We gave everyone in our company a secret mission

Optimation Group's AI Hackathon

There is a version of the AI adoption story playing out quietly in many organisations right now. The technical team builds. Everyone else watches. A two-speed organisation takes shape, not by design, but by default.

The people with the most valuable instincts for where AI can help are not always the ones who write the code. They are the ones who know the workflows. Who live the problems daily. Who understand, better than anyone, where time and energy quietly disappear in a given process. That knowledge exists throughout any organisation, and when you give it the right tools and the right conditions, the results tend to surprise people.

So earlier this year, we created those conditions. We closed the laptops on business-as-usual for a day and ran Optimation's first company-wide AI Hackathon.

Why a Hackathon?

We could have run workshops, assigned online courses, or set up a lunch-and-learn series. All of those things have their place. But none of them do what a hackathon does.

A hackathon compresses the learning curve because the stakes are real and the timeline is short. You aren't exploring AI in the abstract. You are building something specific, for a real problem, that you will present to your colleagues by 2pm. That constraint changes everything. It creates conditions for genuine curiosity, for trying something you might otherwise have talked yourself out of.

There is also something that happens when people work together under a little pressure and a lot of creative licence. The conversations get better. The questions get braver. And when it's working well, people stop noticing that they are learning.

That's what we were after. Not a training event. An experience that changes how people see their own potential.

Before the day: getting the environment right

This part does not always get talked about, but it matters as much as the day itself.

Before we asked anyone to build anything, we made sure people were building in an environment that was genuinely safe, secure, and set up for success. We chose Claude as our AI platform, configured our connectors, and locked things down appropriately. People needed to know that what they were doing was encouraged, that it was secure, and that they could push and experiment without accidentally breaking something important.

That psychological safety, the sense that it is okay to try things here, is the prerequisite for everything else. Think of it like a test kitchen. You need to trust the environment before you can be free in it. We did that groundwork upfront, quietly, so that on the day itself the only thing people had to focus on was building.

We also made sure everyone had at least a baseline familiarity with the tools before they started. Not expertise. Familiarity. Enough to know where to begin. The goal was to remove friction, not create it.

Designing the day

A few deliberate choices made the day work. We would make each of them again.

Team composition was intentional. We did not let people self-select into groups of close colleagues. Teams were mixed across two dimensions: AI proficiency and domain depth. You want people who can move fast with the tools sitting alongside people who have deep, specific knowledge of a business problem. Neither alone produces the best result. The person who knows a workflow cold but is new to AI is extraordinary at spotting where an agent would actually add value. The person who has been working with AI for months is extraordinary at making it real quickly. Together, they are formidable.

The theme was doing real work. Secret Agent was not just a fun wrapper, though it was definitely that. The theme was chosen to create a particular kind of energy. When people are having fun, they are more daring and more curious. They try things they might dismiss as too simple or too ambitious in a more formal setting. That willingness to have a go, especially at the start of an AI journey, is worth more than any amount of structured instruction.

The brief had clear boundaries. This is the counterintuitive one: more creative freedom does not always produce better results. We gave teams a tight framework for what makes a good agent.

• Be task-based: one repeatable, specific thing done well, not a sprawling ambition.
• Drive real impact: either efficiency (saving time, reducing cost) or effectiveness (improving quality, insight, or customer experience).
• Provide rich context: the best agents need great inputs. Build in an area where you have data and domain knowledge to draw on.
• Solve for what you know: you are the expert in your corner of the business. The best ideas come from exactly there.

That framework was not limiting. It was liberating. It meant teams could stop staring at a blank canvas and start asking the right questions about their actual work.

What happened on the day

Kick-off at 9:30am. Teams formed. The brief went live.

What followed was one of the most energising days we have had as a business. By lunchtime, something had visibly shifted. People were no longer asking what AI can do. They were asking where in my week have I just accepted inefficiency as normal?

At 2pm, teams presented. Awards were on the line: the Speedy Gonzales Efficiency Engine for the biggest productivity boost, the Maverick for the most creative leap, and the People's Choice decided by company vote. The competitive energy was excellent. The costume commitment was variable. The quality of thinking across the board was impressive.

Some of the sharpest, most immediately usable agents came from people across every function of the business: delivery leads, client-facing roles, operations people who know their workflows with a precision that is genuinely hard to replicate. That is not a coincidence. That is what happens when you give domain expertise the right tools and a structured challenge.

By 3:30pm we were celebrating. The conversations over drinks were, in a lot of ways, more valuable than anything built during the day itself, because people were already thinking about what they would do next.

What we learned

1. Doing beats watching, every time.

You can run all the lunch-and-learns you want, but there is no substitute for actually building something. The learning compresses dramatically when there is a deadline, a team, and a little healthy competition.

2. The value is not in doing more work. It is in doing the right work better.

Several people came out of the day talking about their roles differently, about where their time had been going, about what it might feel like to genuinely have capacity back. That is not a small shift. That is the beginning of a different way of working.

3. Fun is a design principle, not a bonus.

The Secret Agent theme, the costumes, the awards, the energy of the day, these were not decoration. They were structural. When people are having fun, their guard comes down, their curiosity goes up, and they try things. For many in our team, this hackathon was their first real attempt at working with AI. The fact that it happened in an environment that felt playful, safe, and collective made all the difference.

4. The skills that matter most are changing.

AI literacy, knowing how to prompt, direct, and tune these tools, is becoming a core competency across every role, not just technical ones. But so is being deeply human. Judgment, empathy, genuine client relationships: these do not get replaced by AI. They get amplified by it, when people have the capacity to actually use them.

Could this work for your organisation?

Yes, and the approach is more transferable than you might think. A few things we would point to as foundational.

• Start with the environment. A secure, well-configured platform for experimentation is not a nice-to-have. It is the prerequisite.
• Mix your teams deliberately. Domain expertise and AI proficiency sitting together is where the interesting things happen.
• Make it fun by design. Playfulness is not decoration. It is how you get people to try things they would otherwise talk themselves out of.
• Keep the brief tight. Constraints are liberating. A focused challenge produces better thinking than an open canvas.
• Lead with the why. Be clear on what good looks like for your people, not just your processes. What would it mean for no one to be left behind in your organisation's AI journey?

If you are somewhere in the space of knowing you need to do something about AI but not being sure where to start, we would genuinely like to talk. No pitch, no pressure. Just a real conversation about where you are and what might actually be useful for your team.

That's how the best missions start.

Get in touch with our team. We are building this in real time too, and are happy to share what we are learning.