Stop Box-Ticking, Start Winning
Free your teams from the bureaucratic hurdles that kill innovation
Hi there,
Today I’m going to write about why box-ticking and risk-aversion is the biggest threat to innovation, especially when it comes to innovation driven by AI. If people wanting to manage risk get too focused on process, then challengers will disrupt incumbents at an almost unfathomable pace with accompanying societal disruption through substantial job losses.
The Big Company Puzzle
Business is all about trial and error. Fresh ideas are a key ingredient of success. If you’re not figuring out how to deliver faster, better and cheaper for your customers, one of your competitors will!
Big companies have a lot of resources - yet they often struggle to come up with great new ideas and execute them successfully. This is partly because as any firm grows in size, it becomes more and more bureaucratic unless it has a founder or CEO with an iron focus on efficiency and winning.
This development over time of processes, procedures and controls to regulate change and coordinate across functions, countries and customer types, can lead to a box-ticking mentality that stifles creativity.
What the rise of AI and its rapid increase in capability means for innovation is that bottom-up trial-and-error needs to be met with top-down understanding and enabling of creative freedom - otherwise innovation will die on the vine.
This isn’t to argue that we don’t need rules - we definitely do - but we need to think of them as guardrails and not iron-clad must-dos that override the natural entrepreneurial trial-and-error process that delivers enormous value for society.
The Incumbent's Burden
In regulated industries especially, there’s nothing inherently objectionable about good corporate governance. You need rules and checks and balances to protect shareholders and meet wider societal expectations. You want to make sure someone did the testing before rolling out a new credit risk model or customer complaint process.
The risk of having all of these frameworks and their assorted processes and controls is that following the process and engaging in theatrics to demonstrate that the process was followed to the letter can become more important than achieving or delivering real outcomes to customers.
A good example of this is how some firms responded to COVID-19 - many pivoted almost instantly to remote work and had few issues because they had the foresight to be well prepared for business continuity or disaster recovery type scenarios - the challenges they typically faced were vastly higher demand on remote access portals and the like, but they had thought ahead and were mostly fine.
Many other firms didn’t have that experience. A lot of their focus on business continuity and disaster recovery was cosmetic - they definitely had the frameworks and the printed out procedures to follow - they may have even warranted the same to their regulators and auditors - but they hadn’t invested the money required in building out the technology to support their policy ambitions. Some were even stuck in the middle of on-premise to cloud migrations and were stuck in hybrid mediocrity that was exposed under pressure.
A lot of innovation in the corporate world is like this: there is a goal, but it’s not bankrolled properly. When you have to fix legacy technology debt it’s not an overnight process. I sometimes think a lot of market valuations of companies with a lot of technical debt are interesting, especially in regulated fields, like, that’s cool your market cap is such, but you have billions of dollars or tens of billions of dollars in technology underspend? When will that be sorted out?
The challenge is the internal hurdles faced by innovators - a lot of process and politics has to be navigated before permission is even granted to try something so neutered and watered down it’s likely pointless. I think that most technology debt or legacy debt is just a symptom of the bureaucracy that has arisen inside big firms - from old banks to new big tech firms alike. The funny thing is that if these firms enabled AI tooling and let their engineers get on with it, they could do amazing things, unfortunately it’s “not in the budget” this financial year.
The Challenger’s Advantage
One reason why startups do cool things is because they haven’t developed the cruft of process and interaction overhead that kills off experimentation. Without entrepreneurial drive, vision and embrace of trial-and-error, so many things in our world would never have happened.
The main advantage of being a challenger in the AI-era is that freedom and speed is both wider in the sense that a tiny startup has so much more leverage from AI tooling, and faster because speed is limited less by human capability than by the human’s ability to structure and nudge AI tools to deliver the results the startup wants to achieve.
The future is “tiny teams” who can leverage AI to create working products for customers in much less time and at much lower cost. A challenger isn’t going to waste time filling out project governance paperwork. They have a vision and issues in Linear and they are smashing through the backlog at pace.
Startups still need to do testing and have standards and keep quality high - but they are able to sprint towards the next milestone of innovation instead of trudging through the mud weighted down by process. They’ll need to develop their own controls and guardrails but these will be much more highly automated and responsive to supporting innovation at pace.
Shaking Up Finance: Crypto and Stablecoins
A good example of this challenger vs incumbent pattern is just in stablecoin usage for retail cross-border flows. I’ve written elsewhere about the use of stablecoins in more institutional contexts, but the fact of the matter is that compared to Western Union or sending a wire transfer via a bank or even using a “fintech” like Wise or Revolut, the cheapest/fastest/best way of moving money cross-border in the presence of a fiat on-off ramp in either country is sending a stablecoin.
The entire legacy system of correspondent banking and multiple intermediaries taking large percentages of value at each step of the process has been sidestepped. Similar to how Uber stretched the rules globally then nudged regulation to a position that favoured it and other ride share firms, stablecoins have already won cross-border retail flow - especially now that debit card products attached to wallets with stablecoins are so widely available.
The Next Challenge: Getting AI Right
AI presents huge opportunities. Every big firm has a backlog of customer needs it hasn’t deliver on because it can’t afford the technology spend. If they upgraded their AI tooling and let their best engineers loose, they could rapidly solve long-standing problems that lead to things like manual processes and workarounds driving up operational risk and leading to customer disappointment.
I think the major risk with AI is excessive worrying about potential problems. Many of the common objections to using AI in the workplace have already been asked and answered: you can use models without them training on your data, you can deploy local open source models if you so choose, you can maintain all of your data sovereignty and restrict internal usage - all of your controls can be met if you have capable enough folks in your technology function who are “at the frontier” and not lagging behind.
The problem though with trying to do a top-down controlled-and-structured rollout of AI is that you lose the innovation from bottom-up problem solving. People who aren’t that technical or slightly technical are close to problems in your business that could easily be automated if an engineer gave them 30 minutes attention but the internal structure of big firms makes this sort of engineer-and-frontline worker alignment impossible because of the fixed cost of project overhead and controls around change.
Imagine instead a world where inside agreed guardrails, people in customer service were allowed to build their own internal agents to help them solve problems faster and people in product were able to get to Minimum Viable Feature type working prototypes without needing to talk to designers or engineers.
Too much caution around how you control innovation will just give more room for challengers to win. Take a highly regulated business like auto insurance: there is nothing stopping a tiny startup team vibe coding an end-to-end auto insurance flow and getting it to market other than innovation.
You can use AI to build the mobile app. You can use AI to build the underwriting engine. You can use AI to build the customer claim flows. You can use AI to manage the payments and investment of premiums in T-bills. The entire operating model of a startup auto insurer could be hacked together in a couple of weeks using AI - getting the regulatory approvals is the hurdle (for good reason - but that won’t be a good enough reason for much longer).
From a creative destruction point of view, this is the real risk of AI disruption: in an age where a viral TikTok can drive unimaginable sums of marginal revenue to an app or consumer brand - how is it not feasible that challenger firms that are too slow to move could lose entire categories of product to AI-first challengers?
I think about the rise of the digital broker-dealers over the last decade - built on APIs from US brokers - they just built marketing wrappers that made it easier for retail investors to buy US stocks and trade US options and added on FX margin and other products over time - these sorts of tools now face the next level of disruption as startups will again - use AI tooling - to replicate their entire capability set within days or weeks.
Where To Next?
I’d be interested to hear from you whether you think the rise of AI challengers is realistic or not and whether you agree or disagree with my assessment that box-ticking and a focus on risk management will hurt incumbents, especially in highly regulated industries.
When I left my corporate job a few months ago, I knew I wanted to spend more time investigating and experimenting with AI tooling because I could see the writing on the wall - knowledge work is at a cross-roads - and the rapid increase in AI capability will lead to enormous disruption in the coming years.
Many highly educated and highly skilled folks who think they are most insulated from this technology are actually the most exposed - developers, doctors, lawyers, engineers, consultants.
We recently welcomed our second child, and the amount of dinosaur technology still in complete control of the Australian public-private healthcare system is amazing - pen, paper, systems that don’t talk to each other, multiple channels of communication between doctor and patient that aren’t linked, multiple forms and invoices across different payment channels - a complete and utter customer experience nightmare. The medical care and staff were great - everything else, ouch!
So maybe my new litmus test for AI agent capability is when I can use voice prompts just chatting to a major consumer facing model to navigate a system like this where it goes off and interacts with the right agents at the hospital and the insurer and all of the agents share medical information securely so you don’t need to repeat yourself and that patient identify verification is better than “tell me your D.O.B”!
The whole interacting-with-the-medical-system experience highlighted how much low hanging fruit there is for builders in healthcare innovation to fix when they can deploy AI tooling as leverage to solve problems.
There is a much better faster cheaper future possible - it doesn’t all have to be doom and gloom around job loss and automation of knowledge work - removal of drudgery and reduction in error rates is another set of possible outcomes.
Enjoy your weekend,
Brennan
Edison tried hundreds of times before he came up with a working prototype of the light bulb - our very symbol of any great idea. If he'd been fettered by top-down procedures, he'd never have seen the light of the day!