Skip to main content
Welcome to Deloitte
If we have selected the wrong experience for you, please change it above.

UK perspective: Preparing for a silicon based workforce

Preparing for a silicon-based workforce

Agentic AI can be a catalyst for transformational growth and competitive advantage within UK organisations. But how do CIOs overcome the hurdles to seize the opportunity?

The momentum is undeniable: Gartner predicts that 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028, up from none in 2024. But few organisations are realising true ROI.

Those seeing the most success are treating this as a business transformation opportunity, re-thinking operating models, re-designing organisations and building agentic-compatible architectures. It’s a major shift in how work happens across the enterprise.

The ‘reality gap’ and ‘paralysis point’ are blocks to ROI

Many organisations are seeing a 12 to 18-month lag in ROI, while some are experiencing double the cost of technology per head, without clear value capture. There are fundamental reasons why:

Traditional enterprise systems weren’t designed for AI agents.

Current architectures create friction – most organisational data isn’t positioned to be consumed by AI agents that rely on accurate data and sufficient business context for reasoning, planning and decision making.

Enterprises struggle to establish appropriate oversight for systems designed to operate autonomously. Traditional governance models are simply not fit for purpose.

For CIOs, there can be a paralysis point – it can feel like you are sprinting in the fog.

  1. Where do you start and how fast should you go?
  2. Which providers and partners should you back? Where should you build your own capability?
  3. How do you prepare your data?
  4. Can you justify investment and manage costs, given ambiguous and immature pricing models for consumption-based AI services, for example API tokens? What happens if the price of the token skyrockets in the future?

The lack of visibility can make commitment challenging.

Real value lies beyond cost reduction

In the UK and the rest of Europe there appears to be a stronger focus on using AI to cut costs, while we typically see the US leaning more on AI to unlock growth. Concentrating on cost savings alone can jeopardise long-term investment. Underinvesting in growth opportunities could mean the UK gets left behind while others reap the benefits from greater AI innovation.

To help unlock growth opportunities, it’s important to consider the following:

To recognise maximum ROI, find opportunities to generate revenue, for example by leveraging AI agents to support customer acquisition, better conversion or cross and upsell opportunities.

How can you use agents to be even better where you’re already winning? Agentic AI should fuel your strategy and create new opportunities.

Is it your data, your people, your relationships? Pinpoint this, then deploy agents to double down to optimise your greatest assets.

Ask where can you create capacity – and what could you do with it?

Change has to be systemic – or it won’t stick

“We adopted some AI tools” is not a strategy. Given the rising cost of AI inference and token economics, business cases that save 10% to 15% are unlikely to be economical in the face of unpredictable tokenomics.

AI adoption is now as much a capital allocation challenge as it is a technology or change management issue. Strategic investment decisions are required to ensure resources are directed towards initiatives with the greatest potential impact.

Systemic change in the agentic age means embedding agents into the very fabric of the organisation, with robust structures for regulation and governance. This includes defining decision ownership, creating roles such as Agent Supervisors, ensuring agents comply with regulatory requirements and designing end-to-end lifecycles as we have with human employees. Ongoing monitoring of agent performance, costs and data sovereignty is essential, supported by resilient frameworks for oversight and ethical management.

Curate an ecosystem for competitive advantage

Your technology ecosystem can become a competitive advantage, particularly as cost and data sovereignty become more crucial.

When curating your tech ecosystem, you should:

Identify where architecture, integration and data flows could accelerate – or constrain – agent adoption.

Consider not just solutions but also data security, pricing and long-term value.

These partnerships will shape your ability to innovate and compete.

The organisations that act now will be better placed to shape their ecosystems deliberately, rather than absorb higher costs, weaker negotiating positions and reduced strategic flexibility later.

Deloitte is helping CIOs identify where real ROI lies and move from paralysis to strategic implementation.

The evolution to agentic AI will see a wholesale change in how enterprises operate, compete and create value. Much like the dawn of digital, with the rise of self-service and the app economy, it will reveal a wealth of new economic opportunities. But compared to digitisation, this transition is moving at a far greater pace.

Success relies on mastering several vital elements:

  1. The foundations of agent-native process design.
  2. Multi-agent orchestration.
  3. Collaboration between humans and agents.
  4. Understanding, deploying and managing a silicon workforce as part of a hybrid approach.

To unlock maximum value, recognise the unique strengths of your organisation then focus capital, plus the power of both humans and agents, in that direction. Those that do, will thrive.

A CIO’s perspective

Salesforce CIO Paul O’Sullivan shares his insights on the agentic ‘rollercoaster’ and what tech leaders need to prioritise to grasp this opportunity.

“I ask CIOs one simple question, ‘if you were going to build an agentic enterprise from scratch, would you do it based on the business processes you have today?’ The immediate answer is no.”

When GenAI exploded on to the scene, it captured imaginations worldwide. Since then, we’ve been through the hype cycle, felt the fear factor and come out the other side. Now, from the proliferation of pilots, we’re starting to see real value creation.

Agentic AI isn’t faster automation; it’s agents taking the initiative. The opportunity is enormous, but it needs a clear compass.

As a CIO, Paul’s AI journey has been guided by two principles:

  1. Technology for technology’s sake will never go anywhere. There has to be a direct benefit for customers or employees.
  2. Cognitive offloading – implementing AI and moving on – cannot be an option. Explainability is fundamental.

Ultimately, for Paul, agentic transformation will rely on humans.

“For me, people are still at the heart of everything I do in my team and in my business. They are our greatest asset if we support and nurture them.”

Embracing a hybrid future

Leaders must see AI agents as a 24/7, silicon-based workforce that complements human employees.

This transition from augmentation to enterprise-wide autonomy calls for new management approaches, human-AI collaboration models and talent strategies. It requires auditability and observability – people must be able to observe and interact with agents, correcting them if necessary. There must be feedback loops, coaching, reinforcement learning and human-centred design.

Foundations for agentic success

A catalyst for change that’s both horizontal (across channels and customer touchpoints) and vertical (across systems of engagement, intelligence, work and agency), agentic AI has the potential to impact everything.

Agents can save us thousands of hours of effort. They can deliver frictionless, hyper-personalised customer experiences that build trust and loyalty. But alongside a hybrid workforce, success requires some foundational elements.

CIOs should shift from basic IT enhancements to reimagining processes and operating models, focusing on agent-compatible workflows rather than simple automation. Strategic redesign will be a differentiator.

Robust data architecture and strong oversight are essential. Paul believes governance and control are at the root of everything. From there, the priority is data – your USP – and how you and your agents use it to proactively engage with customers.

“If you can harness the power of your data to delight customers through incredible experiences, you’re going to win all day, every day.”

From pilot to production: The evolution of Agentforce

Agentforce is now actively running across Salesforce Help (Salesforce’s online helpdesk), handling more than 90,000 customer conversations a week. For Paul, the pilot proved the importance of real-world testing and feedback.

“For us, this has been about leading from the front as a business that’s talking to customers about the importance of being an agentic enterprise. We’ve had bumps along the way, but we’ve learned valuable lessons that are informing our strategy.

Before releasing Agentforce, we did all the right things – we generated thousands of synthetic test cases to make sure responses were accurate. But there’s nothing quite like consumer interaction. When we launched it, we quickly realised that customers were asking different questions.

For example, we wanted to minimise hallucinations and put a very tight guardrail in place, so the AI wouldn’t refer to our competitors. But in reality, customers wanted to know how to integrate Salesforce with other products and services, and they weren’t getting the responses they needed.

We made some really smart changes and, suddenly, we were seeing higher levels of accuracy in our response rates. It went from low 60% to 86%.

AI is an iterative process. We added the trust layer but this isn’t like legacy IT; you can’t build it, deploy it and walk away. Agents are an extension of your enterprise, so you need to train and refine them like you would your human workforce. We all do compliance training and we all know the corporate messaging – agents need to know these things too.”

Driving innovation and harnessing creativity are now essential expectations for tech leaders. Once seen as delivery engines, CIOs need to be more knowledgeable about the broader context and impact. They need to be a champion for responsible business and a brand ambassador; a strategist, educator, advisor and advocate.

The priorities for Agentic AI transformation are clear:

  1. Understand the problem: Projects fail when they’re not linked to business value – don’t do anything because of a fear of missing out.
  2. Know when to start: If programmes are in flight today, don’t be afraid to pivot so they too can become agentic.
  3. Create the headspace to look holistically: Don’t just point and shoot. Consider enterprise and value chain transformation and how external providers, like Deloitte and Salesforce, can bring outside-in perspectives.
  4. Think about controls, guardrails and the human impact: Look beyond the technical aspects and focus on the operating model, change management and human-centered design.
  5. Drive strategic alignment: There will always be areas of an organisation that resist change. Without strategic alignment, you can only go as fast as your slowest-moving part.

This CIO perspective explores one of the key trends featured in our Global Tech Trends report: The agentic reality check: Preparing for a silicon-based workforce. Read the full trend here.

Did you find this useful?

Thanks for your feedback