AI feels like one of those rare technology shifts where adoption moves gradually, then suddenly all at once. We are still early, but it is increasingly obvious that the gap between organisations embracing AI and those treating it as a novelty will become material over the next few years.
The statistics already point to this acceleration. Stanford’s 2025 AI Index found that 78% of organisations globally were already using AI in at least one business function, up from 55% a year earlier[1]. McKinsey’s latest enterprise research suggests companies are now beginning to see tangible returns, with leading adopters reportedly generating roughly US$3 of value for every US$1 invested into AI initiatives[2].
What is more interesting is that adoption is no longer confined to large technology companies. Smaller businesses are increasingly moving faster than large corporates because they are less constrained by legacy systems, bureaucracy and entrenched workflows. JPMorgan research showed newer business cohorts adopting AI materially faster than older cohorts[3]. Surveys across small and medium enterprises now show AI usage becoming mainstream for customer service, marketing, administration, operations and content generation[4].
Financial institutions are also accelerating deployment. Deloitte recently noted that two-thirds of banks and insurers are now using AI or machine learning models in some capacity, with adoption among smaller banks rising from 22% to 52% since 2023[5]. Anthropic has now launched dedicated AI agents for financial institutions capable of building pitchbooks, auditing financial statements and preparing credit memos, with firms including Goldman Sachs, Visa and Citi already using the technology[6].
For me personally, the journey started much more simply like as a casual acquaintance that somehow became my work wife.
Like many people, I initially used ChatGPT almost as a replacement for Google Search. I would use it for personal and professional ad hoc research, quick summaries, drafting content and brainstorming ideas. I would often cross-check outputs with Grok to compare responses and triangulate information.
What surprised me was not necessarily that the answers were better. It was the speed.
Tasks that previously required opening 20 browser tabs, reading multiple articles and manually synthesising information could suddenly be completed in minutes. Once you experience that compression of time, it becomes very difficult to go back.
One unexpected area where AI became incredibly useful for me was health data. Over time, I started uploading blood tests and other health information into ChatGPT to analyse trends and patterns. My GP historically would often say there was “nothing really to worry about” aside from slightly elevated cholesterol levels. In contrast, AI tools were able to contextualise markers longitudinally, explain interactions between different indicators and highlight possible areas worth monitoring or improving. AI became a second layer of analysis that helped me ask better questions and engage with my own health more proactively.
At Aura Group, despite our Aura Ventures team specialising in investing in AI focussed companies (including the likes of Haast which just announced a $17.2m capital raising at a $98m valuation as a 3 year old company) our own adoption journey has been slower.
As a regulated fund manager our obligations around data security are high and so we have experimented with Microsoft Copilot, which in my experience felt relatively underwhelming and constrained compared to the frontier models now available. More recently, we have been trialling Claude by Anthropic, and the feedback internally has been overwhelmingly positive. Almost everyone I speak to who uses it seriously seems to rave about it producing amazing outcomes.
The biggest shift for me, however, came after participating in a charity auction where one of the prizes was an AI training course.
That course fundamentally changed how I thought about AI.
Prior to that, I was largely using these tools in a 1:1 way — research, summarisation, drafting content and asking questions. The training reframed AI not simply as a chatbot with some memory, but as infrastructure capable of automating parts of a professional workflow.
Instead of asking, “What can AI tell me?”, the better question became:
“What repetitive parts of my workday can AI perform autonomously?”
That mindset shift led me down the path of building and training agents collaboratively with team members to automate repetitive tasks, systemise workflows and improve operational leverage and we are still learning how to do this better. In many ways, the future value of AI may not come from individual prompts, but from building systems of agents that operate continuously in the background.
Most recently, I personally signed up to Manus AI, the increasingly talked-about agentic AI platform founded by Chinese entrepreneurs before later relocating operations to Singapore. Manus gained global attention after Meta reportedly agreed to acquire the company in a deal valued at roughly US$2 billion, before Chinese regulators intervened and ordered the transaction unwound amid escalating geopolitical tensions around AI and technology sovereignty.
What makes Manus different is that it feels inherently agentic. Rather than simply answering questions, it performs tasks.
As a bit of a funny experiment, I asked Manus to help create a family coat of arms and build a family office website for my family. After less than 30 minutes and fewer than half a dozen prompts explaining my family history (I was born Year of the Pig, my wife Rabbit, my twin daughters Tigers, my son a Snake and my beloved dog a Yorkie), our cultural heritage (Chinese), geographic affinities (traversing between Singapore and Australia) and sharing my LinkedIn profile, Manus had generated the coat of arms, built the website, automated most of the domain registration and deployment process, and had the site live, all for around $60. The outcome is here à www.ngcap.co and my family’s new coat of arms here 😊
I was genuinely impressed (sorry web design industry) and this continues to make my wife smile every time she sees it.
The implications for productivity are enormous.
Small businesses will likely benefit disproportionately because they can suddenly access capabilities that previously required teams of designers, analysts, developers, marketers and administrators. Financial institutions will use AI to compress workflows, improve compliance, automate reporting and enhance customer engagement. Professional services businesses will increasingly operate with smaller teams but materially higher output.
At the same time, there are legitimate concerns around governance, hallucinations, cybersecurity, privacy and workforce displacement. Even recent research points to a growing “transformation paradox”, where organisations know they need AI but struggle to structurally adapt around it.
But despite those challenges, I increasingly believe AI adoption is no longer optional.
It is becoming foundational.
The organisations that learn how to integrate AI into workflows, decision-making and execution today will likely build structural advantages that compound over time. Those that delay may eventually find themselves competing against businesses operating at a completely different level of speed and efficiency and ultimately be left behind.
[1] https://hai.stanford.edu/ai-index/2025-ai-index-report
[2] https://www.businessinsider.com/mckinsey-ai-adoption-return-on-investment-analysis-enterprise-2026-5
[3] https://www.jpmorganchase.com/institute/all-topics/business-growth-and-entrepreneurship/understanding-ai-use-by-small-businesses
[4] https://usmsystems.com/small-business-ai-adoption-statistics
[5] https://www.deloitte.com/dk/en/services/consulting/perspectives/ai-adoption-in-financial-institutions-balancing-growth-and-governance.html
[6] https://www.reuters.com/business/finance/anthropic-deepens-finance-push-with-10-new-ai-agents-banks-insurers-2026-05-05/