1. Introduction
We’re at the Emirates Stadium, sitting pitch side with cameras, lights, and an absolutely massive LED screen looping AI content above our heads. This episode the Impact of AI: Explored podcast is on tour and we have landed at the Hello AI Collective’s AI summit (https://www.helloaisummit.co.uk/)
Eleven months ago, we recorded with Dan just as “Hello AI” was an idea on a slide deck; today, his brand is literally wrapped around the stadium for the Hello AI Collective Summit, sponsored by Microsoft and Lenovo.
In this episode, we talk about what has changed in AI since that first conversation, and more importantly, what actually works when you try to apply AI inside real businesses. We dig into data foundations, the myth of “time saved” as ROI, operational AI agents, and why customer experience is still where the biggest value lies.
If you’ve been wondering how to move from “AI is a cool demo” to “AI is a core part of how our business runs,” this is for you.
2. Meet the Guest
Our guest is Dan Boyles, founder and CEO of Hello AI, the company behind the Hello AI Collective Summit at Arsenal’s Emirates Stadium. When we first spoke with Dan, Hello AI hadn’t even launched; it was an idea for a new kind of AI company focused on outcomes instead of tools. Less than a year later, he’s hosting a full day summit with eight speakers and a stadium full of people there for one reason: to figure out what AI actually means for their organisations.
Hello AI doesn’t sell a fixed product or a specific tech stack; Dan’s team sits in the messy middle between business strategy, data, and AI, helping organisations of all sizes—from two person companies to enterprises with thousands of employees—build AI operations that are grounded in their own data. Their tagline, “reason, solution, scale,” sums up the philosophy: start with the reason, design the solution, then scale responsibly instead of chasing shiny tools.
3. Setting the Stage
We’re at a point where AI has moved from novelty to expectation: people are using tools like Copilot, Claude, and ChatGPT every day, sometimes with or without IT’s blessing. That creates a strange tension: leaders know AI “should” be important, but they struggle to pin down what success looks like beyond vague time savings.
In this conversation, we wanted to cut through the noise and talk about what actually matters if you’re serious about AI in your business today. From data strategy and governance to customer experience and operational AI agents, our goal in this blog is to give you a grounded view you can use as a checklist: are we fixing real problems, or just adding more tools?
4. Episode Highlights
Highlight 1: “Time saved is not ROI”
Dan is brutally honest about one of the most common mistakes companies make with AI: they treat “saving two hours a week” as a business case. As he puts it, that’s not ROI—because we immediately fill that time with something else, often without a clear link to revenue, cost reduction, or risk mitigation.
Instead, he pushes organisations to ask harder questions:
- Does this AI project increase profits by a measurable percentage?
- Does it allow us to avoid a hire or consolidate multiple outsourced contracts?
- Does it fundamentally improve a key process like sales, support, or operations?
The message: AI is not a free pass to ignore the basic rules of business cases we’ve used for decades.
Highlight 2: “Sort your data, then nothing is off the table”
The other major theme is data. Dan is clear that most AI projects don’t fail because of the model; they fail because the data is fragmented, messy, and inaccessible. He describes the biggest wins coming from organisations that bite the bullet and centralise their data into something like Azure SQL and Fabric, bringing together on prem systems, Citrix environments, and random OneDrive or desktop files into one governed layer.
Once that’s done, “nothing is off the table”: you can put OpenAI style models, Azure OpenAI, Copilot, or custom GPTs on top of that foundation and actually trust the outcomes. But you don’t get the magic until you invest in the boring data work.
5. Deep Dive: Data, Risk, and Operational AI
If there’s one big idea running through the entire conversation, it’s this: AI is just software, and all the usual rules of good IT and business still apply.
Data as the real AI platform
Dan repeatedly comes back to the same point: stop treating AI like a glorified Google search and start building it on top of a robust data platform. In practice, that means:
Consolidating data from on prem apps, desktops, OneDrive, Citrix, and SaaS into a central store such as Azure SQL, then ingesting and modelling it with tools like Fabric.
Using that central layer as the single source of truth for any AI agent, chatbot, or Copilot style experience you build.
Applying all your usual governance, DR, retention, and security controls there instead of sprinkling AI directly over production systems.
Do that, and suddenly you can plug in different models (OpenAI, Azure OpenAI, Anthropic, custom LLMs) without re architecting everything each time the market shifts.
“Don’t do an Ultron, do a Jarvis”
When we push on the risk side—like Amazon’s infamous agent deleting production resources—Dan’s response is very simple: “don’t be an idiot.” You wouldn’t give a brand new, untrained junior unrestricted access to production systems; why would you grant that to an AI agent?
The analogy he uses is straight out of the Marvel universe: Ultron is what happens when you let an unchecked system roam free; Jarvis is what happens when you design guardrails and a well defined operational role.
The practical guidance:
Never put unrestricted AI directly into production environments.
Treat AI as a very clever intern at first, not an autonomous executive.
Define what it can and cannot do; log, monitor, and test changes before scale.
From AI agents to AI operations
In our previous episode, we talked about AI agents mostly as advanced automations. This time, Dan’s view has evolved: he sees the future in “AI operations”—AI embedded into the operational fabric of the business, handling repetitive, rules based work across systems.
Will it replace hosting a podcast or strategic consulting? Probably not. But will it replace chunks of operational consulting, repetitive analysis, and routine process execution? Very likely, especially as tools like MCP and multi tool orchestration mature.
6. Real-Life Stories and Examples
A stadium full of AI use cases
The Emirates stadium setting isn’t just a nice backdrop; it’s a symbol of how fast this space has moved. When we last spoke, Hello AI was pre launch; now, there’s a full day summit with sponsors like Microsoft and Lenovo, eight speakers, and a crowd that ranges from tiny two person firms to huge enterprises.
Dan’s point: in the last 9–11 months, there hasn’t been a single sector where he couldn’t apply AI in some meaningful way. The pattern is always the same—start with a specific outcome (more revenue, less cost, better fan engagement), then work backwards into data and operations.
Avengers themed AI agents
One of the lighter moments in the episode is when we talk about how people name their AI agents. Dan admits he has multiple “Avengers” running: Jarvis, Thor, and others, each with different responsibilities. Underneath the humour is a serious point: these agents are becoming part of people’s workflows and identities, not just one off scripts.
He even jokes about having an Iron Man helmet in the car, ready to bring onto the pitch for a photo—because if you’re going to talk about Jarvis and Ultron all day, you might as well commit.
Customer experience over shiny tools
A recurring theme is customer and fan engagement, especially in industries like sports. Dan argues that businesses are under investing in this compared to clubs and brands that are obsessive about fan experience.
He describes a simple lens:
From the moment someone reaches out, how easy is it for them to get what they want?
Does AI make that experience faster, more consistent, and more professional?
For Hello AI, that means aiming for the same great experience whether a prospect has 10 pounds, 1,000 pounds, or a million pounds to spend. AI helps do the legwork so the team can stay focused, but the human experience is still non negotiable.
The love-hate relationship with AI support
We also get into the very human side of AI customer service. Most of us don’t mind a bot when we’re dealing with something transactional—like checking why a payment failed or confirming broadband issues—as long as it’s fast and accurate.
But when an AI tries too hard to be “personal” (“Hey, Daniel!”), it can feel hollow. That’s where the balance matters: use AI to deliver quick, reliable data, and keep humans in the loop when nuance, trust, or empathy is required.
7. Key Takeaways
- Data first, tools second: Centralise and clean your data before you obsess over which model or vendor to use.
- “Time saved” isn’t enough: Build AI business cases around revenue, cost, or risk—not vague productivity gains.
- AI is still just software: All your existing practices—governance, DR, change control, access management—still apply.
- Don’t build Ultron: Start with constrained, well defined roles for AI agents and expand as you gain trust and telemetry.
- Think AI operations, not just agents: The biggest near term impact will be in operational, repeatable processes across the business.
- Customer experience is the real battleground: Use AI to show up consistently and professionally across every touchpoint, without losing the human element.
- Treat AI like a clever intern, then a colleague: Give it structured tasks, monitor outcomes, and gradually increase responsibility as guardrails prove themselves.
8. Closing Thoughts
Recording this second episode with Dan at the Emirates felt like a full circle moment: from “we might start a company” to a packed summit with Hello AI branding wrapped around a Premier League stadium. It’s a reminder of how quickly AI is moving and how important it is to stay grounded in fundamentals like data, governance, and real business outcomes.
If there’s one question we’d love you to reflect on after reading this, it’s this: What is the actual business problem you want AI to solve, and is your data ready for it?
We’ll continue exploring this in upcoming episodes, including deeper dives into AI operations, MCP style multi agent ecosystems, and how to design AI experiences that customers actually want to use. If you’ve got stories—good or bad—about AI in production, we’d love to hear them and maybe feature them in a future show.
