Over the past year or so, I’ve been playing catch up. A new job in a new industry with new channels and new people. At the same time, new AI tools were arriving like a slurry of overexcited hubris, promoted in the media, by influencers and over Slack channels every single day.
One week Lovable was the big ‘I AM’, the next Clawbot was everyone’s best mate. Somewhere underneath, Claude started to lay the foundations to eventually become the daily driver of record.
Everyone’s got a take on AI. The tech bros say it’ll 10x your business. The naysayers say it’ll take your job, fry the planet, and hallucinate its way through your token budget in hours. Meanwhile the consultants are charging telephone numbers to tell you that you must have an AI strategy (preferably theirs) before you go obsolete.
While there are a gazillion upsides to all this frenetic, competitive land grabbing, nothing this universally adopted comes for free. The tokens paying for the earth-warming server farms are subsidised to the hilt, loss-leading in the finite style of Icarus, except instead of flying closer to the sun, the server cities are bringing the sun to us, scoffing greenhouse-gas-exuding fossil fuels at the same pace a Scottish football fan chugs Budweiser.
It’s not the app-building, vibe-coding, deck-doctoring utopia the LLM turbo-bros would have you believe, so I thought it would be worth doing a teeny tiny reality check to look at the potential downsides next to the amply promoted upsides before we all float away on the hype bubble (or drown under rising sea levels).
The Value / Cost Of AI Adoption
The real question is what this stratospheric innovation will actually cost you / us / our kids, and whether the people pushing you to adopt it faster have factored any of that in (TL;DR they haven’t).
| Item | The value | The cost |
|---|---|---|
| Productivity | Teams produce more stuff in less time, with AI handling first drafts, summaries and repetitive tasks at a pace no sentient being can match. | Output quantity goes up while quality gets harder to police. When everyone hits publish faster, the race to the bottom gets mega competitive. |
| Cost efficiency | Real reductions in costs for repeatable tasks. Tools once reserved for enterprise (personalisation engines, A/B testing at scale, 24/7 chatbots) are now available to all. | The cheap tools are a subscription trap. Stack them up – ChatGPT, Midjourney, Claude, HubSpot AI, Notion AI, Lovable and so on – and you’re quietly spending £500 to £1,500 every month. |
| Personalisation at scale | The ability to segment, tailor and adapt messaging at a granularity that was only previously feasible for enterprise upwards. | Bad personalisation scales mistakes. When your AI works off bad data, you reach more people with the wrong message, damaging brand trust, potentially irreversibly. |
| Speed to market | Campaigns that used to take four weeks now take four days. Product copy, ad variants, social content, email sequences – all compressed. That is a real and valuable gain. | Speed without judgment is just noise, but faster. When every business in your category runs the same tools with the same prompts, everyone sounds the same, putting you in a soup of compromised mediocrity. |
| Environmental impact | Marginal efficiency gains in certain sectors such as logistics and energy grid management, where AI can reduce waste (yeah, I’m clutching at straws here…). | Training a single large language model produces roughly the same CO₂ as five cars over their entire lifetimes. Water usage for cooling servers is measured in billions of litres. We’re writing the world’s epitaph (but at least Claude will write a good one). |
| Decision-making quality | AI can surface patterns in customer behaviour, market trends and campaign performance that humans would take weeks to find. | When it’s not trying to please you, AI hallucinates, and the cost could be reputational, legal, actual or all three. |
| Competitive advantage | Early movers in genuine AI integration are building compounding advantages in speed, cost and customer insight. | The businesses that come out ahead will be those that combine AI with what it cannot replicate: relationships, creative risk-taking, ethical judgment, brand trust and that old chestnut – massive marketing spend. |
| Talent and workforce | A good marketer using AI is more creative and more strategic when not wasting time on manual tasks. | Job displacement is real and unevenly distributed. The jobs that go first tend to belong to juniors – the same people who would have become your senior team in five years. We’re making a hole that something has to fill. |
| Data and privacy risk | AI enables deeper customer understanding when built on first-party data and ethical data practices. | Most businesses are not doing this properly. Feeding customer data into third-party AI tools raises serious GDPR exposure that most marketing teams have not thought through. |
| Brand trust | AI can help brands be more consistent, responsive and present across channels. | When you find out your 30-minute complaint call about that service you pay loads for has been with a bot, how do you feel? Exactly. |
Looking For Goldilocks
So where the hell is the sweet spot?
If you don’t have the budget to deploy AI well, then you definitely don’t have the budget to do it badly. The ‘not too hot, not too cold’ line is extremely thin, so pick two or three tools that solve specific problems in your workflow, then measure actual output quality, not just speed. Don’t let subscription creep turn ‘AI efficiency’ into a new leaky-bucket cost centre that nobody’s managing.
The 2035 Scorecard
If current trajectories hold, this is where we’ll probably end up:
- Productivity gains: real, measurable, and likely to plateau once the tools mature and adoption equalises.
- Cost savings: real for large companies, but mixed for SMEs once true implementation costs kick in.
- Misinformation risk: AI output volume is scaling faster than human verification can ever hope to. And don’t get me started on agentic mediation…
- Brand differentiation: the brands that survive will go balls deep into what AI can’t do.
- Regulatory pressure: the compliance cost of AI misuse is so far down most priority lists as to be entirely irrelevant. THAT WILL CHANGE.
- Environmental damage: well, if you’re comfortable kissing goodbye to Southampton, Dublin, Cardiff, Bristol, Peterborough, Hull and chunks of London, then there’s absolutely nothing to worry about.
AI amplifies whatever you point it at, then tells you you’re doing an awesome job.
If there’s one thing us human beings share, it’s that we bloody LOVE a compliment, so be enthusiastic, sure, but it’s time to be a little cynical too, and stop believing all the guff we get served in exchange for a poorly executed prompt.
Which Brings Us To ‘Doc Diarrhoea’
…one of the most common (but least acknowledged) downsides of the AI era. With everything becoming so bloody easy and efficient, I’ve seen a massive increase in this phenomenon. Every day, and after every meeting, people scamper off with their Gemini, Firefly or Granola notes and start ‘being productive’. New GTM drafts, research papers, campaign ideas, customer analysis and white papers are sharted out like confetti at a wedding. We’re all generating far too much for the sake of production – a clear instance of ‘just because you can, doesn’t mean you should’.
In my last role, I hit a staggering thirty-eight hours of meetings one week, and it rarely dipped below thirty. Every day, my team and others would share such docs, spat out by Claude and their agents.
Nobody ever read them. Nobody had the time.
In a world where AI is meant to make us more efficient and productive, Doc Diarrhoea is doing the exact opposite, so in my humble opinion there’s a serious need for tighter RACI processes and protocols to pivot this wasteful enthusiasm in more positive and productive directions.
It might seem a bit weird that a guy on the cusp of launching an AI training school for local SMEs is being such a Debbie Downer, but I’m firmly of the view that honesty is the best way for us to escape the hype bubble faster.
We need to craft a pragmatic, world-friendly version of ‘The Age Of AI’, to mitigate Doc Diarrhoea and to progress in a more holistic, less power-hungry, greed-fuelled, ‘profit at any cost’ way. To achieve that, we need to collectively work towards a more sustainable, positive, people-friendly way of harnessing this incrementally powerful tech (especially before Gen AI kicks in…).
The next ten years will sort businesses into two groups: those that use AI to do more of the same at lower cost, and those that use it to do something better.
The first group will compete in a race to the bottom, while the second will make themselves harder to replicate whilst benefiting humanity and the planet. You need to decide which group you’d rather be in, and you need to do it soon.
Harry Lang launched The Oxford AI School to help as many local SME professionals as possible learn the basics of AI in a way that’s practical, fun, easy and relevant.
You can connect with Harry on LinkedIn and follow the school on X, TikTok and YouTube. This piece first appeared in Marketing Unfiltered #88.