In my last column, I wrote about the emerging research on the impact of generative AI on the ability of people in organizations to learn, develop, and maintain critical knowledge and skills. While organizations chase the promised immediate-term bottom-line benefits of GenAI, and indeed the renewed interest in ‘traditional’ AI, there is a risk that this may weaken the foundations for medium and long-term viability of organisations.
The question and call to action I posed was simple: What are we doing now to make sure that the workforce of five years’ time has the competencies and capabilities to carry out their job roles, including having the knowledge and ability to identify and remedy defective outputs and intervene in a timely manner to correct errors in processes?
Since that column was published, another question has loomed on the horizon — what happens if, or when, the AI bubble bursts? After all, in the last few weeks we’ve seen investors like Michael Burry move to short their investments in AI-related stocks, and Softbank has liquidated their position in Nvidia (albeit to invest in a different strand of AI technology). This on top of OpenAI hinting at and then walking back a need for US government underwriting of their fundraising, and Peter Thiel also dumping all his Nvidia stock.
Is It Happening Again?
My career in the telecoms industry began in 1998, during the exciting era of copper wires, 56k dial-up connections, and the early days of people figuring out how to profit from the “World Wide Web.” There was a huge rush of investment and sky-high valuations, not only for companies creating web-based services, but also for those providing the critical infrastructure, for example networking firms and fibre optic providers.
However, when growth slowed, many of these businesses collapsed after taking on heavy debt based on overly optimistic revenue forecasts. The results were chaotic, yet some companies survived, restructured, and continue to operate today in new forms and at varying scales. The dotcom crash had a devastating impact, wiping out numerous companies and forcing countless others to retrench and restructure. But it also left behind valuable assets.
Beyond advancements in internet-related technology and the understanding of how to actually do business online, many countries ended up with a surplus of ‘dark fibre’ (i.e. unused or underused network infrastructure) which has eventually helped enable faster mobile communications and widespread internet access. Modern 4G and 5G masts, for example, rely on high bandwidth backhaul connections provided by this infrastructure.
Equally, the global financial crisis that came less than a decade after the dotcom collapse caused significant global economic hardship that continues to resonate and echo. But it triggered the early realisation that data was a thing that needed to be governed because lax oversight of data was a contributory factor to poor risk management in banking and all that flowed from that. This triggered regulatory responses to mandate governance standards for data and gave initial impetus to data governance in the financial services sector and a grudging realisation that for trust in socially and societally impacting systems you inevitably need governance, rules, and regulation.
From collapse came further innovation. From failure came the foundations for future success.
Do Hype and History Rhyme?
The Irish poet Seamus Heaney’s poem “The Cure at Troy” (much quoted by various US President’s over the years) tells us that, from time to time, “hope and history rhyme.” But is the same true for hype and history? When the current AI hype deflates or pops, what will the equivalent “leave behind” be that the next wave of progress will be built on? When the financial engineering that is fuelling this investment bubble falters, what will the fallout be and what will support the recovery? Will we be left with “dark data centres” dotted around the world like so many long-abandoned shopping malls?
Slightly ironically to answer this question, I have to act a bit like a Large Language model and develop a model based on what has happened in the past and then infer the next most likely words in this story. Combining the themes from the dotcom collapse and the global financial crisis I infer a possible future state. (And, like an LLM I might be wrong or hallucinating.)
My Inference
The dotcom crash left behind a legacy of technology and technical innovations. A likely legacy of the Generative AI boom, and the big long-term win for organisations from the current AI hype cycle, will be the renewed focus on data and content not as ‘process exhaust’ but as the thing that powers organisations and enables people to get stuff done. While organisations now are adopting Generative AI with a frenzied mix of strategic intent and FOMO, if the tide goes out it will be those who invested in their data and data-related capabilities to make the AI thing work better who will thrive.
Essentially the data, and how we think about it and manage, it is the ‘technology’ in this revolution.
The long-term winners are likely to be the ones that:
- Take the opportunity to tame their disparate data and content, using AI to augment human efforts in tagging, categorising, and classifying data and improving quality of data.
- Take the chance to implement appropriate organisational and technical governance over data and AI, not as a technical competence or technology requirement, but as an organisational capability.
- Invest in redesigning how staff learn how to do their (newly augmented) jobs while still developing the competences and capabilities to be good leaders and managers and have the knowledge and skill to identify defective outputs or intervene when the process is going wrong, including critical thinking and systems thinking skills for and about data and content.
Bluntly: If the leave-behind of the AI bubble is that organisations start doing the things we’ve been trying to get them to do for the last 30+ years in data and information management, then that would be a valuable legacy. And it is a legacy which would deliver organisational and social benefits through improved understanding of how data influences decisions and outcomes that impact people.
What About Regulation?
The regulatory shifts that occurred after the global financial crisis, driving the early adoption of data governance as a formal discipline, will also likely be echoed in the future in the governance of AI. I say this notwithstanding the current wave of de-regulatory lobbying that is underway in the EU and elsewhere in respect of data privacy laws and AI regulation. After all, excessive deregulation is considered by many to be a root cause of the global financial crisis and the instability that ensued.
I personally think that there will be a recognition of regulation as an antidote to hype and a need to innovate responsibly, particularly with technologies that can have significant social and economic impacts. But we may need to see homeopathic dilution of existing regulation (as we experienced in the financial services sector in the 1990s, pre-crash) before we see the pendulum swing back.
The lasting foundation for the future lies in adopting strong organisational behaviours and good governance of data and AI. Legislation sets minimum standards, but organisations should invest in data quality, metadata management, and data governance not just for compliance, but to build robust and adaptable foundations that can withstand potential disruptions like an AI bubble collapse. And doing these things increase the probability of success in your AI and Generative AI adoption.
Keeping Your Swimming Trunks On When the Tide Goes Gut
The dotcom boom had many justifiable failures. It also had many successes that just took a lot longer than they had thought to make things work, and perhaps not in the way that they had originally envisaged. Softbank and others moving their investment focus doesn’t mean that the AI bubble is going to collapse (yet), but it should be a signal that we need to start looking for what of value might be left behind and building towards that.
For most organisations, it will be the investment in the human capital, paying down of technical and data debt, and thinking critically about data and its quality and governance, and the impact on business continuity and resilience of building in reliance on AI, Generative AI, and Agentic AI that will likely be the factors that mean your togs stay on when tide does go out.
But that’s the thing about tides. If you are still standing and hang around long enough, they usually come back in.
When the dotcom pioneers reached to develop a commercial model and innovative businesses for the internet, we got investment in communications infrastructure and the seeds of data analytics technologies and the mobile internet. Today’s pioneers are reaching to create a simulation of an intelligence that will either supplement us or supplant us. Perhaps what we’ll get is a move up the tech stack to the data layer in the wider thinking and longer term success there, even as the headline goal remains tantalisingly just out of reach?
This column is based on a blog post originally published on the Castlebridge website. Castlebridge will be discussing this theme at our annual Data Leaders’ Summit in Wexford, Ireland March 18th, 2026.
