Becoming AI-native

What this means for growth, agencies, and employees

“The way I interact with a computer has completely changed forever.”
– Mark Holland, our Head of Creative Strategy

Something changed over the last month. I and another colleague have become AI-native.

My colleague Mark, our head of creative strategy, told the team at Friday Beers & Cheers “the way I interact with a computer has completely changed forever.”

As soon as he said it I realised it was the framing I’d been mulling on for a week or two.

It’s statements like this that start to turn heads away. You risk sounding like a hype merchant or a peddler of NFTs. But this time it does feel true.

By this stage I’ve heard a lot of people reference Matt Shumer’s essay Something Big Is Happening¹. The intro lands the plane:

Think back to February 2020.

If you were paying close attention, you might have noticed a few people talking about a virus spreading overseas. But most of us weren’t paying close attention. The stock market was doing great, your kids were in school, you were going to restaurants and shaking hands and planning trips. If someone told you they were stockpiling toilet paper you would have thought they’d been spending too much time on a weird corner of the internet. Then, over the course of about three weeks, the entire world changed. Your office closed, your kids came home, and life rearranged itself into something you wouldn’t have believed if you’d described it to yourself a month earlier.

I think we’re in the “this seems overblown” phase of something much, much bigger than Covid.

This is a bit of a longer read (12 minutes or so) than usual and so am giving you a bit of a breakdown of what to expect here:

• The job market today

  • Defining AI native

• I think we’re months away from AI being able to do most of our work

  • Creating a handover podcast for a client

  • Identifying product focus based on retention, LTV and cross-sells

• The new skills we need to learn

  • Taste, multi-disciplinary learning, and critical thinking

  • Product & engineering

• What I think this means

  • What this means for agencies and growth (and by that token Ballpoint)

  • What this means for employees

• Closing thoughts

Anything shared at this stage is really a reflection of current understanding. And so more so than anything else I encourage all of you to research this yourself and form your own viewpoint.

---

The job market today

A lot of this conversation can often be too abstract. And so today I want to really go through specifics for our industry.

First let’s consider the current state of the employment market.

The IPA release the annual Agency Census² data in February. Key insights from it:

1. Employment in creative and non-media agencies is down 14.3%

2. Overall agency headcount is down 6.7%

3. Part time headcount is down 9%

4. Under 25s is down 19.2%

5. Open job roles decreased 41% down to 680

6. Graduate recruitment is 56% down

7. 24% of agencies expect to reduce workforce this year due to AI

Jack Dorsey just reduced headcount at Block by 4,000 down to 6,000. Block haven’t replaced those jobs yet he just knows its coming and is pre-empting.

This pattern is being seen across the wider economy. Job vacancies are down and layoffs are happening, in large part, due to the expected impact of AI rather than its current state. And as the Economist reported last week, we have yet to see the AI impact in macroeconomic data on productivity³.

The current wave of job cuts, when AI is referenced, for now seem to be excuses.

And you can tell this because when you talk to people in different industries, when you hear how the economy is still working, you know that the world is not yet AI-native.

Defining “AI-native” or “AI first”

I’ve been mulling on the idea of what being AI-first means or being AI-native. I see this as the following:

AI-Native. adjective.
Definition: 'Someone whose starting point for problem solving is giving the job to AI, rather than asking AI to help them with their own solving of that task.’

For the last month, I’ve been spending at least two-to-three hours a day, seven days a week experimenting with AI. I’m already blown away.

AI has been able to either (A) dramatically reduce time of a task, or (B) do something to a level we never would have been able to do without.

But importantly my default is not ‘ask how AI can help’ but ‘instruct AI to do the job.’

So what does this look like today?

---

We’re less than a year away from AI being able to do up to 90% of the current jobs

When I think about the strategic and executional work we to do in growth, performance marketing, and as an agency today, I think we’re months away from AI being able to do all of it.

The more I’ve used Claude Code, the more I’ve seen how much it can do. And I don’t think it’s there yet, but the leap from today until that point now feels measurable. It also may be something we can already achieve, if we put enough work into it.

I imagine within the next few months, if not by the end of the year, we’ll be in a place where almost all the current types of work that performance marketing entails will be significantly improved by AI.

I imagine the time required will go down by maybe 25-90% depending on the task itself, and capacity for what’s possible will increase infinitely.

But where are we today?

Here are a couple of specific use cases I’ve played around with recently.

Creating a handover podcast for a client

Background
We were migrating one client from one growth strategist to another. This is a process that historically requires the departing growth strategist complete a handover.

The old way

• Sitting down and from memory writing out what the last 3, 6, 12 months looked like

• Rewatching recordings of the quarterly/half-yearly reviews or the ‘Bigger’ meetings like strategy resets, and adding more specific notes from those to the handover

• Doing a load of fresh long term data analysis, getting charts showing monthly performance for the last year

• Taking screenshots from reports and dashboards

• Doing a summary of key players and providing verbal background to it

• Summarising creative wins

• Sitting down and drafting a long memo or deck

• Sharing the deck. Getting the person to read it

• Meeting to discuss

Done well, it’s probably 8-16 hours of work for the person creating the document.

Anyone looking at this methodology knows it’s flawed.

Humans are choc-a-block full of biases. Our brains prioritise and store information relative to those biases. And so we will inherently leave so much context out during this process.

The AI-native way

Flipped around, I wondered if there was a better way to solve this.

I’m a huge fan of the Acquired podcast series: an entire business story including an investment thesis over a four-hour period. There’s a strong narrative interspersed with business analysis. It’s captivating, you actually listen to it as entertainment, and retain more because humans like stories.

We’ve got 3 years of:

• Growth meeting recordings

• Experiment writeups (~1,000)

• Meta, Google, Shopify, COGS data in BigQuery

• Internal and client-facing Slack conversations

• Growth models and forecasts

So I fire open Claude Code, stick it in GSD planning mode and begin discussing the problem at hand.

My initial prompt was something like this.

“I need you to help prepare for me a 45-60m handover podcast. One of my colleagues is departing and another taking over their account. I think the Acquired podcast series is exceptional at podcast storytelling for business stories and I want us to take inspiration from that. You can connect to the Fathom MCP to access all of the call archives (there’s about 150 of them, so you’ll need a good systematic way of analysing them all), there’s Notion where there’s over 1,000 experiments (again with lots of depth, video, statics included, so please take that into consideration), Slack conversations – both with the client and our own internal, recognising the nuances between those two, my Google Drive contains lots of individual reports often needed for annual reviews and one-off reports, and their BigQuery database contains data from Meta, Shopify, with key reports on acquisition, growth, and LTV. The podcast should take the form of a two way conversation with one person the knowledgable expert, and the other an inquisitive newbie. I imagine you’ll use Elevenlabs for the generation itself.”

GSD now went back and forth on clarifying questions. We went quite specific in certain areas as to what we should include in the final output.

I should point out, I’ve got a claude.md file set that explains how Ballpoint work as an agency, types of work, and importantly a writing style guide that was generated on over 100,000 words of copy I’ve written by hand.

About 20 minutes in, I realised this should be a repeatable process and so I interrupted it to tell it to turn this into a skill.

The output was great. I was going to skim listen to it to QA it but ended up listening to all 45 minutes because it was genuinely captivating. As CEO, I couldn’t possibly have this level of context of a client anymore, but every single sentence spoken was so rich in information I felt like I lived through the choices, the turns, the ups, the downs of that client’s journey (and our relationship with them).

The person who received the podcast said it was the most comprehensive podcast she’d had so far. As a bonus we had a knowledge file which she could now quiz and ask questions to: a knowledgebase informed by 100 hours of interviews, 1,000 experiments, and thousands of messages.

It took me four hours to make this. But this was probably my fifth proper AI project. I’m confident I could get to it again now in less than an hour. It also is now reusable.

This was 2-4x faster than the old way, but infinitely better in terms of output.

Identifying product focus based on retention, LTV and cross-sells

Another brand we work with has hundreds of SKUs across dozens of categories. We’ve long tracked LTV and cohort retention. It is after all that LTV we use to inform what the CM3 target should be.

But we’ve never sat and really analysed those customer journeys.

Why?

There’s also no real plug and play SaaS that does this. No company’s product catalogue is uniform enough for this. Sure if you’re a single-product LTV, out-of-the-box Shopify plugins will give you an answer, but beyond that you’re out of luck.

And for us to do it would cost a lot and take up a lot of time. It’s what I always think of as ‘exploration mode’ in data analysis. You don’t quite know what you’re looking for but you’re giving room to be inquisitive.

But in a client-agency relationship when its effectively paid on time, it’s harder to warrant a cost for something that might yield no results.

The AI-Native way

AI doesn’t have these time limitations.

So once again, I open up Claude Code, I get into gsd:plan mode and start explaining what I want to do.

“Client X has hundreds of SKUs across dozens of categories. They anecdotally find a lot of their customers migrate between product categories as their knowledge and needs expand. But it’s very difficult to analyse this. We want to know if there are clear user paths through products, if there are patterns we can analyse, and therefore see if there’s different ways we can market or focus with retention. Additionally, we want to find out payback and LTV per product. I imagine you’ll want to connect via the BigQuery MCP to access the tables we have we have these tables xxxx, yyyy, zzzz that contain CM2 per order info, spend levels, and product categories. We’re connected to the Shopify API so if we need extra data not in those tables already we can grab that too, just let me know.”

Again this took me maybe 3 or 4 hours to get a decent output for it. Still a large amount of time, but this is because I’m doing this while learning. Now I’ve gone through this process I can prompt better and faster.

The result in this case was identifying some products which seem to have better payback periods and LTV profiles that *aren’t* the key area of focus, so we’re now validating this and seeing if the pattern remains with new focus. Are those entry products correlative or causal?

Let’s find out.

---

I write twice weekly here every week. This is the weekly Tuesday essay: a key learning from my 16 years experience growing consumer businesses. Every Friday is a weekend debrief: key stories, stats, and reads for the weekend.

Posts are (currently) free, but the archive is paid.

Subscribe as a paid member for full archive access now.

---

The new skills we need to learn

Taste, multi-disciplinary learning, and critical thinking

Scott Barker wrote in his Substack last week on How to prepare for the next decade⁴.

There’s lots in there to dissect from a mental health, identity, and career mindset. But pulling out some practical elements to do with skills for work, I thought this section was great:

“What happens when the skill you spent your life learning can be done at 80% proficiency by a novice in a matter of minutes. We’re already there across most disciplines. AI and robotics will eventually flatten all surface-level skills.

The skills that will matter in the future will be things like judgement, taste, the ability to make connections across disciplines, story telling and moral/ethical reasoning. Going back to the car example, the skills you will need will more closely mirror a navigator than that of the actual driver.

Taste, judgement, cross-disciplinary connections, story-telling and moral reasoning are much harder to foster. These require life experience, failures, interests in many different areas and a lot of time spent in contemplation. It will require you to understand how you actually view the world, what you stand for and who you are.”

Reading this I remarked on how well place people in growth will be.

Growth to many is just rebranded marketing. But I’ve always seen it much deeper.

• Growth mindset comes from Carol Dweck and is the idea that you can constantly self-improve and learn, and abilities are not fixed

• Growth in tech brings together multiple functions of marketing, product, data, and engerineering

• We’re storytellers by nature in our advertising and copy

• We’re visual by nature

There is still lots to learn. Taste is something that needs to be learnt through experience. Reasoning requires active work. And judgement comes by default from experience.

How do we build a world where we foster this stuff together?

So much of this also fits into the adaptation to harder skills as well.

Last year, when I called for the death of the t-shaped marketer, I did so because I thought that old version was too narrow. Instead, I called for the creation of the ‘growth artisan’.

This is the idea that a single person could grow something from start to finish. In current terms that means Shopify, conversion, customer, copy, ads, measurement.

Already with AI, knowledge is at our fingertips, but soon the ability to create within those disciplines will be as well.

As we all play with AI within our current areas of expertise, we know that out of the box, the context is to hallucinatory and broad to be useful. But with the right degree of context, memory-setting, and relevance, it becomes powerful.

Product, design & engineering

I asked Claude Code if I should learn to code the other day, and it told me I didn’t. I did smile at that.

Whether or not I need to, only time will tell.

But I know that being a nerd and someone who broadly understands concepts of TCP/IP, networks, the internet, the web, databases, and high-level computational science has been useful for me as I go.

Having grown up messing around with the command line helps too.

So too does the fact that I used to sit next to a graphic designer, am good friends with a product designer, and I flirted with the idea of being a product manager.

Having some grounding in those functions has been useful. And I believe it’s going to be beneficial for anyone using AI to understand them to some degree.

---

What I think this means

What this means for employees

If we look 10 years out, I don’t think there’ll be an AI advantage. It’s available to everyone. And currently there’s no switching costs between the core platforms.

As it currently stands, I fundamentally believe that learning to use AI properly is going to be the same as how everyone had to learn to use a computer. There will be people better at it than others, but if you want a job you just need to know how to use it.

We’re dedicating a lot of time internally to get everyone up to speed on Ai.

We’re not hiring a chief AI evangelist. We’re not outsourcing this. We’re not going to create a function responsible for it.

It’s in everyone’s OKRs. It’s on everyone’s roadmap. It’s in everyone’s progression plan.

We’re betting that everyone has to become AI Native, not just a handful of people in each company.

If we hit OKRs, then by about mid-summer, we should be a fully AI-Native agency.

I think businesses that don’t do this eventually will go extinct. Lots will go out of business, and they’ll get replaced by native/first businesses.

When we look at those job stats, when we think about the future Matt Shumer writes about, and the preparedness that Scott Barker encourages upon is, a few things become clear:

1. The nature of ‘work’ will change forever

2. We are all going to have to reskill

3. Junior team members are going to be hit the hardest and so will need the most support

4. We need to change the internal view of what success looks like

5. Lots of jobs will go

Like many others before, I’d encourage anyone reading this to go deep soon.

What this means for growth (and agencies)

So much AI narrative at the moment is about cost reduction. But to me that doesn’t play out too much.

a16z are looking for AI native agencies, and the reason being is that for the first time agencies could have software like margins at scale.

It feels a bit like that’s a view of agencies from those who haven’t dealt with demanding clients. If they had, they’d know that no client is ever going to let their agency get away with 90% margins. Long term, it just doesn’t feel like there’s a business case for cost-efficient focused agencies.

For me, I see more excitement as we did in my LTV example above. We are at day zero with our AI work, but already it’s enabling us to do things that we just couldn’t have done before.

There are services I’d love for us to offer because I know they’re in the interests of the client, but the client maybe can’t afford them, or we can’t hire the person for a short budget or one-off project to do it. And so the work doesn’t get done. AI solves this.

And then we have the longer term view. In a few years time, when everyone has access to this. Once the great resettling has separated the wheat from the chaff, and the AI-refuseniks go out of business, what happens then?

If advertising is about how we sell products to consumers, then creativity is how we do that and stand out against our competitors. That need won’t ever change. Humans will always desire more, and when they do they’ll always buy more. Companies will always want to sell more and beat more competition. And so the need to stand out remains. Creativity is the answer, we will all just be more excitingly armed.

---

Closing thoughts

This feels like the most exciting period of my career to date. It also feels like one of the most anxiety-inducing ones. What I’ve become deeply aware of this last month is that we can no longer keep our head in the sand.

At Ballpoint, we have already begun the work of trying to internally revolutionise. This to me feels like the ‘go mobile’ moment at Facebook all those years ago.

Personally, I’m experimenting with almost every spare minute I have. And I encourage everyone close to me, on my team, and more broadly to do the same.

I am a deeply optimistic person, and I remain so. I think there is huge upside to be had. But that is not inevitable, and it won’t just happen to us, we have to take that opportunity ourselves.

I’m going to be writing more this year on how this journey goes.

In the mean time, I’d love to hear what you all think.

Josh

---

If this sounds exciting to you, then maybe you want to work here? We’re currently hiring for a head of growth level paid social expert, a data engineer and analyst, and a PPC manager to join our London office.

And if you are the CMO at a consumer company that feels stuck and fed up that the current way isn’t working, then maybe you’d like to work with us. We don’t have all the answers yet, but we’re sure as hell going to try find them as fast as possible. If so, drop me an email josh@weareballpoint.com

---

I’m building Ballpoint: the growth agency I always wanted to hire when I was a DTC founder and before that a head of growth. We have scale brands from £1m → £50m through digital advertising. If you’re looking for support, then you can email me on josh@weareballpoint.com.

In the meantime, if you enjoyed this please consider giving it a like, a comment or forwarding it to a friend.

1. https://shumer.dev/something-big-is-happening/

2. https://ipa.co.uk/news/agency-census-2025

3. https://www.economist.com/finance-and-economics/2026/02/22/the-ai-productivity-boom-is-not-here-yet

4. How to prepare for the next decade