Skip to main content

One post tagged with "GenAI"

View All Tags

· 13 min read
Chris Tacey-Green

title image reading "The anatomy of paved paths" with lego behind it

Oh the irony

This was the first image that came up when I asked ChatGPT to generate an image for this blog post. I thought it was too beautifully ironic to ignore. Maybe AI can be funny?

Context

Firstly, some important context. I use GenAI tooling almost every hour of my waking day. I think it has incredible value already, and huge amounts of future potential as we build around it. I'm not one of the people saying the tech is a stochastic parrot and has no value. I'm very much on the hype train.

Also, not that credentials are required in order to have an opinion, but I have some old-school experience with AI. I was working on ML algorithms over 10 years ago. I remember the excitement I had when I discovered GANs (Generative Adversarial Networks) and was convinced this was our path to superintelligence. The first GPT I started typing away with was GPT3 (we've come a long way).  More recently, I've been building multi-agent platforms in my spare time. I'm not going to write about any of that today, though. There are many many many people who've written great pieces on how the tech works. I may write on the tech in another post, but not today! This braindump is much 'softer', based on where I see the technology industry going, and crucially, where I think individuals and organisations can differentiate themselves in this environment.

A note on AI usage within this post

I started writing this blog with AI. I then questioned what the point was of doing that. I write in order to bring my ideas together. It's not necessarily about the produced blog post; there's value in the writing process. So I used AI to help with a suggested structure, and to help bring some of my voice-notes-whilst-driving together. I then deleted the whole lot and re-wrote it once I had the structure in my head. The only input AI has had in the finished post is correcting any terrible grammar, but I ensured it left the wording raw. It will read like it's come out of my mouth, rather than a beautiful piece of AI generated content. I hope that feels refreshing, and real :)

The Vibes

Lovable. Replit. v0. Bolt. If you've been anywhere near the AI conversation online, you'll have seen wild claims of someone using these tools to "Build a $10m ARR SaaS Product in 2 days". If we ignore the exaggeration, and the heaps of clickbait, there is genuine value in the tooling. A lot of humans benefit from visual aids when making decisions. There's a reason that presentation decks exists; they help to tell stories and deliver thoughts where a written memo would be lacking. (apart from Amazon apparently, who only use memos (this has been disputed by internal employees)). By getting all stakeholders in a room, smashing thoughts into Lovable and seeing the results of your stream of your discussions, you will drive better and faster decision-making. It's the agile dream, what a feedback loop!

However, we must remind stakeholders that prototypes are not working software

One of these vibe-coding tools (I will keep them nameless) emails out a newsletter showcasing AI-built apps on their platform. 5 'top picks' from the thousands that had been generated. As I tapped through on my phone, the lack of care smacked me in the face. Two of the apps loaded with errors. Another two weren't responsive for mobile, so loaded with overflowing text and buttons all over the place. Only one app genuinely delivered. Now, is the tooling going to improve? 100%. But I was shocked that these were the top picks, chosen to advertise the power of vibe-coding. Remember, this post isn't about the tooling. It will change. What we need to talk about is what differentiates you in this new world...

Attention to Detail.


Attention to Detail

There are three areas I'm going to pick on as examples, although I think we could create a much longer list.

Attention to Detail on Facts

GenAI is hugely disrupting search. Perplexity popped up as an answer engine, and demonstrated what the combination of LLMs and traditional search could do. No need to click through and read 10 different web pages from your Google search, Perplexity will do that for you and synthesise it all into a beautiful answer. More recently, the big LLM providers are all adding research (or deep research) functionalities, which can spend minutes researching a topic for you, bringing together information and sources, and presenting it to you. 

So why do we need to pay attention here?

Even with the great work these agentic approaches pump out, hallucinations are still common. I have no doubt that this will improve in the products, but I still regularly get research responses where citations and quotes from sources have been entirely fabricated. How do I know? Because I check them. And you should too. You've saved an awful lot of time in aggregating this information, please spend an appropriate amount of time fact-checking it. Paying attention to the detail within the generated content will, unfortunately, differentiate you from a large number of people, who blindly take these responses as facts. As an aside, I've worked a lot on building out multi-agent frameworks that can reduce these poor citations. There are flows that can help, but for now, fact-check yourself. You will still be accountable for whatever you do with this information, so you should have a healthy amount of skepticism and apply critical thinking. This will give you the best of both worlds: significantly less monotonous work, and high-quality outputs that reflect your care and skill.

Attention to Detail on Working Software

Everyone (with the financial means) now has access to vibe-coding tools, and AI Coding Agents (Cursor, Windsurf, Github Copilot, Manus etc). Time to POC has plummeted. So in a world where everyone can build an app, how do you differentiate yourself? It's the detail and care over the software you're building. I'm not suggesting you need to think about these things when you're living the agile dream, prototype-vibing with your stakeholders. Focus on the decisions that need to be made then. But once those initial questions have been answered, you are not done.

  • Have you genuinely thought about the features you're building, or were they simply features the AI decided would be good?
  • Have you thought about the integrations and data you need to make this real?
  • Have you thought about how this software is going to run across the multitude of platforms that your users will expect to use it on?
  • Have you considered the security model that needs to be applied to each of the components of your workload?
  • What scale are you wanting to build for at this point in time, and do the decisions you've made align with that?
  • Have you thought about the user experience you're trying to achieve here, or have you simply accepted the generic template that the vibes built? (more on user experience later). 

The list goes on.

This, in my opinion, is where you differentiate yourself from others. Use the power of these tools, and then take them to the next level. Use your experience. Use your knowledge of the problem space. I genuinely think that by doing this, you'll produce exceptional software, at a pace that you never could previously. And crucially, it will stand out amongst the sea of AI generated software where this care wasn't taken.

Side note - Code accountability

This really shouldn't need saying. Whether your code came from Lovable, Cursor, Claude or StackOverflow, you are accountable for it. This does not change. It will feel a little odd at first. The code in your IDE has been written by someone else. But if you think about it, this is something we've all been doing for years right? Your team members don't write code in the same way as you, but you review it, you pair with them on it, until you're at a point that you're happy to have your name stamped against it alongside them. That's exactly what you should be doing here, if you're applying attention to detail.

caution

Those who are fully vibe deploying to production - you are wild. I hope you only do this for online games, and nothing that handles my payment processing...

Attention to Detail on User Experience

If you've been using vibe-coding tools for a while, you'll find that you can spot similar apps in the wild quite easily.

"Ah that's a Replit app"

"That looks like it was from Lovable"

They are intentionally opinionated, which is why they create such incredible applications in such a short amount of time. Those opinions also mean that they're less flexible and less creative when it comes to your problem. They have standard components to apply. The niggling feeling in the back of my head kicks in here. I think this generic nature of applications works if you're using these UIs for internal use-cases (e.g. Making it easier for an operational team to do a certain task). But when we're talking about our paying customers, in a lot of cases this is going to prevent your organisation from differentiating.

If you simply need a shopfront for your hand-built tables, your differentiation is your woodworking skill. But in a lot of SaaS applications, your differentiation is your user experience. You win because users love it when they visit your site. They feel comfortable. It feels intuitive. They 'feel' your brand. This comes from that oh-so-fluffy world of UX. It comes from designing the journeys that your users take through the application. It comes from attention to detail. As with all of my caveats, the vibe-coding tools may well become easier to give this flair to your applications. But it should never remove the responsibility from you. As soon as that happens, you've stopped thinking about your users. I think caring deeply about your user experience is going to be one of the key things that differentiates you and your organisation going forward. Hold onto this, prioritise this, and you'll see long-term value.

A side note on user experience

Can we talk about chatbots? I get the feeling that everyone thinks adding chatbots to their website is an amazing idea and we all want that. I'm not sure about you, but I can't think of anything worse. If our idea of utopia is that now, instead of opening 5 websites and using a well-designed journey through it, I now open 5 websites and hope for the best that my worded message to the chatbot is understood correctly: I'm out. Don't get me wrong, I'm all-in on agents doing tasks for me. But it shouldn't be an agent per website that knows nothing about me. It should be a personal agent, that uses APIs (or MCP if you like) to integrate with the 5 services for me, through one single agent experience. Maybe I'm out on my own with this opinion, but I really do not want website-specific chatbots, and I already avoid them wherever I can.

A second side note on user experience

I keep hearing customer service being touted as the great opportunity for organisations to automate with AI. Why?! For a lot of companies, customer service is one of the rare opportunities that you get to actually communicate with your customers. It's so hard to get feedback and interaction with customers. It's something that startups are desperate for. And you're going to automate this away with an AI Agent?!

This leads me onto my bonus differentiation:

Being great at interacting with other humans

For individuals, this is something that will become more and more important. Certainly within the tech world. You may not need to be typing away as much code in your day, but your knowledge of the business, the processes, your users and their pain points, is all going to be a differentiator for you. You will produce better AI-generated solutions because of it.

For organisations, I'm convinced that AI fatigue is going to hit all of your users within the next few years. They're going to be tired of interacting with AI agents during their workday, only to talk to more of them when they're trying to get help from your company. I think once the excitement about chatbots has died down, the organisations that have prioritised human interaction in the client-facing parts of the offering, will thrive. They will be seen as unique. People will appreciate speaking to a human. They'll find it refreshing, and it'll feel like a premium experience.

Okay, so how do we make that a reality?

People are expensive, and replacing them with AI Agents will be entirely achievable in the near future. But let's just think this through for a minute longer. My suggestion would be to automate everything behind the human. Make it that they don't need to fight with systems, trawl through data, and spend half their day doing administrative work. Build agents into their interactions, so that the human is getting hints, answers and realtime data fed to them as they're talking to the user. Make them superhuman. Don't make them a robot. I'm not saying this for empathetic reasons, I'm saying it from a commercial perspective. I'm betting that in a few years time, people will be desperate for human-interaction.

Final Thoughts: Attention to Detail is your Differentiator

In a world where more and more of the world is generated by AI, attention to detail will become your differentiator. As an individual, this means fact-checking and holding tight onto your accountability of what you produce with AI. It means caring and paying more attention to your user experience. As an organisation, it means all of this, plus being crystal clear where your brand differentiates. And if you are lucky enough to interact with your users, do not replace that interaction with AI!

If you read to the end, thanks for spending that time! I'm genuinely interested in your thoughts, so ping me wherever you find most convenient. This was a bit of a braindump of the pinball thoughts that have been bouncing around my head for a few weeks. Less structured, less deliberate, but quite fun to write!

Until next time x