AI may be the best thing to happen to agency work

I started using AI and saw the value immediately

I'll start with a disclaimer - this post wasn't written by AI.

I've been using AI extensively over the past few months. Why? Because I firmly believe computers should do as much work as possible, leaving us humans to check, edit, make decisions and ultimately - do less work.

At a more fundamental level, in an industry plagued by a lack of regulation and transparency, AI frequently outperforms agency folk, delivering results with greater speed and efficiency. Many in the field still operate on intuition or 'flying by the seat of their pants', so when a technology like Chat-GPT arrives on to the scene, I celebrate a disruptive entrant - the industry needs it!

For those that still want to pursue a job in agency land, I promise you, once you start using a tool like Chat-GPT, you'll never go back. It simply makes your job easier and your results better.

I'll share more about some of the awesome examples I've experienced over the coming months, but for now, I want to address how things haven't really changed in the past 5 or so years.

State of the agency

I've been fortunate enough to contract for a few agencies this past year (and if you're a subscriber to my newsletter, you'll know why). Most have reminded me what it used to be like when I was Head of Design Thinking at an agency, and how, over a number of years, changed how we work for the better. The current crop of valuable agencies haven't changed, they've just scaled.

This is a sweeping generalisation, but what I see in product strategy and design are:

• too many people working on the project (8, when there could be 2 or 3)
• estimates based on days, rather than following a roadmap of a few sprints
• a complete mess of project management tools with Slack, Confluence, Google docs, Slides and even email on top (instead of say a Notion one pager with clear goals, roles, deliverables and a timeline)
• the above point means that no one can find anything at any time
• time taken to make documents and decks to justify high fees and time spent
• no post-deliverable support
• bizarrely, the agency isn't gaining the full lifetime value of the client

So yes, a bunch of what I see as problems. However, they seem to be present in most UK top-tier agencies and I suppose it's because no one is looking at the whole. There's an incredible opportunity to deliver an alternative, and that's exactly what we're aiming to do with Skysoclear.

Whilst a lot of these things can be just a general tool and process problem, the amount of time and the quality of work can be greatly improved with AI.

Being agile to the need

Let me share a story. Within the second week of a website redesign project, we hit a problem. The designer needed to understand the value proposition of the client to put together a landing page wireframe, and because we were limited to an MVP with only a few weeks left, no time or budget had been allocated to running a value proposition workshop.

So what did I do? I used Chat-GPT to ask how I could do value proposition. It gave me some set parameters and explained what it was. Simply though, it stated that we only needed a few things:

• a headline
• a subheadline
• 3 or 4 benefits
• imagery

Great I thought. I'll take 20 minutes of the next check-in and we'll run a workshop on it.

And this is where I realised the value of AI....

Happy at the prospect that this could be done quicker and better, I referenced the current website and asked it to create a value proposition for it. Within seconds, it gave me it, in that same format.

Why was this a useful exercise? Well, essentially, clients currently look to agencies for expertise and advice. You can run a workshop for the client to fill in their own value proposition and up vote the best suggestions, but that's led by several biases, the knowledge and commitment of the people in the room and ultimately, it's subject to subjectivity. 'Innovation theatre' some could say.

Sharing the pre-filled value proposition and allowing the team to silently note take, share their feedback and edit what Chat-GPT had given us, was in my view, more informed and more valuable. We edited a couple of words to give it more brand direction and committed to it. Value proposition done. Next!

This example was using just plain ol' Chat-GPT 4. No plugins, no workflow - just one prompt, one 20 minute activity and we were done, and anyway, they could adjust it later if they really wanted to.

Removing the busywork

As part of client work, you would probably have received or commissioned pages upon pages of market research, survey data etc. Data-driven research is more popular than ever, but no one is going to read a 40-page document or be able to share the key highlights. Wouldn't it be great if you could just get the key learnings in a bullet-list type format within a few seconds?

Enter Chat-GPT again.

This time, you'll need a pdf plugin, and within a few seconds, you'll receive a list of the highlights. You've removed the hours-long busy work, you've used technology to analyse and create a bitesized summary, and especially if you serve clients, you've created time to get ahead of the task list or take an early lunch break!

These are just the early days, but I've already seen examples of UX audits that look at screen-grabbed landing pages and give a one-page summary of areas to improve, workflows that can create templated documents (think Canva and Chat-GPT combined) and I'm not even talking about production (I'll save that for another post).

What agencies need to do now

First, do what I do and start using it, even Chat-GPT 3. Once you've had a few goes, you'll understand what I've been talking about here.

Then, decide as an agency if you want to tell your clients that you are using AI or not. I suspect most won't, to preserve integrity, status-quo and to justify their fees. They may even increase their margins.

But I think AI for agencies is going to increase transparency, consolidate a number of time-consuming human processes down to a few seconds, and in some cases, it's already doing a better job.

Do you agree? Let me know on X.