As we enter a tumultuous World in 2026 I thought it worth reflecting on the impact AI made in 2025 and what, if any lessons we learnt along the way,
By the end of 2025, the narrative around artificial intelligence shifted. Early excitement, hype and experimentation gave way to hard lessons in operational value, governance, and integration. Organisations discovered that deploying AI at scale is fundamentally a business transformation challenge and not just a technology change. Companies need to want to change if they are to leverage AI as an advantage. They have stopped ‘just going along with it’ to now looking for the economic benefits of using it.
There is also, in my view a slight undercurrent of caution and when AI is discussed in context of, for example a legal firm, there seems to be a reluctance to embrace the many benefits quite possibly because of the potential cultural/personal impact of staff/clients.
Below are five key lessons from 2025 that could shape how business leaders approach AI in 2026.
1. Proofs of Concept Abound—but Many Failed to Deliver
In 2025, the transition from pilots to scalable systems was a persistent challenge for organisations. Many AI initiatives that looked promising in controlled environments stalled when put into production.
Context and Source
McKinsey’s 2025 state of AI survey highlighted that while AI adoption increased, the transition from pilots to scaled impact remained a work in progress in most organisations. (McKinsey & Company)
Illustrative Example
Numerous studies throughout 2025 confirmed the pattern: companies build numerous PoCs, but only a small percentage reach sustained business outcomes because they’re not designed with integration, data quality, or ownership frameworks in mind. Reports indicate roughly only 54% of AI projects make it to production, with organisational readiness cited as a major cause of failure. (IBM). This certainly rings true in my experience, where a company looks to deploy AI without understanding the underlying requirements around data, culture and processes
Why this matters
This pattern forced a reset: in 2026, successful AI initiatives should start with operational metrics and production pathways, not just exploratory pilots.
2. Fitness for Purpose Outperformed “Bigger Models”
In 2025, business leaders learned that larger models do not automatically deliver better value—especially in real operational environments where accuracy, trust, cost, and control matter more than raw capability.
Context and Source
Across industries, organisations focused on aligning AI with specific business needs and data domains rather than chasing the latest large models, as documented in year-end use-case surveys showing a shift toward targeted workflows such as fraud detection, customer service, and predictive analytics. (Databricks)
Illustrative Example
Enterprise AI blueprint collections in 2025 emphasised workflow-specific AI applications that reliably integrate with core systems such as CRM, ERP, and customer service platforms—rather than generic models running in isolation. (Google Cloud)
Why this matters
In 2026, organisations will increasingly prioritise domain-tuned models and embedding AI into existing systems rather than deploying large, standalone models that don’t provide stable ROI.
3. Redesigning Workflows Adds Real Value
Organisations that rethought their processes to leverage AI saw greater benefits than those that simply grafted AI onto existing workflows.
Context and Source
Industry surveys of AI use cases show that companies are increasingly embedding generative AI into operational workflows such as document processing, customer service automation, and decision support—transforming work rather than automating single steps. (econocom.co.uk)
Illustrative Example
Databricks’ review of top AI use cases in 2025 pointed to real-world deployments across analytics, automation, and operational improvement functions that demonstrate the value of aligning workflow redesign with AI capabilities. (Databricks)
Why this matters
In 2026, workflow integration—not task automation—will be the catalyst for measurable productivity gains.
4. Governance and Board Oversight Became Core Business Issues
As AI moved deeper into business processes, governance, compliance, and risk oversight moved into the Boardroom.
Context and Source
AI risk and governance began showing up in annual disclosures and board discussions, with nearly half of companies in some surveys reporting board-level oversight of AI risk and strategy. (Harvard Law Forum on Governance)
Further, corporate governance thought leadership emphasised that Boards must actively shape AI governance posture and review it regularly to keep pace with evolving technology risks. (WTW)
Illustrative Example
Increased AI oversight disclosures in 2025 filings demonstrate that companies are starting to embed AI risk into enterprise-wide governance and risk frameworks, covering accuracy, data use, and cybersecurity. (EY)
Why this matters
For 2026, organisations can no longer treat AI governance as a bottom-up IT project; it must be a strategic, Board-level competency.
5. Adoption Spread Rapidly—but Shadow AI and Risk Emerged
Usage of generative and agentic AI tools expanded dramatically in 2025—but this created shadow AI use and data governance challenges.
Context and Source
Reports from late 2025 indicate widespread adoption, with usage rising significantly year over year. (Netguru)
At the same time, security reports observed sharp increases in policy violations related to unsanctioned AI use—highlighting risks when employees adopt tools outside controlled enterprise environments. (TechRadar). We also saw this with the rise in Cloud file sharing sites like Dropbox and group messaging systems such as WhatsApp
Illustrative Example
Security reports highlight that as more users adopted generative AI tools, organisations experienced rising incidents where regulated or sensitive data was entered into public AI platforms—creating compliance, IP, and cyber risk challenges. (TechRadar)
Why this matters
In 2026, organisations must balance speed of adoption with governance and risk control to scale AI responsibly.
Overall: 2025 Was the Year of Operational Reality
Across industries, one message was unmistakable:
AI has stopped being a novelty and become a line-of-business issue—driven by workflow integration, risk frameworks, and operational governance.
As AI seems to be gaining traction in businesses, we might expect to see more caution or reluctance as people wrestle with the notion of hiring freezes (as AI will do those jobs now), the cultural impact of not hiring the next intake of young/future professionals and maybe the realisation that our own offices jobs may be impacted.
Let’s see what 2026 has in store…
Simon
