16 Jun 2026
by Jiahao Sun

Agentic AI needs privacy-preserving infrastructure for real-world success

Read this guest blog by Jiahao Sun, CEO of FLock.io, for techUK’s Tech and Innovation Focus Week 2026.

Agentic AI is being deployed by forward-looking IT departments aiming for genAI-enabled workforce augmentation. Traditional automation was the beginning, streamlining repetitive tasks and improving efficiency, but agentic AI takes it a step further. Now, organisations are seeking self-managing systems that can adapt dynamically and understand intent.  

For autonomous AI systems to progress from concept to reality, robust data privacy protections are non-negotiable. This transition from demos to responsible, scaled use cases requires building the right underlying infrastructure. 

What is agentic AI? 

Most companies are still at the elementary stage of content generators, chatbots, and digital assistants based on prompts and predefined rules. More adventurous companies are gaining new core competencies such as decision support analysis. But the winning 35% of organisations told BCG they are already using agentic AI, and another 44% said they plan to do so soon.  

Now, the goal is for agents to be trusted to use high-level goals to plan and execute entire workflows and apply judgement shaped by a company’s knowledge, with minimal human intervention.Agentic AI’s name comes from the broader concept of agency: a capacity to act independently, differing from passive systems that are merely reactive. The three key characteristics setting agentic systems apart from their predecessors are:  

  1. Goal-oriented autonomy that enables independent decision-making without continuous human guidance  
  2. Contextual reasoning from integrating memory and planning 
  3. Multi-agent collaboration 

Infrastructure and robust governance are the bottleneck 

Despite the momentum, we are seeing a bottleneck: 83% of organisations state they require immediate infrastructure upgrades to support production-grade autonomous systems, according to BCG. They need to embed the company’s proprietary intelligence and business context without exposing sensitive data, then prepare for a radically altered enterprise. A lasting competitive advantage will not be secured by building superficial application layers on top of centralised monopolies. 

Centralised AI fails privacy requirements 

Privacy-preserving technology is needed for agents to move from demos to real-world use cases. The problem is that traditional, centralised AI introduces privacy and security risks when sensitive data is sent to external or centralised environments for processing. In certain sectors with highly sensitive data like healthcare and defence, the risks this carries are particularly acute. 

By contrast, federated learning (FL) lets multiple entities collaboratively train a model and only collects the updates, not the raw data, instead of sending sensitive data to a single central location. It brings the training process directly to the data. FLock.io demonstrated this privacy-preserving machine learning technique with Sarawak AI Centre – a governmental organisation in Malaysia – as a sovereign AI proof of concept for the public sector. 

Furthermore, decentralised agent swarm networks are explained in a paper published by FLock.io last year. While there is a growing interest in multi-agent systems enabling autonomous collaboration among agents, most still have centralisation and static workflows. FLock.io’s design allows the network to be communicative without compromising privacy. 

With more autonomy comes greater need for governance  

As things stand, most companies are building processes with a human-in-the-loop who has the ultimate authority. As AI takes on more decision-making, organisations are expecting a shift in governance. Systems are becoming more autonomous, allowing them to adapt when faced with unexpected scenarios. But with that comes a need for governance, to carefully monitor agents and ensure they act within boundaries. To ensure this, FLock.io has switched from simple prompt engineering to harness engineering.  

The name of harness engineering comes from horseriding – a horse is powerful, but without reins it might run off in all sorts of directions. Likewise, an engineer can wrap a harness around a model to keep it in check, namely the infrastructure you build around the model (systems, guardrails, testing, tools, constraints and feedback loops). The goal is to ensure AI agents deliver reliable, production-grade performance.  

The most significant opportunities for agentic AI adoption across UK sectors 

Sectors that are heavily reliant on administrative processes see the most significant opportunities for agentic AI adoption. 

Public sector  

Governmental organisations, from healthcare to defence, would benefit from using agentic AI to reduce administrative strain on public services. But despite the demonstrable benefits of AI in every industry, the privacy risks make it easy to understand why the public sector has been wary of implementing it. 

With rising pressures on budgets and staff, agentic AI offers a more adaptive and autonomous solution to optimise beds, equipment, booking, staff schedules and space. Early trials, such as the Manchester University NHS Foundation Trust’s collaboration with Microsoft, suggest that autonomous agents can reduce the time staff spend on routine HR and administrative tasks by up to half, allowing clinicians to focus entirely on direct patient care. Patient waiting times have dropped by up to 30% in some AI-enabled settings, yet challenges remain. 

Data protection remains a major concern for healthcare providers adopting new digital technologies. Sovereign AI is the only way forward for governments – that is, to independently develop, deploy, and govern AI using their own infrastructure and data instead of relying on a foreign corporation. 

Finance  

In wealth management, accounting, banking, and building societies, multi-agent swarms can autonomously manage complex compliance and regulatory workflows. By analysing transaction histories locally and applying judgements shaped by an institution’s protected knowledge base, agents can detect anomalies in real time without exposing customer data to external risks. 

Financial operations are governed by ultra-strict mandates (such as FCA rules and anti-money laundering regulations). If an agentic system requires exposure to raw transaction logs, personal identifiable information (PII), or institutional trading records to detect market anomalies, routing such data through third-party centralised systems can create significant compliance and security risks. 

Logistics  

Agentic AI can independently orchestrate end-to-end supply chains. It can predict customs bottlenecks at UK ports, dynamically rerouting freight, and autonomously manage supplier constraints based on real-time disruption data. Centralising this data risks catastrophic commercial leakage, exposing trade secrets to competitors utilizing the same foundation model. 

By championing decentralised, privacy-preserving, and securely harnessed agentic networks, the UK can foster an ecosystem where enterprises confidently transition from sandbox pilots to responsible, scaled, real-world deployment. 

Author

Jiahao Sun

Jiahao Sun

Founder and CEO, FLock.io



ai_icon_badge_stroke 2pt final.png

techUK - Seizing the AI Opportunity

The UK is a global leader in AI innovation, development and adoption.

AI has the potential to boost UK GDP by £550 billion by 2035, making adoption an urgent economic priority. techUK and our members are committed to working with the Government to turn the AI Opportunities Action Plan into reality. Together we can ensure the UK seizes the opportunities presented by AI technology and continues to be a world leader in AI development.  

Get involved: techUK runs a busy calendar of activities including events, reports, and insights to demonstrate some of the most significant AI opportunities for the UK. Our AI Hub is where you will find details of all upcoming activity. We also send a monthly AI newsletter which you can subscribe to here.


Upcoming AI Events

Latest news and insights

Subscribe to our AI newsletter

AI and Data Analytics updates

Sign-up to our monthly newsletter to get the latest updates and opportunities from our AI and Data Analytics Programme straight to your inbox.

Contact the team

Kir Nuthi

Kir Nuthi

Head of AI and Data, techUK

Usman Ikhlaq

Usman Ikhlaq

Programme Manager - Artificial Intelligence, techUK

Sue Daley OBE

Sue Daley OBE

Director, Technology and Innovation

Visit our AI Hub - the home of all our AI content:

Seizing the AI Opportunity generic AI campaign card.jpg

 

Enquire about membership:

Become a techUK member

Our members develop strong networks, build meaningful partnerships and grow their businesses as we all work together to create a thriving environment where industry, government and stakeholders come together to realise the positive outcomes tech can deliver.

Learn more


 

 

 

 

 

Authors

Jiahao Sun

Jiahao Sun

Founder and CEO, FLock.io