It’s only in the last few years, as auto-save has become dominant, that we no longer ‘click save’.

But for a good 20-year run, people clicked on a little icon of a floppy disk not knowing why that button represented the action of ‘saving.’

We became an AI-native agency a few months ago. I’d also define that a few of our clients are AI-native versions of whatever their company is as well.

What’s interesting is seeing how two AI-native companies interact with each other. Or more specifically how the humans within those businesses react.

What’s immediately becoming clear is that so much of the way that organisations, teams, clients, partners operate, is because of the natural limitations of prior working conditions. And actually in this new world, just like the floppy disk, likely will soon become redundant.

Subscribe now

One Claude negotiates with another Claude

For the last 18 months, AI-written email has become the commonplace.

But like all AI-native evolutions, in the early days it was “I want to say something like this, re-write it so that XYZ happens.”

Today, those who are expert do this differently.

I’ve had a few email negotiations this year where the AI usage was far more evolved. You can tell through writing style and verbosity that it’s AI immediately, but actually beneath the tells, there was significant substance to what was being asked.

On our side, we have operations and business leadership AI that understands different service offerings, understands the systems, what clients actually use, and our ‘rules’ which we may have. So a response is not just me saying “I want xyz, go write me an email to get it,” the response is based on understanding the full context of our business.

What feels odd is that at the bottleneck of this was two or three humans acting as negotiatiors-cum-conduits.

In corporate life, you would have an army of lawyers and dozens of stakeholders on either side fighting these corners, but ultimately all comms leading through a single point of contact.

That is where this new negotiation world feels, except instead of the lawyers and stakeholders, you’ve got agents representing those views.

But I wonder if there’s a future where the human isn’t needed at all.

What if each side set their desired outcomes, their guardrails, their limits, and let the agents discuss freely?

Without human emotion wrapped up in it, it’s likely that more equitable outcomes for all parties might come out on top.

It feels to me we’re in the early floppy disk era with negotations, where we’re still mimicking the styles of the past, but soon, we might evolve into a new future.

Subscribe now

Catching out each others’ hallucinations

The other trend I’ve noticed is with hallucinations and in particular how they get caught out.

The old way

Go back two years, and an agency like ours when performance wasn’t good may have done a deep dive of data analysis. One or two people would have taken a few days to go deep on some data and write up a report. The report would be shared, and the client likely would have questioned on the data if they knew something to be off. But otherwise, the assumption was on the fact the data was right.

What this looks like in the new world

Here’s a new world example.

• Performance is bad

• Both client and agency go and get a bunch of analysis done by asking Claude to go deep

• Long form reports get written up and shared with each other

• Each party does some form of checking the other’s workings

• Desired outcome is discussed

Now one point of tension I notice during these times is what I think of as ‘trying to catch out hallucinations.’ Very often one party will see something, and question it. Maybe they know it’s not right, maybe it doesn’t feel right. Maybe they’ve asked their AI to cross-check and validate. Whatever it is, something has changed.

What this could look like going forward?

At present we’re in a murky area where the level of AI use varies wildly from company to company.

For those at the beginning of their AI expertise, you ask the question “Why is performance bad?” and the answer you’ll get is highly plausible, but ultimately worthless.

For us, approx half our time is spent on either data engineering or context engineering. This can mean all sorts of things from:

• Using a ‘deterministic’ script or programme over an AI one wherever we can

• Using evals in the system both at building and execution level

• Introducing adversarial agents to challenge data and scenario plan

Now when all companies are at this level, it means that the above example I gave is a great way for companies to collaborate. Company has all of its own context and data and customer understanding, third party provider has domain expertise and market knowledge. Both parties with good AI systems suddenly produce exceptional pieces of analysis collaborating together.

For working relationships, this default requires a connection of deep trust. But also a shift in what data means.

My guess is its one where that will take the form of: ‘we’re both incentivised towards the right goals, we’re both working towards things together, we’re both using AI to the best of our ability, we’re both using the best data we have at any time.’

Should we feel each other’s analysis into our own systems to validate it? Yes, absolutely. It’s almost a requirement to have additional adversarial agents investigate. Especially because what may be true with one set of parameters, won’t be true under another.

I think psychologically, this will be a big shift.

It puts the onus less away from ‘getting something wrong is bad and a sign of a bad job’, and more ‘so long as the system is structured properly, we recognise this is a part of it, let’s work together to solve the problem and move forward.’

This sounds easy in practice, but I think difficult in reality.

Subscribe now

The end of dashboards and reporting as we know them

In a year’s time, I imagine we’ll barely use dashboards and reports with our clients.

Here’s why.

Dashboards and reports are the way we showed metrics in a world where that was the best way to do it.

But dashboards and reports are rigid at best, and potentially dangerous at worse if they teach you to have narrow focus.

We are already at the stage now where we can spin up a dashboard of any kind in seconds.

Real world example from this week:

Growth strategist meeting
GS1: ‘we’ve just had a rocketship with client X, what do you think we can expect for budget?’

GS2: ‘it was pretty powerful for client Y, let me show you’

GS2 prompts our growth strategist agent to pull the data for said client. It spins up a chart that shows spend growth annotated with the launch of the ad in question.

A hypothetical future scenario:

Maybe we’re in a client meeting, we show data on a new LP we’ve launched. We’re focused on CM3, but client wonders about LTV. Somebody quickly prompts our agent for cohort retention for that product mix, spins up a quick dashboard and we can all look at that data together.

Or a third scenario:

A client hears from a customer they liked purchasing two products together, and ask in Slack for us to dig into that data.

Old world, we then go and do it and add it to the dashboard.

New world, we just get the one-off report easily and paste right in.

Subscribe now

We’re currently replicating our existing modes of business – but soon we will jump them entirely

We don’t use rigid reports because they’re the best way of doing a thing.

We use them because it’s at the tradeoff of: worth the time invested vs outcome of what they can display.

But today AI changes that value exchange because on both sides there is now little limit.

So much of working culture is based on the limitations of where tech was.

That’s about to change.

And I for one am very excited to see where it’s going.

Leave a comment

I’m the founder of Ballpoint. We’re a growth agency that help you scale spend from £50k per month to £500k per month through creative-driven performance marketing. If you want to see what working with an AI-native agency might mean for you, then drop me a line.

When I was a teenager I wanted to be a video games programmer. I bought C++ for Dummies and tried to learn how to code. I couldn’t, for some reason my mind just wasn’t set up for it.

Over the years I’ve tried python and javascript, and each time I get bogged down in trying to memorise syntax, and frustrated. The thing I never wanted to do necessarily was to code, but it was build the thing that came from the code.

AI has taken that wall down. And I’m having the most fun I’ve ever had in my career.

But getting here has meant mourning something first.

I’m 37. Like a lot of millennials who entered professional services in our early 20s, I saw a career path in front of me. My first CEO was in his early 30s. I grew up with nothing and that felt like proof of what was possible if you worked hard and played the game. The path was clear at every rank. Execs became senior execs, then managers, then senior managers, associate directors, directors. Then you either left to start something or you went into leadership.

62% of millennials say work is central to our identity, behind only family and friends. But when I sit down with friends, work is very often one of the primary sources of conversations. We have wittingly or not built a sense of self around that career ladder.

We became defined by our work, and in particular, the workplace has had a cult of management for a long time.

Everyone who goes through the path recognises how odd it is. We spend years learning to execute well – often with very specific training at the start. Only then to be thrust into a management role which none of us are prepared for.

Tech was at the forefront of pushing back on that, inventing the role of the ‘individual contributor’ or IC. After all, not every great software engineer had the want or ability to manage people, and so the IC path allowed them an alternative. You still gained the status of evolving with your job, but it was now linked to building.

At Ballpoint, we’ve always been very top-heavy. Growth strategists run strategy on the accounts, the client relationship and are still in the weeds. But for us, this was about the belief it produced better results for clients – rather than a prediction on an AI future.

But, I think that’s about to be the reality for most professional services businesses.

Think about what a growth strategist actually does. At the core, strategy is deciding what to say no to.

The more experience you have:

• the faster you pattern match,

• the better your decisions,

• the quicker you change direction when something isn’t working.

Two years ago, when a strategist pointed in a direction, a more junior person ran the analysis and built scenarios. Now a good AI does that analysis, and a great AI comes back with experiment ideas the strategist with 10 years of experience hadn’t thought of.

The middle layer of execution and organisation: making sure shit gets done, and carrying out the core analysis tasks, can now be done almost infinitely better with AI.

The result is senior people are doing more hands on work than ever, not less. But it’s not just about the fact these people are doing more hands-on work, they’re also expected – at least at Ballpoint – to be building the systems that improve over time as well.

Two months ago we set a new bar at Ballpoint. Stop using AI to assist your work, and start getting AI to do the work for you. Over 80% of tasks at the agency now begin that way. And where we don’t do it, it’s deliberate. For me, longform writing is the primary one. Writing remains the best way I have ever found to think about a subject.

Getting people to get AI to work for them was phase one. Phase two is getting the AI to do better work than humans. And it’s in these systems where we now expect everyone to plan, research, build, test and iterate.

As of today, I’m most weighted in that direction. A glance at my timesheets show that 20 hours a week go into building. That’s a third of my time spent on executional work. My 22 year old self never would have expected that.

Our head of creative strategy is responsible for building the systems that power his creative strategists. Our head of creative studio for implementing systems that produce results faster. Every single person has an OKR this quarter focused on building.

Subscribe now

What happens to juniors in this world?

So how do you train someone into this world?

I watched a documentary years ago about a restaurant chasing its third Michelin star. Junior chefs weren’t allowed to do anything for days except chop onions. They might have been sous chefs elsewhere but in this kitchen you couldn’t touch a stockpot until you’d chopped onions for days and days and days.

I think one version of the future of training looks like that.

Last year, I touched on this in my article on the rise of growth artisans. The idea that individuals will be able to own the entire growth field – and that training has to be done almost as an apprenticeship along the way.

You learn performance marketing by hand. You listen to customer interviews and write the JTBD specs yourself. You write copy yourself. You export the data and analyse it in a spreadsheet. Only once you’ve done it by hand and you’re expert at doing it that way, do you give it to AI.

The cost implications of this are obviously huge. (And for any government worker out there with a grant, we will happily explore this as a training scheme with you). But you’re essentially doing apprenticeships accepting that grads won’t create comparative value for a long time.

There’s a second version that mirrors the multi-disciplinary nature of our new roles.

There’s a future where actually doing that stuff manually is like knowing long division. Maybe it’s worth learning at school to exercise the brain, but in day to day life it’s just never going to come up.

Maybe it doesn’t matter if you’ve never built an ad campaign or written a line of copy or spoken to a customer in real life. The journey from 0 to good enough will be fast, and then better development of AI systems will get you further.

Here the training is broader but maybe shallower. We might learn product management, customer research, AI engineering, copywriting, a bit of Python, graphic design, ethics, taste. They’d be multidisciplinary by default, because the senior role they’re being trained for is multidisciplinary by default. And this idea, especially that of developing taste, is something I’m focused on at a senior level too.

Mourning our old identity

The old world of work we know is about to change, and I think the unspoken thing at the moment is the knowledge we’re going to need to mourn our old expectations and identity.

Last week, Tom Blomfield of YC said that middle management will die. I reckon it will too, though probably slower than some might predict. I think corporates and the humans inside them have too many incentives to protect the status quo.

But for most, I think it’s important to come to terms with the fact that our old career trajectory is about to change. And I think the idea that there are people who just manage other people and don’t do more hands on work themselves or build themselves, will soon feel out of date.

Not every CEO will be a builder. But I think the best ones will be – or perhaps as Jones Road Beauty’s Cody Pfloker just experienced, it’s the AI-enabled CEO who does the job of exploration, only to rebuild the function internally.

But while not every CEO might end up a builder, I do think everyone else will. And I’m fortunately here to say, having come out the other side of that, it’s a pretty fun world.

Ideas were always cheap, execution now is as well. The question is what will you do with it?

It’s everywhere. X, LinkedIn, Substack, even the NYT got in on it.

The truth is nobody really knows what’s coming. But the possibilities are starting to feel genuinely endless.

At the beginning of March, I published Becoming AI Nati…

Read more

“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.”

Read more