Today is the day where I fully realised that we have reached AGI (Artificial General Intelligence). Its defined by a “software that can learn, think, and do any intellectual task that a human can do, across any topic, without needing special training for each new job”. In the area of marketing and other types of knowledge work, we have achieved this.
I’m saying this because I realised that its outsmarting me in my analysis approached due to its combination of resoning and tool use. As a specific example I want to mention an SEO due dilligence that can cost tens of thouands of euros and hundreds of human hours that I can now complete in an afternoon using AI.
The key to making this work is giving the AI the relevant context. At the point of writing this, 1M token context windows are the max we can work with before conversations are getting repacked. The smart approach here is to use data conenctors of various tools (DataForSEO, Google Search Console, Google Analytics, Databases (e.g. Firebase), Google Ads, Meta Ads and any other relevant tool that contains data that humans would take decisions on.
This data needs to then get in-gested into a system where the AI can model it with python to produce smaller data cuts in hundreds of iterations, joining it with other relevant data based on its advanced reasoning. This Data with all its detail needs to then get enriched with human knowledge that defines the guard rails. For example why have certain things not been implented previosuly, whats the goal and focus of the company, why are we not serving a specific market, which legal issue is preventing us from acting in a certain way (?).
The data and context needs to then get fed to a knowledge graph that is getting steered by an LLM. This combination of knowledge and reasoning is what I have been thinking about since 2022 when I hypothesised with a friend about GPT 2s impact while eating a kebab in Amsterdam. He mentioned this to me a few times over the years and today was the day when I let Fable 5 on Ultracode build this.
Marketing is becoming a systems engineering problem rather than a creative production problem.
The outcome is better than what I’m capable of. Its better and faster than any colleague, even the ones that are deep in the topic and are a CMO since 20y in the same company. The AI is outsmarting them because it has the better context. Its not only able to answer questions better (e.g. how do we best decrease CAC), its also better in actively suggesting new persepectives (e.g. you have 2 Meta campaigns that are wasting budget, reallocate it to Google Search).
This system that I have in front of me is different. I can’t play it. Its smarter in every aspect of marketing. Due to its write access, it isnt only able to suggest, it also just implements directly with the tools. Because of its access to GA4, it can map conversions to the Ads. If something is missing, its using its Google Cloud access to code the missing event triggers into the website. It just fixed my merchant center feed for my marketplace. I only gave it the goal to grow the platform so it went to adjust campaign budgets, optimised conversion rates and fixed bugs that harmed growth. I was able to solo build an entire marketplace and get it to 1000 registered users within 8 weeks.
The tool provider who is able to sell this to our marketing community is the next unicorn.
