I’m calling bullsh** on headlines like this:

AI costs literal peanuts.

OK, so if your company is building AI, yes, the AI might cost more (no-duh) but there’s no way for a normal company, or a normal individual that AI costs more than you do!

Now that I’ve got that off my chest. The real question is in fact:

“What’s an acceptable amount of AI usage?”

Or, a more on the nose question:

“How the f*** do I know whether I’m using AI enough / in the right way?”

Some people will openly share that they’re going all in.

Some are more reserved (whilst secretly using AI more than they let on).

Here’s my view:

If you’re not consuming up to your AI usage limits you’re not using AI enough - why?

If you don’t, you’re not getting a feel for it’s true capability.

That doesn’t mean playing $200 a month for a Claude Max plan.

And, strangely enough, it also doesn’t mean using AI for everything.

Remember when you used to build your your own PC - OK, maybe you didn’t.

But if you went into the systems settings on startup (I think you had to punch F2 like a billion times) sometimes the settings would allow you to ‘overclock’ to make your CPU run in overdrive (before it exploded).

You can’t make Claude run faster, but you can stop it from hitting it’s limits.

This is as simple as allowing it to bill you when you go above your usage.

Just go into your Claude usage settings.

I pay $20 per month for Claude, and have only ever had to top up with like $5.

The screenshot below is from a team of 15 people absolutely rinsing Claude for a month.

Less than 600 EUR to give an entire team free reign over AI for a month isn’t expensive.

Plus, what you can’t see here is that fact that 80% of that paid usage can from 2-3 people (namely Nicolas and myself using AI to generate models, powerpoint, and endless deep research reports for content) - not to mention all of my Claude Cowork testing.

It’s only when you’re using AI to it’s limits that you’ll know:

One week going hard on Claude will make you feel soooo much better about the whole “falling behind” nonsense.

And it’ll give you a much better view of the sort of work you DO and DON’T want to source to AI - More on that in a bit.

Take Claude Cowork.

I went down that rabbit whole with my ghostwriting work.

What did I do:

The results were literal dog 💩.

They didn’t have any depth.

They didn’t use my clients’ voice

They didn’t follow the non-AI writing guidelines

“You just need a better prompt Adam…”

No dice, Bob. I’ve been writing prompts since ChatGPT was a baby, and writing 5-10 mega-prompts per day over 3.5 years puts me into super-user territory.

When you look at the influencers I work with, there’s some common threads.

Most are collaborating with Claude → but have a high-quality bar and are still prepared to to call out when something is not good enough (they’ve not completely drank the Kool-aid just yet)

Most of them still have team members (often outsourced) that they’re encouraging to use AI - and by the way, they’re not letting staff go, if anything they’re taking on more staff as they’re growing so quickly!

The bigger names are doing similar things.

Stephen Bartlett told his team to “use AI agents before hiring”[1] and then raised an 8-figure round to scale the business with more humans on board.[2]

MrBeast’s Beast Industries just posted two AI-specific roles (Head of AI-Native Production and AI Enablement Lead) describing AI as “not a tool but the foundation”.[3] He’s hiring people to figure out where it fits, not firing them to replace with bots.

Just like any Level 5 Leader (look it up - Good to Great by Jim Collins) there is no ego with their usage.

If something doesn’t support the cause and inspire high standards there’s no room for it.

But how do you decide what to use AI for, what to outsource, and what to hire for?

Well…

I’ve recently taken on a VA in Bangladesh.

I’ve had to get over my own bias on “not needing anyone else”.

But, why do I need someone to check my e-mails when I can just can an AI to triage my messages according to a set of rules?

Well, you don’t NEED to…

But as I’ll explain, there’s a big different between the tasks you CAN outsource to AI and Automation vs the ones that you SHOULD.

E-mails are the gateway into the way you conduct your business.

Are they not useful for someone else to learn from, that can then also help you with other things, that - in turn - add up to more time and energy saved?

This is what it’s all about.

‘Buy Back Your Time’ by Dan Martell, is a blueprint on how to, well, to buy back your time actually.

The principles are sound, and the ‘DRIP’ quadrant is an evergreen methodology that relates to how you should think about work.

Dan Martell's DRIP (Delegate, Replace, Invest, Produce) is a brilliant framework - but it was built when delegate meant only handing a task to a person.

In 2026, delegate is a progression:

And for me, it needs to starting with ‘AI-Sourcing’ - which I’m yet to trademark but might.

Before sourcing anything, sort every task into this 2×2.

For everything in the 🔵 SOURCE cell, pick the right tier. Start at the top - only move down if the tier above can't hit the quality bar.

Best for:

Examples: research, transcription, first drafts, formatting, scheduling, data entry, summarisation, light analysis.

Best for:

Examples: editing, light design, inbox triage, image sourcing, content posting, light client comms, AI output QC.

Best for:

Examples: account management, key client relationships, brand strategy execution, hiring, ops leadership.

Best for:

So, I’ll let you into a secret (that’s not so secret)

That entire last section was generated by Claude within Notion - could you tell by the Em Dashes? lol.

(apart from the AI sourcing comment)

And I only half agree with it.

The examples are also rubbish, so if I were to update, it would need to take into account my taste, and real life domain expertise of using AI.

For now, I’ve left is as an example of something over-engineered that lacks depth, that’s half useful, but probs not something I’d trademark (yet).

But what you can do is take the bare bones, and create your own method.

Or maybe just steal this next one…

So, instead of me landing a complicated matrix and workflow method on you.

Here’s a super simple way to think about it.

For any tasks that NEED to be done (that you don’t love and earn you cash) do this:

If you were to think of this as a pyramid. the biggest section would be AI and automation at the bottom. Followed by outsourcing, followed by skilled internal resource, followed by you at the top.

When you go through your AI-First progress, you’ll have to figure our what works for you.

Here’s a few observations I’ve made from my own work.

90% of the world has a f***ing clue what they’re doing with AI.

Even the Bartletts and MrBeasts of the world are still trying to figure this out (albeit with bigger budgets).

Either way - you’ll never regret spending more money to outsource a load of work you don’t want to do yourself.

Outsourcing can be to a thing or a person.

The thing is ultimately more scaleable so start there.

Just don’t engineer out either yourself or the skilled people with taste that bring about incredible amounts of value to the table.

Until next time.

Adam

—

P.S - Want me to help you find you best AI use cases? Reply with what you’re struggling with, and I’ll help where I can.

P.P.S - 3 extra goodies for you:

1. Claude for e-mail subject lines:

How I get 52.98%+ opens on a newsletter to 190k subs here

2. Turn followers into brand partnerships (without relying on agencies)

5 day e-mail course on how I've delivered $500k+ in 6-months for influencers here

3. Resource vault

Everything else I've produced over the past 3 years here.