What makes agentic AI a potential gamechanger? Traditional AI models require explicit instructions or serve narrow functions. With agentic AI, autonomous agents handle multi-stage tasks, learn from data and adapt their actions accordingly, and work across different enterprise systems. These ‘digital coworkers’ can be set to work on multiple tasks, from updating databases to managing internal communications or even engaging directly with customers.
How can organisations get ahead of some of the challenges of agentic AI to ensure they realise these benefits?
Scaling agentic AI
At first glance, deploying AI agents might appear simple: just plug in a model and let it run. The reality is far more nuanced. One of the biggest hurdles is the sheer complexity of many processes. It’s easy to underestimate just how interconnected and layered workflows that have developed over time can become. A simple customer request, for example, may touch a number of different systems and multiple subprocesses, each with its own exceptions, data dependencies and rules. Mapping these intricacies is essential before AI can meaningfully participate.
Another critical barrier to adoption is cultural. There’s a common fear that AI agents are designed to replace people. That’s why it’s essential to emphasise through persistent and effective internal communication that these tools are about enhancing human productivity. They free-up teams from monotonous tasks so they can focus on higher-value work.
Reinventing, not just augmenting
Agentic AI isn’t just a tool for optimisation, it’s a powerful lever for reinvention. Rather than digitalising existing processes, start from the business outcomes you want to achieve and work backwards. For example, instead of mapping how customer service processes work today, start by asking: “What would be the fastest, most empathetic, most accurate way to solve a customer problem?” Agentic AI can then be configured to find the best pathway. That may mean it bypasses many of the steps that a legacy workflow would require. The mindset shift, from augmentation to reinvention, is where agentic AI’s true power lies.
Delivering value – agentic AI in action
What could this look like in practice? Think about the simplest HR request, such as checking on the amount of annual leave available and booking a few days off. Under the surface of this apparently straightforward action, there’s a lot going on. An agent could take on the entire chain of automated tasks, from calculating remaining days, checking team schedules, notifying managers and booking the time off.
Handling customer enquiries in a contact centre is another powerful use case. Today, up to 80% of calls are for routine enquiries. Agents could handle these easily, handing over more complex and involved cases to people. More specifically, in healthcare it’s possible to use agents that can optimise patient pathways, streamlining processes from admission to post-care follow-ups, achieving better outcomes while reducing overheads.
In financial services, we are helping one institution deploy agents that are mapping data lineage and categorisation across multiple systems. That’s essential to trace how information flows through platforms vital for compliance, security and operational efficiency. We’re also seeing agents now able to create truly personalised offerings in marketing and advertising. Such mass personalisation at scale has been an aspiration for some time. Agentic AI is making it happen.
Trust me, I’m an agent
As AI agents proliferate across departments and systems, one question looms large: How do we trust them? Trust can’t be left as an afterthought. It has to be built in from the start, all the way from rigorous testing and validation protocols to cyber security measures that prevent malicious actors from hijacking agents. Cyber and risk teams have a vital role to play in ensuring that AI solutions are not only effective but also verifiable and safe.
Once agents are in production, the right governance frameworks are also essential to monitor agent behaviour, evaluate the quality of the decisions they make, and maintain transparency with stakeholders.
Building for industry
Together with Google Cloud we are combining our strengths to solve these challenges head-on. As part of our partnership, we are working on the Agentic AI Development Program, which tasked PwC teams with developing 100 AI agents across multiple use cases and industries. These agents are engineered to address specific tasks in lines of service such as accounting and tax, regulatory, audit, legal and compliance as well as for specific industry applications - ensuring that they're practical, compliant and relevant from day one.
Why Google Cloud?
Purpose-built for scalability and speed, Google Cloud’s technology spans from infrastructure to large language models (LLMs) to orchestration tools like Vertex AI, allowing developers to go from idea to working prototype in minutes rather than months.
Crucially, Google’s technology is also designed for accessibility. Natural language interfaces and visual tools mean even non-technical users can interact with, guide and benefit from AI agents. This democratisation of development is a game-changer because it allows business users to participate in the creation process without needing to know how to code. That means faster iteration, user-centred solutions and faster time to value.
The power to solve longstanding challenges
Agentic AI is going to help solve some of the major challenges across your business, from finance and marketing to predictive maintenance and supply chain.
As you start to develop your approaches to building agents, it’s important to take a systemic view of development. If it’s left to each function to create their own agents, the results are likely to reinforce existing silos and agents will fall short of delivering the value of which they’re capable. Orchestration and integration across systems will be key.
What are the key actions to consider now? Start small but think big, using focused use cases to demonstrate value and deploy those wins to build confidence and uncover larger orchestration opportunities. Focus on business outcomes, not the tools themselves, so that business needs drive agent design. And to establish and continually earn trust, build governance and validation into the agentic AI program from day one.
Agentic AI is not simply another technology tool to add to the stack. It’s a fundamental shift in how work gets done. By moving now to understand, deploy and scale these technologies your organisation can gain a substantial competitive edge.
PwC and Google Cloud are combining world-class engineering with unmatched industry expertise so you can transform your business. If you want to learn how agentic AI could transform your organisation, get in touch with Clara van Heck, Google Cloud Alliance Director, PwC United Kingdom. Tel: +44 (0)7483 148936
Contact us
