What businesses are learning as AI becomes part of everyday work

This month, we brought back the New Icon Innovation Breakfast.

We invited a small group of people from across different sectors to join us at our Bristol office for coffee, croissants and an informal conversation about innovation and AI.

There were no presentations, pitches or planned conclusions. The idea was simply to get people with different experiences around the same table and talk openly about what they're seeing: what's working, what isn't, where the opportunities are and what they're still trying to figure out - and there was plenty to talk about!

The conversation moved from AI agents running parts of day-to-day operations to changing customer expectations, the future of junior roles, vendor dependency, human judgement and whether businesses are moving so quickly that they risk losing sight of what they were trying to achieve in the first place.

Here are some of the things we took away from the morning.

Using AI isn't the same as changing how your business works

Most businesses have experimented with AI by now.

But there's a big difference between using ChatGPT or Copilot for individual tasks and embedding AI into the way a business actually operates.

Around the table, we heard examples of AI being used to prioritise workloads, manage diaries and inboxes, analyse marketing performance, support outbound activity, create proposals and help review large amounts of information.

That's a long way from asking ChatGPT to rewrite an email.

But one point came up repeatedly: before you can successfully introduce AI into a process, you need to understand the process itself.

What actually happens? In what order? Where are decisions made? What information is needed? Where does someone's experience or judgement come into it?

If you can't map the workflow without AI, adding AI isn't necessarily going to fix it.

In fact, you might just end up doing the wrong thing faster.

The boring work might be the important bit

There's understandably a lot of excitement around what the latest models can do. But some of the most important work involved in implementing AI isn't particularly exciting.

It's documenting processes. Structuring data. Understanding workflows. Deciding where responsibility sits. Establishing what happens when the AI gets something wrong.

One attendee talked about the advantage of already having heavily documented standard operating procedures before introducing AI. That groundwork made it considerably easier to identify where AI could take on parts of the process.

It's a useful reminder that successful AI adoption isn't just about having access to the best model.

The quality of what sits around it matters just as much.

AI is changing what expertise looks like

The changing nature of work became one of the biggest conversations of the morning.

Across creative work, marketing, technology, professional services and other knowledge-based roles, AI can now complete tasks that previously took significant time or specialist resource.

That changes the economics of delivering that work.

A smaller number of experienced people, supported by AI, can potentially achieve what previously required a much larger team.

For businesses, there are obvious advantages. More capacity. Lower costs. Faster delivery. The ability to spend more time on work that creates greater value.

But it also led us to a much harder question.

If AI does the junior work, how do people become senior?

This was one of the most interesting questions raised during the session.

Experienced people are often very good at using AI because they know enough to recognise when it's wrong.

They've spent years doing the work themselves. They've seen what works and what doesn't. They've made mistakes. They've developed judgement and instinct that isn't always easy to document.

But much of that experience was built by doing exactly the kind of work we're now starting to automate.

If junior lawyers don't review documents, junior marketers don't build campaigns, junior developers don't work through problems and junior creatives aren't producing and refining work, how do they develop the judgement they'll need later in their careers?

There isn't an obvious answer.

It may mean businesses need to think differently about apprenticeships, mentoring and training. It may change what entry-level roles look like entirely.

What seems clear is that removing junior tasks doesn't remove the need to develop senior knowledge.

Producing something convincing has become incredibly easy

AI is very good at sounding like it knows what it's talking about.

It can produce a polished strategy, report, analysis or recommendation in seconds. And if you don't know the subject particularly well yourself, it can be difficult to recognise what's missing.

That creates an interesting shift in where value sits. Producing the first version of something is becoming easier - knowing whether that version is actually any good isn't.

Several people around the table came back to the importance of experience and context. Someone with years of knowledge in a particular field can challenge an AI response, spot a questionable assumption or recognise that an answer simply doesn't work in the real world.

As the barrier to producing work gets lower, judgement could become one of the most important skills we have.

Everything is getting faster. Should everything get faster?

Speed came up again and again. AI can shorten the distance between an idea and execution dramatically.

You can generate options, test them, analyse the results and change direction much faster than before. In some situations, that's hugely valuable.

But there's another side to it.

When new analysis and recommendations can be generated constantly, businesses can find themselves responding constantly too.

There's always another piece of information. Another recommendation. Another way of approaching the problem.

One comment during the morning summed up the risk well: if the strategy is right but the plan isn't working, change the plan. Don't keep changing the strategy simply because there's new data available.

AI can help businesses move faster, but knowing when not to change direction may become just as important.

What happens when your business depends on one AI?

Another practical concern was dependency.

If an AI model becomes embedded across your workflows, what happens if its pricing changes? Usage limits change? The model changes? Or another provider suddenly becomes significantly better?

It's easy to see similarities with previous conversations around cloud vendor lock-in, except AI can become intertwined with a much broader range of day-to-day business activity.

One approach discussed was to avoid thinking in terms of being an “OpenAI business” or a “Claude business” altogether.

Different models have different strengths. One might be better for a particular form of analysis, another for written communication and another for image generation.

Building around the capability you need rather than a particular provider can give businesses more flexibility as the technology continues to change.

For organisations handling sensitive data, there are further questions around security, compliance and where information is being processed.

The more important AI becomes to an organisation, the more important those questions become too.

Replace, augment or do something you couldn't do before?

A useful distinction in the conversation was thinking about the role AI is actually supposed to play. Sometimes the goal is replacement. There are tasks that can be automated entirely, particularly where businesses need to reduce costs or remove repetitive work.

Sometimes it's augmentation: helping someone do their existing job faster or better.

But perhaps the more interesting opportunity is using AI to give people capabilities they simply didn't have before.

That could mean analysing information at a scale that wasn't previously practical, giving someone access to expertise or insight they didn't previously have, or allowing a business to tackle a problem it couldn't justify tackling before.

Rather than starting with “What can we automate?”, businesses might get further by asking: What can we do now that we couldn't do before?

Could AI actually make businesses more human?

For a conversation that spent a fair amount of time discussing jobs being automated, one of the more optimistic ideas of the morning was that AI could give people more time to be human.

One example shared involved AI saving someone around 15 hours every week on a repetitive part of their role.

The interesting question isn't only what that saves the business.

It's what that person can now do with those 15 hours.

There will always be a temptation to fill that capacity with more output. But it could also create more time for conversations, relationships, problem-solving and the work that repeatedly gets pushed aside because people are too busy.

For businesses where trust and relationships matter, removing people from the process entirely may not even be desirable.

People still want to work with people.

And as AI-generated content, communication and experiences become increasingly common, genuine human interaction may end up becoming more valuable rather than less.

We're still figuring this out

We didn't leave breakfast with all the answers - that wasn't really the point.

What we did leave with was a much wider view of the questions businesses are starting to grapple with.

AI isn't just a technology conversation anymore.

It's a conversation about how businesses operate, how we develop people, what customers value, where expertise comes from, how quickly we should move and which parts of work we want to remain distinctly human.

The technology will continue to change. Probably faster than any of us can comfortably keep up with.

The challenge for businesses is making sure that while the tools change, they're still clear on the problem they're trying to solve.

Thanks to everyone who joined us and contributed to the conversation. We'll be back next month for another Innovation Breakfast and another topic to get stuck into.