Inside Out AI
TLDR; As the SaaS heavyweights continue to roll out AI features, brands face the real risk of diluting their differentiation. By design, AI gravitates to a bland median, so when everything from your customer support to eCommerce copy to social posts is generated by generic AI tools, the customer experience will become grey and boring. Brands with true differentiation should be building AI from the inside out, baking in their uniqueness through workflows and models that they own, deploy and improve. And this is possible – the speed and ease with which agents, workflows and models can be built means you can unbundle the AI from your SaaS tools and not fade to grey.
Bundles of tools
A foundational belief pumping up the enterprise AI bubble is that knowledge work is nothing more than bundles of tasks. This isn’t the whole truth, but it’s true enough that there are some tasks in some workflows that have discrete input and output stages that are well suited to LLM-driven automation.
This has the SaaS big dogs frothing. Salesforce has Agentforce™, Shopify has Magic™, Hubspot has Breeze™, Canva has Magic Studio™. All of these hastily named AI units are breaking off small chunks of workflows and shipping LLM-powered tools that remove some time, effort or expertise out of a person’s job, or alternatively create a more customised output where a generic one previously stood.
For brands big and small, this is how AI is entering the picture. Unlike the world of coding, people working on brands don’t suddenly have an AI sidekick to do most of their job for them (don’t at me, developers). Instead, a cute little ✨ icon is popping up offering to magic up a small piece of workflow.
Your tools are not differentiation
These SaaS platforms are ubiquitous. Every brand from the greatest to the blandest are using them. And until now, that’s been fine. Apple and Nike don’t differentiate because of their CRM or eCommerce or social media tools, they differentiate through them. The brand is delivered not simply by a set of guidelines, but by its people – their norms, culture, ways of working, artefacts and rituals.
These elements are often not documented. They are (to quote Dennis Denuto moreso than Andrej Karpathy) the vibe. They are what sustains great brands and allows them to be above average year after year after decade.
But by design, LLMs are very much average. Language models are trained on everything we have ever written flattened down to a multidimensional lattice of numbers. These models distill patterns across contexts and are exceptional at connecting concepts, but their natural gravity is always towards the median.
Whether it’s trying to get an LLM to write in your voice or making your customer support chatbot not sound like a chirpy moron, fighting this gravity takes a lot of work. At risk of being the 872th person to quote Brian Eno on this topic this week, he does nail it:
“what you spend nearly all your time doing is trying to stop the system becoming mind numbingly mediocre. You really feel the pull of the averaging effect of AI, given that what you are receiving is a kind of averaged out distillation of stuff from a lot of different sources”
When every brand is using the same platforms, and everyone is clicking the same little starry icon to invoke the AI magic, everybody is hurtling towards the same mind-numbing mediocrity.
Building AI inside-out
Through their ubiquity, SaaS platforms are driving outside-in adoption of AI for brands. They are using their scale and distribution to make this path for AI adoption feel inevitable. This is, after all, the playbook that defined their success – very few brands have the scale to build their own CRM, email or customer support platforms, so relying on the VC-backed tech companies to build this infrastructure seems like a great idea. And it was a good deal when it was just infrastructure. Roll a platform out, your people adapt their workflows, and hopefully your brand essence survives.
But now with AI, brands must own the solutions from the very beginning, especially for anything touching the customer experience. Building AI agents, flows and models is nothing like building a massive SaaS platform. And so brands now have the opportunity to build inside-out rather than accepting the magic-wand AI offered to you (and every one of your competitors) from the outside.
This may sound hard, but it has never been more feasible and accessible to build your own tech solutions, bespoke to your brand. You can bake in what differentiates you without needing VC money and SaaS scale. The reality is that most in-platform AI tools are nothing more than text boxes (we’re no closer to solving the takeoff and landing problem), so integration isn’t a blocker.
Building inside-out has another huge benefit – it builds your capability. Understanding how an LLM interprets your brand when generating text and images, what specific phrases nail the tone or send it somewhere weird, and how to “work beside” the AI and not under it, these are critical things to learn and understand deeply. Because even if you do lean into the default AI tools built in to your platforms, you’ll at least have some skills to fight the gravity of blandness.
This isn’t for every brand. If you compete on price and convenience, the mediocrity tax of default AI might be fine for you. But if your differentiation comes through your customer experience and not just what you sell, if your brand is built on distinctiveness, then accepting generic AI is accepting the erosion of everything that makes you valuable.
Over the next few months and years, I’m expecting to see new brands rise because they’ve understood how to deliver distinctiveness through AI built from the inside out. At the same time, more than a few great brands will fade to grey by deploying off-the-shelf AI with a single click – a click that smoothly erases everything that ever made them distinct.
- November 2025