For a while now, AI has been stuck in a familiar place: plenty of pilots, plenty of proofs of concept, and far fewer examples of it genuinely changing how organisations operate day to day.
That’s what made Cisco’s AI Summit feel different. The conversation wasn’t about what might be possible with AI, but what’s already starting to happen as it moves into real enterprise environments.
The takeaway was simple. AI is no longer something organisations are experimenting with at the edges. It’s becoming something they need to run, govern, and trust at scale.
And in 2026, that shift is starting to feel unavoidable.
Agents are edging closer to becoming co‑workers. Infrastructure is turning into a competitive differentiator. And trust, in data, systems, and outcomes, is no longer optional.
The organisations that adapt fastest will be the ones best placed to move from curiosity to capability, and ultimately take the lead.
From pilots to production
Zooming out, a broader transition is underway. Model capability is accelerating quickly, but most organisations are still working out how, or whether, they can safely absorb it.
That tension is why 2026 is shaping up to be a pivot year. It’s the point where AI stops being treated as an interesting experiment and starts being treated as production infrastructure.
Several themes are driving that change. Agentic AI is pushing automation beyond individual tasks and into full workflows. Infrastructure demands are rising faster than many expected.
And perhaps most importantly, the gap between what AI can do and what organisations are ready to deploy continues to widen.
The constraints holding AI back
Despite the momentum, adoption isn’t frictionless. Three constraints kept resurfacing.
Infrastructure pressure
AI’s appetite for compute, power, and space is growing fast. Data centre capacity, energy availability, and even cooling systems are becoming limiting factors rather than background considerations.
There’s also a growing trend with ‘edge AI.’ Combining edge computing and AI introduces a new way for data processing away from a centralised cloud. Although AI on the edge results in more secure, low-latency data, it’s another factor to consider regarding infrastructure pressure.
Cisco is investing heavily across its stack. From chips and optics to edge token generation, these technologies help relieve some of that pressure and support AI at enterprise scale. Cisco’s own Unified Edge also takes into account the new paradigm of placing AI at a network’s edge.
Trust and confidence
Trust remains a sticking point. Concerns around data security, infrastructure resilience, agent behaviour, and governance continue to slow decision‑making.
That’s the gap Cisco is aiming to address by putting enterprise‑grade security and guardrails around AI systems from the outset, rather than bolting them on afterwards.
The data challenge
Data can be a challenge. Securing data, creating data, accurate data – all a constant factor when dealing with data. There’s already a growing trend of using synthetic, machine‑generated data in place of human-generated data.
Those concerns will have long‑term implications for how models are trained, evaluated, and trusted.
Cisco’s broader message is that constraints around machine-made data are real. However, they’re not insurmountable. With this year’s summit, the company positioned itself as a partner for organisations trying to work through them, rather than avoid them.
With 70% of Cisco’s product code now generated by AI, you can take their word for it.
High capability, low absorption
One phrase that kept resonating was “high capability, low absorption.”
In simple terms, AI is advancing faster than organisations can realistically take it on board. Models are becoming more powerful, but deployment is slowed by:
- Security reviews.
- Fragmented data.
- Governance gaps.
- Cultural hesitation.
- Systems that were never designed to support autonomous agents.
This is where competitive advantage starts to shift. It’s becoming less about access to the latest model, and more about whether an organisation can manage the change that comes with it.
OpenAI CEO Sam Altman has warned that businesses unable to adopt AI co‑workers risk falling behind. Cisco’s AI Readiness Index reinforces that concern, showing that only 13% of organisations believe they’re currently capturing real value from AI.
AI can drive efficiency, but without strong leadership it can just as easily erode trust. That’s why leadership, not technology, continues to emerge as the strongest predictor of successful adoption.
The step change with agentic AI
The conversation around AI tools is also starting to feel outdated.
Agentic AI represents a shift towards systems that can plan, reason, and act across multiple steps with minimal human input. Instead of supporting individual tasks, these systems are designed to own entire workflows.
As a result, enterprise software is being rethought. Multi‑agent orchestration and agent‑driven feedback loops are becoming more common, and there’s even talk of future “agent social networks”. These are systems that coordinate and learn collectively.
Of course, that raises new questions. Power and manufacturing constraints haven’t gone away. Data governance becomes harder, not easier.
Infrastructure moves centre stage
Even as models become more efficient, infrastructure demands continue to rise.
Low‑latency compute, sovereign cloud capabilities, and AI‑optimised networks are becoming table stakes for many organisations. AWS CEO Matt Garman has suggested that AI inference will soon be embedded in almost every application.
It’s raising the bar for how cloud platforms are designed.
At the same time, practical constraints around infrastructure remain. Ideas like space‑based data centres still feel some way off.
Against that backdrop, hyperscalers such as AWS and Microsoft are competing on speed, power, and silicon strategy. Simultaneously, countries like Saudi Arabia position themselves as emerging AI superpowers, leveraging energy and land availability as strategic advantages.
What this means for businesses in 2026
For organisations trying to prepare, the direction of travel is becoming clearer.
Leaders should be thinking about redesigning a small number of core workflows for agent‑driven execution, as suggested by Box CEO Aaron Levie. Governance, security, and access control need to be treated as foundations rather than afterthoughts.
Employees will need support to work alongside AI in ways that enhance, rather than replace, human judgement.
Early adoption of agentic tools may also create meaningful competitive advantage. It comes at a time when AI becomes embedded across enterprise communications and customer experience.
With the right tools and knowledge, businesses can keep pace with AI. Business leaders can attend industry events like GX Summit 2026 to understand how AI and other topics will influence success. Without those insights, businesses are at risk of losing this competitive advantage.
What’s clear is that AI is no longer waiting on the sidelines. In 2026, it moves decisively from experimentation into enterprise reality — and the organisations that respond thoughtfully will be the ones best placed for what comes next.
