At the moment there is an ongoing argument in tech about whether 'founder mode' or 'manager mode' is the best way to build a company. Neither of those really matter, because all of these companies really operate in FUBAR mode. When it comes to the implementation of AI and the future of business building, this should be the mode every business is worried about. Here's why.

There’s a common line in the press and social media commentary on AI and business that ‘AI isn’t ready for enterprise’. It says a lot about how many people think LLMs function, how they think Gen AI needs to be integrated into businesses - and even their understanding of how enterprise businesses actually operate.  

If you want to understand how poorly most people have grasped the impact AI could have on business, the statement 'AI isn't ready for enterprise' is the place to start. The logic goes that AI is not ready for enterprise because the output people are getting is not accurate enough to make their jobs easier and businesses more efficient. Microsoft recently had a contract for co-pilot canceled because the output was the same as a ‘middle school presentation’. Spend five minutes searching for co-pilot on social media and you’ll see similar responses. 

Advocates of this position essentially believe AI will be ready for enterprise when the models have improved and are betting on those improvements solving the accuracy problem. But it’s not that AI isn’t ready for enterprise, it’s that enterprise business structures as they stand will never be ready for AI.

Most people don’t realize this because their experience of enterprise is siloed – they’ve often worked in an individual department. If they're in the investment community, they may never have worked in an enterprise-scale organisation at all. If we start from a place of believing AI automation of a business is possible then it's irrelevant either way.

There there is a much deeper problem that needs to be dealt with – the knowledge that underpins the organisation, that defines it, and its processes, is often a chaotic, self-contradictory mess of disconnected documents, fragmented files and siloed concepts.

If the business knowledge is a mess when it's presented to an LLM - the LLM sees it as FUBAR Mode. If you're unfamiliar with the term FUBAR, it's a military term and it means things are not particularly good. But here's the way an LLM defines it.

You can get away with FUBAR mode – albeit inefficiently – if you're using humans to power your company. But if you want to use an AI system based on your knowledge base then you have absolutely no hope of true automation if it's FUBAR. For AI to truly deliver useful automation for enterprise businesses at scale, their entire businesses need to be ripped up and restructured. True, effective AI automation in business can't be implemented on the mountain of sand that is typical of large organisations - a gordian knot of PDFs and documents.

It’s an interesting thought exercise to try to imagine you’re an LLM making sense of what is essentially informational vomit by enterprise businesses. Your job is ultimately to assess this mess and distill it into rationalized insights and even work out what the next step is to take. Good luck.

It doesn't quite use the right expletive for FUBAR, but nails the rest. If you then ask an LLM to explain the impact of their 'work' in this environment the issue for current implementation approaches is obvious.

Yep that’s pretty much it, that’s the problem. And no you’re not solving it with better RAG, a faster, smarter AI model or a new workflow in a department. 

Automating a large business with AI is going to take an entire new approach to what a business actually is – and that's the real multi-billion dollar opportunity. The future of 'big' business will be one-person teams powered by AI.

It is going to take an entire new approach to what a business is and that's the real multi-billion dollar opportunity - the future of 'big' business will be one-person teams powered by AI.

I’ve been fortunate over my career to work with some of the biggest businesses in the world, across pretty much every department, and they all have one thing in common: every single one of them is an absolute shit show of information. In every enterprise company nobody knew exactly what the strategy was at any given time. There were thousands of decks and pdfs stored in all sorts of locations, all housing contradictory information.

Microsoft, for instance, has 228k employees, all creating and sharing documents. It gets really fun when you consider that some of the Fortune 500 have up to 1,000 agencies working with them across their companies. Thinking about how much contradictory information exists in that environment is like trying to imagine how many stars and planets there are in the universe – the human brain cannot compute, but yet we're asking AI models to do just that.  

In one groundbreaking study researchers followed multi-disciplinary teams at Fortune 500 businesses looking at organizational and execution-based issues within their companies. Those who finally got a holistic view of the mess referred to it as an ‘epiphany’, saying ‘it was like the sun rose for the first time. … I saw the bigger picture’. Many quit their roles after taking part in the projects and moved into organizational restructuring roles to try and make an impact on the mess. This paragraph stands out: 

When run through the findings from one of the teams one CEO responded ‘This is even more fucked up than I imagined’ and admitted his grasp on his control of his company was imaginary. I’ve personally experienced this exact response from multiple C-suite individuals at similar companies, including one that has a lot of vested interest in the success of AI right now.

Ultimately, this mess of information is exactly why the global consulting industry exists – an industry estimated to be worth around $250 billion. Think about that for a second, that’s basically $250 billion dollars being spent globally to help businesses make sense of all their internal messes. Consultancies like Accenture are now bundling ‘AI implementation’ into their scopes and seeing a lot of success, which is an indicator that enterprise businesses cannot figure out how to apply AI on their own.

If you’re reading this and work within an enterprise business, it’s likely the colleagues you work with every day might not know what they are doing. Perhaps they think they know what they are doing, but they are doing something different to what you or someone in another department thinks they are doing. You might even be reading this thinking ‘fuck I don’t even know what I’m doing or even doing it for’, in that case don’t worry, you are not alone. Millions of other people out there are like you. One thing is for sure, you’ll end up as part of an expensive discovery session with a consultant at some point, so either treat it as a therapy session or keep quiet and hope things continue as before (they often do).

It’s time for a prophecy: when AI is built to enable a new type of enterprise, and when the balance tips in favor of AI-ready businesses, McKinsey and Accenture will no longer exist. It will be that much of a disruption to how businesses are built.

Making AI work for enterprise in its current state, has sucked a lot of attention and funding from the investment community. It’s understandable: for many investors enterprise is a safer bet, with higher ticket sales. Some startups have attempted to make Gen AI output better and improve the results, and their attempts often boil down to ‘we’re going to search your documents more intelligently’.

Similarly, many attempts at implementing AI within enterprise businesses are ultimately siloed changes to specific department workflows – they don't concern themselves with solving the broader problem of chaotic knowledge. This is unsurprising for two reasons. Firstly, the use cases are obvious and easier to sell in: ‘We make this thing your team does daily easier/faster’. Secondly, many of these companies are founded with specific experience in those exact use cases or departments. You see it in the Y-combinator releases almost daily, ‘Founder X spent 10 years in Y industry/department and now they want to solve the problems they saw so you can better automate your outbound sales (or insert other workflow)’. This is just driving a minor, incremental evolution in business, when a true revolution is needed.

What will it take to truly automate an enterprise business?

The ‘enterprise’ company of the future will be automated with a single source of truth and far fewer employees. It will not operate in FUBAR mode. Enterprise will not look the same as it does now in any capacity. Building that future takes a radical rethink of what a company is, how you structure the knowledge and how you enable automation at scale.

For this to occur everything must connect, AI must have a clear view of central truth of the business it is building. This is exactly the challenge we are taking on with NOAN. Connected real-time business knowledge is everything, it will power assistants and agents and ensure they all work together. It’s what will enable humans to control entire companies on their own – where one human decision will ripple across the entire automated business. 

To build this future enterprise solution we are starting bottom up, by automating for the small businesses first, because in future $1Bn companies will also be small. So we're starting with one-person teams, because in the future they will be the ones with the power to build vast companies on their own.

Want to work on it? We’re hiring, reach out. 

Want to invest in it? We’re raising, reach out. 

One thing I can guarantee it’s not FUBAR here, we build our own company with our own product. 

Neal