Is AI changing the way we work?

Six rules for using AI at work responsibly.

There is no shortage of opinions about AI. It’s going to transform everything. It’s going to take everyone’s jobs. It will make us more productive, but it will also stop us thinking and make us lazy. It’s an extraordinary creative tool, and also the death of creativity.

The pros and cons of AI have been discussed earnestly and endlessly. Rather than add to that noise, we decided to do something different, knowing that even within our small team, there are differences in the way we feel about AI and how we use it. So, we sat down together, recorded the conversation and talked freely to see whether we could come up with ‘our rules for working with AI’.  From our different perspectives, as creatives, writers and marketers. What we like about it, what makes us uncomfortable and where we think the human bit still matters.

We were hoping for a bit of disagreement. Chantal, Head of Brand and Creative Strategy, was, in her own words, quite keen on “fisticuffs”. No chairs were thrown.

A sceptic, an enthusiast and a frazzled project-manager walked into a bar…

We did not all adopt AI with the same energy. Chris, Campaign and Client Manager, described himself as “a dinosaur” when it comes to new technology and admitted that, when ChatGPT was first shown to him in a pub, he largely dismissed it. “I thought the human brain was far more capable and far stronger than anything machines could produce.”

That changed once he started using it properly.

Not because AI suddenly became cleverer than him (he questions that), but because he found a different role for it. “It feels like a critical friend.” For Chris, AI is a sounding board. Put in an idea, some half-formed thoughts or a different perspective you want to test and it gives you something back to challenge.

Katherine, Head of Communications, sees it more like extra storage. “It gives my brain more capacity. Like plugging in an external hard drive.” Complex projects create a lot of stuff. Meetings. Conversations. Decisions. Background information. Previous thinking. Retaining every moving part in your head is impossible. AI doesn’t just help hold it. It organises it and sees connections that could easily be missed or forgotten.

And for Chantal? “AI helps structure the hell out of me.” Perhaps the least technical explanation of AI-assisted cognitive processing, but an important one. Because one of AI’s most useful jobs is not necessarily doing the thinking. It’s the ability it has to take the thinking spilling out of a human brain, and help make it usable.

That “tidying up” is incredibly satisfying. But what if the mess is the point?

What if something valuable was hiding in the mess? Katherine talked about the remarkably clear meeting summaries AI can create in seconds. Chantal immediately saw the catch. “Sometimes the waffle is the creative moment.”

Human thinking is not particularly efficient. Conversations wander. Somebody makes an irrelevant observation. Someone else connects it to something apparently unrelated. You go off on a tangent, come back again and suddenly there is an idea sitting on the table that nobody could have put on the agenda beforehand. Bingo bongo, creativity!

AI is very good at deciding what appears to be relevant. But creativity is often ignited when humans wander off topic.

That gave us our first useful rule:

01 | Use AI to organise the mess. Don’t let it decide what matters.

More capacity. More time?

This led us to the biggest question: If AI saves us time, why aren’t we all spending Thursday afternoons with our feet up because Gemini, Claude or ChatGPT finished the work early. Katherine summed it up perfectly: “We’re still on the hamster wheel. We’re just going twice as fast.”

There is an enormous amount of debate about AI and productivity. Tasks that once took an hour can now take ten minutes. Research can be accelerated. First drafts can appear in seconds. Information can be processed at extraordinary speed. Yet we’re not sitting back and relaxing. We’re doing more.

This isn’t necessarily a bad thing. Faster medical research, quicker problem solving and the ability to remove hours of administration can have huge benefits. But productivity is not the same as value. The real benefit of saving four hours depends on what you choose to do with that time instead.

More thinking? More conversations? Better creative work? More time understanding the problem before rushing towards the solution? Or just another four hours of output?

02 | Let AI create capacity. You decide where to spend it.

The right answer or the most trodden path?

Then Katherine mentioned using Gemini to plan a holiday. A perfectly sensible use of AI. Tell it where you are going, what you enjoy and how much time you have and it can do hours of research in seconds. Chantal, was quick to question the value. “Isn’t it just showing you the most trodden, most predictable path?” And suddenly we were no longer talking about holidays.

Ask AI for campaign ideas, headlines, positioning, imagery or content topics and it has an extraordinary body of existing material to draw upon. Which is both its strength and its weakness. You ask for ideas, but what you actually get is statistical probability. What people commonly do. What things normally look like. Familiar language. Conventional structure. In other words, a balanced, polished, perfectly competent, perfectly sensible and perfectly familiar answer. Sometimes following the crowd and taking the sensible route is what you need. Sometimes you need differentiation. Standing out doesn’t mean following the crowds. It means taking the turning three miles earlier that everybody else ignored.

03 | Don’t assume that convention is always the right answer. Ask whether the solution works for your organisation, your audience and this moment.

When making things becomes easy, choosing becomes harder.

The discussion became particularly lively when we got onto creative work. If AI can create competent copy and imagery in seconds, what happens to the people who have spent years learning how to create those things?

This sent a little bristle of professional pride around the table ; ) Tools can give people access to production. They do not automatically give people judgement.

Chantal observed that new designers sometimes use every tool in the box just because you can. Every Photoshop trick, every effect, every element. Experience teaches the opposite. “The skill is in taking stuff away.” That feels increasingly important in an AI-enabled world. When generating another ten ideas costs almost nothing, the valuable skill is not necessarily generating idea number eleven. It’s knowing that idea four has something. It’s confidence to bin off idea eight. Or combining two to make one new brilliant idea. AI makes creation abundant. That makes selection, restraint and judgement more valuable.

04 | Generate freely. Edit ruthlessly.

Does it really matter if AI made it?

This was the hottest debate. If somebody uses AI to create a perfectly functional piece of communication, should we care? Chantal’s position was fairly pragmatic. “It’s just another tool. It’s about how you use that tool.” “A badly designed communication is badly designed whether somebody spent six hours lovingly crafting it or generated it in six seconds”. On the flip side, if an organisation with very little budget can suddenly create something clear, useful and effective, then why deny the opportunity?

There is another consideration. People notice. And if something feels generic, artificial or careless, that can influence what they think about the organisation behind it.

So perhaps asking “Was AI used?” is the wrong question. Better questions might be:

• Does it do the job?
• Does it feel like us?
• Is it true?
• Is it distinctive enough?
• Has somebody with the right judgement taken responsibility for it?

Maybe the tool matters less than the thinking surrounding it.

05 | Judge the work, not the tool. Own the outcome.

Advice is not the same as handing over the keys.

There was one point where our comfort levels changed quite noticeably. We’re all pretty happy allowing AI to offer ideas, organise information, challenge our thinking and suggest what might happen next. Handing over the keys is different.

Chris described using an automated service where the lack of human intervention meant that, once the process went wrong, you reached a dead end with no friendly face to provide reassurance. Asking AI “What do you think?” still leaves you with the decision. Asking an AI agent to make the decision, purchase the thing, book the trip or complete an entire process loses you something valuable – control.

Chris has clear boundaries. He likes AI for work but an Alexa in an Airbnb was physically unplugged. “I can bring it out when I need it, but I’d rather it didn’t become a part of my life that I hadn’t invited in.” Chantal, meanwhile, would be first in line to purchase a fully functioning household robot if the price was right.

06 | Hand over tasks, not accountability. And always leave a way back to a human.

Maybe disagreeing was the point.

So, no, we did not always agree. No chairs were thrown. Probably a good thing, on both counts. But isn’t that the joy of being human? We are not robots, developed by taking masses of data and forming it into a generic human form, based on statistical probability. Individually, we have applied our enthusiasm, scepticism, curiosity and professional pride and found a way to use AI. We remain excited about what AI could make possible and have a touch of discomfort about where it might go – we will always be polite towards ChatGPT, just in case.

Perhaps you’re the enthusiast saying: what if we could? Or the sceptic saying: should we? Or the pragmatist, asking: is it actually any good? Somewhere between those three personalities lies a fairly sensible way to use AI. Avoiding it isn’t the answer. Nor is trying to put this particular genie back in the bottle. The best advice seems to be: don’t hand over every task simply because a machine can now perform it.

We started with some questions about AI. We finished, well over an hour later, talking about creativity, trust, loneliness, productivity, independence, science fiction, horses, cars, robots with googly eyes and whether being nice to ChatGPT might eventually save us when our new overlords arrive. Which, thinking about it, rather proves the point about not editing out the waffle.

AI has made producing things easier. It hasn’t made knowing what is worth producing easier.

P.S we call our ChatGPT Champagne SuperNova, Nova for short, and her pronouns are she/her/overlord. She helped structure this blog, and gosh, aren’t we thankful she organised our spaghetti brains. Which feels appropriate really. Although possibly that’s exactly what she wants us to think…