Agencies faced this exact choice-paralysis twenty years ago, when digital and social channels exploded overnight. This time, the stakes for getting it wrong are higher.

When I started Gunpowder in 2012, a large part of my work was helping agencies transition from traditional, mostly analogue new business practices to take advantage of the myriad of digital and social channels that had opened up over the previous decade.

The opportunity was real, but so was the overwhelm. Suddenly there wasn’t one obvious route to new business, there were a dozen: a blog, a newsletter, LinkedIn, Twitter, SEO, PPC, content marketing, marketing automation, and the list kept growing every year. For a lot of agency owners, that much choice didn’t feel like opportunity. It felt like a test they hadn’t revised for.

Some froze. They watched competitors build a following on LinkedIn or start ranking for the right keywords, and did nothing, telling themselves they’d get to it once things quietened down. They never did.

Others dabbled. They’d start a blog with real enthusiasm, post three times, then let it go quiet for six months. They’d set up a company Twitter account, follow a few dozen people, and never tweet again. The pattern was almost always the same: a burst of effort with no strategy behind it, no consistency, and no way to tell whether it was actually working. When it didn’t produce a flood of new enquiries within a few weeks, it got filed under “doesn’t work for us” and quietly abandoned.

And so most agencies did what had always worked before: they picked up the phone, or went back to relying on referrals and the odd bit of cold outreach. Not because those channels were secretly better, but because they were familiar, and familiar felt safer than another half-finished experiment.

The same explosion, twenty years on

Twenty years on, we’re living through it again, except this time the list of what’s suddenly possible is longer and moves faster. A handful of prompts to Claude, or one of its peers, can research a market, profile a prospect, and flag genuine buying triggers before a human would have finished their coffee. The same tools can draft outreach emails, automate the sending of them, and generate the case studies, decks, and social content to back them up. Work that used to take a researcher, a copywriter, and the best part of a week can now happen in an afternoon, sometimes in minutes.

For agency owners who spent the last two decades getting comfortable with digital and social, this should feel familiar. It’s the same shape of opportunity: a sudden widening of what’s possible, arriving faster than most people’s capacity to make sense of it. And if the pattern holds, it comes with the same risk of paralysis, the same temptation to dabble without a plan, the same slide back to whatever felt safe and familiar before.

Where the parallel breaks down

But here’s where the parallel breaks down, and where I think this moment deserves more care than the last one did.

Twenty years ago, the worst outcome of dabbling badly was looking a bit amateurish. A half-abandoned blog or a dormant Twitter account was awkward, maybe embarrassing if a prospect happened to stumble across it, but it did little real damage. The mistakes were slow, and mostly private.

The AI-driven equivalent is neither. An automated outreach sequence built on a shaky prospect list, or a batch of AI-generated content that misreads your audience, doesn’t sit quietly waiting to be discovered. It goes out to hundreds of people at once, under your name, before anyone has had the chance to sense-check it. Inaction is still very much on the table too, arguably more so now, because the sheer volume of options and information can produce its own kind of freeze. But this time, doing the wrong thing badly carries a cost that doing nothing never did: the risk isn’t just wasted effort, it’s damaged trust, delivered at scale, in public, faster than you can walk it back.

What actually separated the agencies who got it right

So what actually separated the agencies who came through the last shift well from the ones who didn’t?

It wasn’t which channels they picked, or how early they picked them. The agencies who came out the other side stronger were the ones who got their positioning and message clear before they touched a single channel. They knew who they were for, what made them different, and what they wanted to be known for, and only then did they choose which channels would carry that message to the right people. Everyone else was doing the reverse: picking up a channel because it was new, or because a competitor was on it, and trying to work out what to say once they were already there.

The lesson isn’t really about channels at all. It’s that strategy has to come before tools, because a channel, or a piece of software, will faithfully carry whatever you put into it, good or bad, clear or muddled. Twenty years later, with AI, the mechanism is identical. AI doesn’t hand you a strategy. It doesn’t know your positioning, your ideal client, or what a genuine buying trigger looks like for your business, unless you’ve worked that out first. What it does is amplify whatever you already have. Feed it clarity, and it will help you move faster and further than you could alone. Feed it confusion, or nothing at all, and it will simply help you be confused, at scale, more efficiently than ever before.

AI doesn’t hand you a strategy. What it does is amplify whatever you already have.

Strategy first, then tools

So what does strategy before tools actually look like in practice?

It starts with knowing, in concrete terms, what a genuine buying trigger looks like for your business, before you ask anything to go looking for one. A prospect changing marketing director might matter enormously to one agency and mean nothing at all to another, depending on who you’re built for and what you’re actually good at. If you can’t describe the signal precisely, no amount of AI-powered research will find it for you, it’ll just hand you a longer list of things that might or might not matter.

It means locking your positioning and message before you let anything draft outreach at volume. AI can write a hundred variations of an email in the blink of an eye, but if the underlying message is vague or generic, you’ve just found a faster way to send a hundred forgettable emails. The message has to be right first. The volume is the easy part.

And it means deciding, deliberately, where a human has to sit in the loop before anything reputational goes out under your name. Research and drafting can happen at machine speed. The final read, the sense check on tone, the decision to actually hit send, that stays human, on purpose, every time. Not because the AI can’t be trusted, but because you’re the one whose name is on it when it lands in someone’s inbox.

Twenty years ago, the agencies who came through the digital shift well weren’t the ones with the most channels or the newest tools. They were the ones who’d done the less glamorous work first: getting clear on who they were, who they were for, and what they wanted to be known for. Everything else followed from that.

The same is true now, except this time the tools are more powerful, the pace is faster, and the cost of getting the order wrong is higher. An embarrassment of riches or fools gold isn’t really a question about AI at all. It’s a question about whether you’ve done the strategic work that makes any tool, however capable, worth using.

That’s exactly what I’ll be covering in Marketing Your Design Business, my three part course for the Design Business Association this September. It’s built for those who want to get their objectives, positioning, message, strategy and plan genuinely clear, so that when they do turn AI loose on research, outreach, and content, it’s amplifying something worth amplifying, rather than helping them make the same mistakes faster and in public.

If that sounds like where you are, I’d love to see you there.