Why Now: The State of AI in 2026
The numbers behind the AI shift, from capital and jobs to regulation and trust, and why governed, responsible adoption is the work that actually matters.
Rich Hay
Co-Founder
On this page
Every few weeks someone asks me whether the AI wave is overblown. It is a fair question, given how much noise there is. Rather than offer another opinion, I want to lay out the numbers. Put the capital, the jobs data, the regulation and the consumer signals side by side and the picture is fairly clear: this is a structural shift, and it is already underway.
Here is what the evidence says, and what I think it means for anyone trying to adopt AI responsibly.
The scale of the capital
Start with the money, because spending on this scale tells you what people actually believe. The World Economic Forum’s Four Futures for Jobs in the New Economy projects global AI capital expenditure of more than $1.3 trillion between 2025 and 2030. That is infrastructure spend, the kind of commitment organisations make when they expect a technology to be a permanent part of how they operate.
You can hold whatever view you like on any single vendor’s valuation. Capital at that scale still changes the economics for everyone, and it makes the case for carrying on the old way harder to defend each year.
How the work changes
The most useful data on what AI does to work comes from the WEF’s Future of Jobs Report 2025, a survey of over 1,000 employers representing more than 14 million workers.
- 86% of employers expect AI and information-processing technologies to transform their business by 2030, the single biggest driver of change they identified.
- The same report forecasts 170 million new jobs created and 92 million displaced by 2030: a net gain of 78 million, but a churn of 262 million roles. That is the largest labour transformation since the industrial revolution.
- AI and machine-learning specialists are among the fastest-growing roles anywhere in the economy.
That last point matters. The headline fear is that AI takes the jobs; the data suggests it reshapes them instead. Every AI use case I have seen inside a large enterprise has created more engineering, governance and integration work rather than less, because someone has to make decades of messy data, legacy systems and compliance obligations fit for a machine to use safely. The work does not disappear so much as move up the stack.
The regulation is in force
If capital is the accelerator, regulation sets the direction. In 2026, governed AI is a requirement.
The EU AI Act is in force, with obligations for high-risk AI systems landing from August 2026. It applies extraterritorially, reaching any organisation whose AI touches the EU market, and the top tier of penalties runs to €35 million or 7% of worldwide annual turnover, whichever is higher. Alongside it sit DORA, NIS2, the EU Data Act and the Data Governance Act, all now in force or imminent.
Closer to home, the FCA’s Mills Review framed firm-level AI governance as a source of competitive advantage rather than a compliance overhead, which is the right way to look at it. Treat governance as scaffolding you put up early so you can build faster later, and you will move quicker than the firms bolting it on in a panic before an audit.
Trust is the bottleneck
The technology tends to be ready before people are, and that gap will shape the next few years. Financial services shows it clearly. The FCA Mills Review found that around one in five UK adults would already delegate a financial decision to an autonomous AI, a figure that rises among people who already use AI tools. Trust is uneven, though: consumers are far happier to let AI explain something than to let it act on their data.
That is the gap where careful, well-governed AI earns its place. You do not close it with a better model; you close it with evidence: explainability, audit trails, human oversight and data you can stand behind.
What this means in practice
Put those four together, trillion-dollar capital, a labour market in churn, firm regulatory deadlines and a trust gap, and the practical question shifts. For most organisations it is no longer whether to adopt AI, but how to do so without coming unstuck.
That “how” is the work we care about at bigspark, and it shapes the way we build:
- Data that is fit for AI. Most AI projects fail on the data underneath the model, not the model itself. Getting structured, governed, well-understood data in place is unglamorous and decisive, and it has been our focus since day one.
- Synthetic data for safe testing. You cannot train, test or evaluate AI on real regulated customer data without taking on risk. Our platform Aizle generates statistically realistic data with no real personal information, the same approach that supports the FCA’s Digital Sandbox.
- Governance you can evidence. Our AI-governance platform Prism tracks AI systems against rulesets and regulatory requirements, so “we are compliant” becomes something you can show rather than assert.
None of this is about chasing hype. It is the ordinary, rigorous groundwork that lets an organisation adopt AI at scale without nasty surprises. The capital shows the shift is real, the jobs data shows the work is changing, the regulation makes governance mandatory, and the trust gap points to who will come out ahead: the organisations that take responsibility seriously.
That is why now.