Artificial intelligence

PwC US has been working with a top 20 global pharmaceutical company to improve the design of their clinical trials for major drugs which could potentially benefit millions of people around the world.

The PwC US team used data and analytics together with the power of GenAI to design a new clinical trial model, building an integrated technology platform which utilised the capabilities of PwC Alliance partners Amazon Web Services and Microsoft. These capabilities enable pharma companies to design clinical trials leveraging data and AI.

The optimised clinical trial design delivered big improvements in three key areas. It dramatically sped up the time taken to test the drug, cutting the length of the trial by 20-25%. It also considerably reduced the cost of the process, and it enhanced the quality of the trial findings by providing greater precision.

Indeed the improvements were so significant that they are estimated to have delivered a 60-70% boost in productivity compared to the old design. The optimised design also delivered a revenue boost in excess of $100 million to the pharmaceutical company because they were able to bring their new drug to market far more quickly than they would normally have been able to.

Most importantly, the optimised design has a wide potential application and can be used to conduct faster and more efficient clinical trials for all kinds of drugs across the pharmaceutical industry, helping to unlock the potential of drug discovery and ultimately benefitting patients everywhere.

A team at PwC India has created a tool to tackle these issues with Cyber Secure GenAI, a security product which has been specifically designed to address the unique challenges and vulnerabilities emerging in the world of Generative AI.

Cyber Secure GenAI is designed to enable enterprises to safeguard and implement guardrails in their Gen AI Large Language Model (LLM)-based applications.

The technology platform provides protection through six modular security membranes including input checks, PII (Personal Information Identifier) checks, datastore checks, generated output checks, connected app checks and hallucination checks. The input membrane, for example, serves as the first line of defence, safeguarding against unauthorised access attempts and malicious inputs. It checks for various threats such as prompt injections, unauthorized code execution, and sensitive data exposure. The output membrane meanwhile checks the legitimacy and accuracy of outgoing data and responses from LLM applications. The other membranes, namely PII, Data and connected apps, perform modular checks at the connected database, data pipeline and other connected applications (Customer Relationship Management (CRM)), Enterprise Resource Planning (ERP) etc) with the GenAI LLM application/s.

Crucially, Cyber Secure GenAI also features an industry-first hallucination checker, an advanced mechanism that detects false outputs and helps to establish a secure and reliable environment for AI applications. This increases the accuracy and credibility of application outputs, helping to prevent the spread of misinformation or distortions.

These six membranes provide a holistic security guardrail for LLM applications — modularly integrated as per the business need from simple to complex architectures. The solution builds trust in adopting newer and innovative technologies such as GenAI within the organisational landscape.

Cyber Secure GenAI will be embedded in all global PwC GenAI offerings to enhance PwC's global cybersecurity services.

The urgency, unpredictability and high costs of a refugee influx crisis make it hard to manage. The public authorities responsible for helping refugees to integrate into local communities have to allocate limited resources quickly, while ensuring that the allocation decisions they make align with the interests of both the refugees and the local community. This is despite often lacking specialist guidance and insights.

As part of PwC’s Solvers Challenge, an annual global challenge that encourages collaboration on innovative solutions to complex problems, a PwC Netherlands team has developed a model which combines data, research, and expert knowledge with a powerful AI model to provide authorities with quantitative insights into potential interventions so that they can decide the best actions to take. The model does this by translating the local community's priorities into weights of key performance indicators (KPIs) and then relating these directly to possible interventions. The interventions proposed by the AI model in most cases outperform those proposed by human beings.

The possible interventions that this model assesses are a mix of investments in education, language and integration, sports, access to transport and employment. The model is based on the generic framework of PwC's Responsible Business Simulator, a modelling tool which helps users make strategic decisions based on facts rather than instincts; however, it is possible to adjust the settings and approach to address refugee crises in other countries.

The current version of the model was designed to support Dutch municipalities in accommodating Ukrainian refugees. It considers the following KPIs for both Ukrainians and local residents: feeling of safety, financial security, social cohesion, health and labour participation, and economic impact on the local area.

The goal of the model is to calculate and compare the impact of the given interventions, thereby allowing decision makers such as governments, municipalities and NGOs to make data-driven decisions in this complex area. This builds greater trust in these organisations and wider society.