but also reducing the administrative and coordination overhead around professionals, so that they can concentrate on the aspects of their profession where human expertise, intuition and empathy are required and most valuable. As the tech becomes more autonomous, the less a human needs to be involved in the admin side of things.

Agentic AI creates an enormous opportunity but also brings with it enormous responsibility. As the amount of autonomy grows, the need for trust, transparency and governance becomes increasingly crucial. The future is not about AI substituting human beings, but about AI helping human beings to concentrate on what matters. Maintaining this approach requires effective governance, safeguards and government regulation.

Lessons from Delivering Transformation

The fundamentals stay consistent, whatever the sector. Start with a genuine business problem, focus on outcomes rather than technology, design for scale and invest in your people as much as the platforms you choose.

The best transformations I've seen have never been about AI alone. They've been about people, process and technology coming together. Before you start, ask yourself: is this the right problem for AI, is there the right information behind it and is there the right fallback if it doesn't work?

Closing Thoughts

All of us on the panel could agree on one thing - AI doesn't create value just because it's deployed. It creates value when people change the way they work.

Thanks to Manchester Digital for hosting such a thoughtful discussion, the questions from the audience and everyone who came up to chat afterwards.

Ready to discuss where your organisation stands in its path from pilot to production? Get in touch with our team today to see how Robiquity can support your journey to scale impactful AI solutions.