Trust in Artificial Intelligence

PwC Switzerland’s Trust in AI services pave the way for robust models and responsible AI practices

Adopting AI with trust and responsibility

Artificial Intelligence (AI) has become an integral part of both our personal and professional lives. Yet, it often raises concerns and reservations. What if your voice-activated application records all your conversations? Or if your automated driving assistant makes decisions that seem unpredictable? To avoid such scenarios, it is crucial to develop, operate, and maintain responsible AI systems.

At PwC Switzerland, we are dedicated to unlocking the vast potential of AI while ensuring its responsible and ethical use. Our Trust in AI services are designed to instil confidence in your AI initiatives by rigorously validating and refining models to deliver optimal output. With a strong commitment to responsible AI practices, we help organisations navigate the complex landscape of artificial intelligence, promoting transparency, fairness, and accountability in the deployment of AI solutions. Trust in AI starts with PwC Switzerland.

Participate in our survey to gain a deeper understanding of your current status, compare yourself to your peers and shape the future of responsible AI.

Start Responsible AI Survey

AI offers enormous opportunities, yet integrating AI into an organisation also entails risks. We are committed to supporting our clients in effectively managing these challenges and building a responsible AI framework.

Yan BorboënPartner, Leader Digital Assurance & Trust, PwC Switzerland

Mitigating the risks of AI through transparency

As with any powerful tool, there are inherent risks in adopting AI. Companies must balance speed and confidence when implementing AI, ensuring a risk-based approach to gain competitive advantages and maintain stakeholder trust. Effective risk management is vital to fully reaping the benefits of this transformative technology.

Discussions about the ethical aspects of AI, such as the potential for deepfakes, bias, and intellectual property infringement, are crucial. And the fears about the technology’s impact on employment and its potential to replace aspects of human intelligence must be taken seriously too.

Building trust in the use of AI entails a full understanding of the risks, and a clear approach to addressing them – as well as stakeholder concerns – in an effective and compelling way. Only by emphasising human involvement, robustness, fairness, security, privacy and governance, businesses can skilfully navigate the AI landscape and realise its full potential.

Read more about the risks associated with AI in our blog How is your organisation dealing with AI?
 

Don’t trade trust for speed

Dealing with AI-generated information requires caution, curiosity, and a discerning eye. Effective risk management is the cornerstone of success. To prosper with AI and secure a competitive advantage, an organisation must establish a robust risk management framework that not only protects against potential pitfalls, but also unlocks opportunities.

Such a risk-based approach to AI also sets the stage for a positive relationship with regulators, consumers, and other stakeholders.

Fostering trust in an AI-driven world requires a holistic view and understanding of potential risks at all stages of the AI lifecycle as well as a strategic approach to proactively mitigate these risks.

Morgan BadoudDirector Digital Assurance & Trust, PwC Switzerland

How to address the key risks of AI

As we’ve been building AI solutions since 2016, we’ve identified key areas of risk and how best to deal with them:

Data protection
Key data protection considerations include conducting comprehensive risk assessments, ensuring transparency, obtaining informed consent, vigilantly monitoring for bias, implementing data retention policies, and regularly reviewing and updating these policies to ensure compliance and security.

Security concerns
Addressing security concerns in AI involves assessing potential risks, implementing robust security controls, training employees on security best practices, keeping software updated with regular patches, and using third-party validation to strengthen security measures.

Over-reliance on generated data
To mitigate the risk of over-reliance on generated data, it’s important to educate stakeholders on the limitations of AI, establish clear guidelines for its use, validate generated data against real-world data sources, and continuously monitor and review the use of generated data in decision-making processes.

Explainability
To address model explainability concerns, it’s important to acknowledge the limitations of the model and its training data. Use interpretable models whenever possible, conduct thorough model evaluations, and stay informed about the latest developments in explainable AI techniques to increase transparency and understanding.


AI in action: a wide range of applications

AI models and algorithms create original data by identifying patterns in existing data sets. These models produce outputs that reflect their training inputs but show distinctive variations. Applications are broad and diverse, including image generation, text and speech synthesis, music composition, and more. But they also carry the potential for bias if not properly trained and validated. Despite these challenges, however, AI models will play a growing role in data generation and decision-making across industries and applications.

From hyper-personalised marketing campaigns and improving customer experience, to streamlining business operations, workforce management and supply chain optimisation, to fraud detection and legal issues, AI is transforming industries.

In marketing and sales, AI can be used to craft highly personalised content across a wide range of media, including text, video, and images. AI can also help create user guides and analyse customer feedback by summarising text and extracting key insights. In addition, AI-powered chatbots can be built or enhanced to provide valuable sales support.

In operations, AI can classify text into specific categories for efficient organisation. AI can also assist in the seamless execution of tasks by generating comprehensive to-do lists. It excels at identifying and correcting errors, anomalies, and defects in text. Furthermore, AI-powered customer support chatbots provide valuable assistance in this domain.

AI offers several benefits in supply chain management. It can improve forecasting accuracy, ensuring maximum availability while minimising product obsolescence. AI also supports the optimisation of input parameters to maximise product quality. And it can anticipate and proactively prevent safety risks in manufacturing processes.

In the context of risk and legal, AI brings several advantages. It can assist in drafting and reviewing legal documents, streamlining processes, researching and comparing legal documents, and efficiently extracting critical information. It can also provide answers to knowledge-based questions by sifting through voluminous legal documents and summarising key sections for quick reference.

AI provides valuable support for a range of HR tasks. It can help generate tailored interview questions to assess candidates, ensuring a comprehensive evaluation. AI can also cleanse and format data, such as CVs, to make them easily comparable. Additionally, it can automate tasks such as email composition and response handling, enabling self-service HR functions.

In finance, AI helps predict customer debt, enabling better working capital management. AI uses machine learning techniques to detect fraudulent patterns, thereby improving security. It also provides valuable insights by explaining variances in financial forecasts and making recommendations for different forecast scenarios and related business decisions.

< Back

< Back
[+] Read More

Potential of Generative AI along the value chain

However Generative AI is not merely a technological advancement, it is a catalyst reshaping the very foundation of how businesses operate. Given its vast possibilities, Generative AI is not limited to certain business functions but impacts the entire value chain of an organization.

Finance

  • Generative AI Enabled Reporting: Automate the process of generating reports and comments for financial reports in natural language using Generative AI. It automates data analysis, generates visually appealing reports, and provides real-time insights. Additionally, it can be used to “Chat with your data” to quickly answer questions of the report user in natural language.
  • Generative AI supported Planning: Leverage (Generative) AI and external data like market- or competitive analysis to dynamically provide contextual information, explanations and charts when users select planning objects. It automates the process of offering additional insights, enhancing the planning experience.
  • Processing Automation: Generative AI can automate the matching process by reading invoices, for example, and comparing them to purchase orders and receipts. Additionally, this can be easily integrated into existing systems and workflows.

HR

  • AI enhanced Recruiting: Automate tasks involved in the recruitment process and streamline the hiring process such as job postings, resume screening, candidate sourcing, and scheduling interviews. Further Generative AI can help by formulating job descriptions. 
  • Employee Engagement: Generative AI based tools can analyze employee feedback, engagement data, or social media posts providing insights to enhance workplace satisfaction and productivity and monitor employee´s engagement.
  • Training and Development: Generative AI can create personalized training programs based on individual employee needs, improving skill development and performance.
  • Talent Pooling: Use Generative AI to aggregate talent information and streamline talent matching, enabling organizations to match project requirements with available skills quickly and effectively.

IT

  • Predictive System Monitoring: In IT infrastructure, AI can support in monitoring and optimizing resource allocation and predict equipment failures or performance issues allowing for proactive maintenance and minimizing downtime.
  • Code and Test Case Generation: Generative AI can automatically generate code snippets or entire functions based on high-level descriptions provided by developers or translate code to various programming languages. Utilize Generative AI to derive and write test scenarios based on business descriptions that are given in user stories. 

Procurement

  • Supplier Proposal Analysis: Generative AI technology can help with supplier proposal analysis by automatically generating summaries and analysis of proposals. This can save time and reduce errors, as well as provide an unbiased and objective analysis. 
  • Automated Order Confirmation Process: Use Generative AI to extract and validate relevant information from unstructured or semi-structured data sources such as e-mails and process the data automatically in the ERP system end-to-end to the customer. 
  • RFX Creation: Generative AI can transform the RFX creation process by automating document generation, optimizing content, personalizing documents, and ensuring compliance. This not only improves efficiency but also enhances the overall quality and effectiveness of the procurement process.

R&D

  • Product Innovation: Generative AI can analyze market trends, customer preferences, and historical data to suggest innovative product ideas and features, aiding R&D (Research and Development) teams in the ideation phase.
  • Simulation and Modeling: AI can assist in simulating and modeling complex scenarios, helping researchers test hypotheses, optimize designs, and accelerate the development process.
  • Automated Experimentation: Generative AI can design and conduct experiments autonomously, accelerating the R&D cycle and improving the efficiency of the research process.

Supply Chain

  • Supplier Performance & Risk Evaluation: Assess the supplier´s performance and identify potential risks and their impact with the help of Generative AI through analyzing and summarizing large amounts of data efficiently. This also helps in monitoring the supplier's compliance with regulations and environmental standards. 
  • Warehouse Operations Optimization: Optimize warehouse layout, inventory placement and workforce scheduling for maximum efficiency through balancing inventory levels and improving the order fulfillment process and adapting dynamically to changing conditions
  • Deviation Reporting: Generative AI can be used to streamline the process of deviation reporting within its warehouses. With Generative AI the system continuously scans and analyzes images and data from warehouse cameras and inventory management systems. It cross-references this data with expected inventory levels and generates real-time deviation reports.

Manufacturing

  • Maintenance Assistant: Generative AI can scan manuals efficiently so that the maintenance worker can ask what he should know related to a certain problem and the solution provides him a summary of the tasks that should help based on what it found in the manuals.
  • Report Creation: Generative AI can be used to create reports such as shift reports to ensure smooth shift hand over or quality incident reports for documentation of quality reason, in a structured and efficient way.
  • Workforce Scheduling Optimization: Generative AI can help in creating the ideal production or maintenance schedule considering different parameters such as workforce availability and skills, material and equipment availability, bottlenecks leading to improved production output and reduced costs.

Marketing & Sales

  • Automated Marketing Content Generation: Generative AI can assist in generating creative content for marketing campaigns, including images, videos, newsletters and social media posts and optimizing content for specific target audiences. 
  • Customer Segmentation: Employ data-driven segmentation and analysis to analyze customer data, encompassing purchase history, browsing behavior, and demographics. Use these insights to deliver personalized marketing campaigns and product recommendations.
  • Customer Retention: Analyze customer behavior, transaction data and engagement metrics using Generative AI, enabling organizations to provide early warnings for potential issues and leverage customer preferences to develop proactive customer retention strategies.

Customer Service

  • Virtual Agents: Generative AI powers chatbots and virtual assistants that can handle routine customer inquiries, providing immediate responses and freeing up human agents for more complex issues.
  • Customer Pattern Analysis: Leverage Generative AI to recognize patterns in customer interactions and inquiries, enabling a proactive approach in addressing customer needs and identifying areas for improvement in customer satisfaction. 

Considering the multitude and broadness of potential fields of application, it is not surprising that organizations are struggling to navigate themselves through the possibilities Generative AI has to offer. In addition, the majority of organizations are seeing the impact AI has and have understood the need for a comprehensive AI strategy and transformation roadmap. However, given common challenges, such as limited IT and innovation budgets, they rather choose to start small with selected use cases to prove the value Generative AI has before deciding about larger investments. Therefore, actively steering the portfolio to maximize ROI is critical to the success of Generative AI in organizations. Some of the key questions and challenges we see currently in the market are the following:

  • Which value-add does Generative AI offer for my organization?
  • Which Generative AI use cases are the right ones for my organization to quickly start, learn and scale?
  • How can we create effective lighthouse applications of Generative AI in the organization to kickstart the broader transformation journey?

Managing AI risks from different functions​

However Generative AI is not merely a technological advancement, it is a catalyst reshaping the very foundation of how businesses operate. Given its vast possibilities, Generative AI is not limited to certain business functions but impacts the entire value chain of an organization.

CFO

  1. Identify internal controls, SEC and Sarbanes-Oxley statutory requirements over financial reporting relevant to generative AI use cases
  2. Inventory financial and accounting tasks in the organization
  3. Plan for process optimization (do or outsource)

CISO

  1. Automate, update, and upgrade cyber countermeasures
  2. Prepare for higher-resolution threat models and insights, predictions and scenarios
  3. Put data loss prevention controls in place
  4. Protect internal/local generative AI models and associated data
     

CDO

  1. Enhance your data governance protocols
  2. Shore up privacy measures
  3. Monitor and protect sharing of organizational data to external generative AI models

IA

  1. Understand company’s goals and uses, as well as enterprise risk management’s view on risks and mitigation plans
  2. Create a plan to audit the core data sets used in training, tuning and running the system and models
  3. Design an audit plan around the generative AI systems and models
  4. Create an audit plan for output of models

CLO

  1. Limit your IP exposure
  2. Guard against improper secondary uses of data
  3. Plan for litigation and investigations

CCO

  1. Keep up with new regulations and stronger enforcement of existing regulations that apply to generative AI
  2. Map your organization’s planned use of generative AI applications to existing laws and regulations
  3. Upgrade your regulatory reporting capabilities
  4. Assess the compliance posture of your generative AI deployments
  5. Establish strong model governance processes

The four dimensions of responsible AI

While identifying risks related to AI and mitigating them, responsible AI emphasizes ethical design, transparency, fairness, and accountability in artificial intelligence systems. It aims to ensure AI respects human rights, avoids biases, maintains privacy, and operates ethically, fostering trust and benefiting society while minimizing potential harms.

Strategy

Take a deep dive into the ethics of data and AI use, and embed them in your organisational values. Stay ahead of the curve in policy and regulation by anticipating and understanding critical policy and regulatory developments, enabling alignment with compliance processes for responsible AI implementation.

 

Control

In terms of governance, it’s essential to establish monitoring mechanisms that encompass all three lines of defence within your organisation. In compliance, prioritise adherence to regulations, internal organisational policies, and industry standards. Within risk management, extend your traditional risk detection and mitigation practices to effectively address the unique risks and potential harms associated with AI technologies.

 

Responsible practices

When implementing AI, it’s important to ensure interpretability and transparency; AI models should make transparent decisions for clear understanding. Sustainability should be a priority, with the aim of reducing environmental impact. The AI systems must be robust and reliable, while also addressing potential biases and ensuring fairness. Cybersecurity is crucial to protect from threats, and privacy concerns must be addressed by data protection measures. Finally, safety is key, with AI systems designed and tested to prevent physical harm and ensure safe use.

 

Core practices

Clearly define the specific problem and assess whether AI/ML is the appropriate solution. Adhere to industry best practices and standards to ensure consistency and quality. Continuously evaluate model performance and make iterative enhancements to improve metrics. Implement ongoing monitoring to identify changes in performance and potential risks.

 


How we are leading the way to Responsible AI

The AI landscape is full of open frontiers, offering vast opportunities, but also significant challenges. A cautious approach to AI is crucial, ensuring that ethics and human input remain central. Building AI systems on a solid ethical foundation is essential, considering their impact on all users.

Our multi-disciplinary teams, with extensive experience in designing and implementing AI solutions, will guide you through the critical steps and provide ongoing support as you embark on a successful AI transformation.
Our approach to AI is based on four pillars: enablement, suitability assessment, third-party assurance, and education.

Enablement

We build digital trust by defining appropriate AI governance structures, steering processes and controls, while considering the unique requirements and risks of the specific company. This involves:

  • Risk management: defining an appropriate AI risk management approach and the related AI guidelines, as well as setting suitable KPIs for effective steering and monitoring of your AI developments
  • Processes and controls: designing processes and controls tailored to individual AI projects, taking into account internal AI guidelines, specific risks, and business needs.

Suitability assessment

Together with you, we enhance transparency and trust by thoroughly examining your AI environment, processes, and controls. Should this analysis reveal any issues, we’ll develop tailored solutions. Our approach includes defining realistic expectations and establishing a suitable risk management strategy by:

  • Maturity assessment: conducting a maturity assessment based on your AI governance and risk management structures
  • AI guidelines: evaluating the existing controls and AI guidelines while considering relevant regulations, standards and requirements – and how they differ from examples of best practice.
  • Compliant processes: defining organisational and technological measures for securing and optimising compliant processes throughout the AI lifecycle.

Third-party assurance

We guarantee adherence to the highest quality standards and transparent implementation of this technology, tailored to your specific needs. To build trust in your AI application, we focus on: 

  • evaluating the fulfilment of specific criteria that the AI service needs to meet according to the established criteria catalogue on the market, e.g. the BSI AIC4;

  • applying best practices from the relevant industries or sectors in line with established standards, framework conditions and criteria catalogues;

  • evaluating and reporting on the compliance of internal AI control systems based on an audit report in line with ISAE 3000 (Revised).

Education

We assist you in ensuring that your employees are well prepared for using AI effectively through: 

  • assessing AI usage and associated risks;

  • enhancing employee awareness with the development of industry-specific use cases and education on AI-related risks;

  • promoting the use of ethical AI and building trust in AI models.


Why we are your trusted AI partner

Trust in transformation – To remain competitive in the digital age, companies need a strategic approach to AI that focuses on security and transparency. Many AI projects fail early, often due to insufficient digital trust across the AI lifecycle, including data protection, bias in model training, and uncertainties in AI performance and robustness. Success depends on defining and implementing the right organisational and technological strategies in services and products. We guide you through the necessary steps to realise the full potential of AI and accompany you on your journey to a successful Responsible AI.

PwC is speeding up its commitment to AI technology, forging industry partnerships, focusing on training, and enhancing its client services — initiatives that have already positioned us as a forerunner in the field of AI. Our approach to implementing AI in businesses is based on building trust and achieving results. These principles form the backbone of our lab-based innovations and our tangible AI solutions. 

In addition to our direct involvement with businesses, we actively participate in discussions and bodies that are shaping the ethical standards and practices of responsible AI.

Contact us

Talk to our AI experts to find out how PwC can help you use AI responsibly and harness its potential in an ethical and conscientious way.

https://pages.pwc.ch/core-contact-page?form_id=7014L000000IIaSQAW&embed=true&lang=en

Get in touch with our AI Assurance experts

Yan Borboën

Partner, Leader Digital Assurance & Trust and Cybersecurity & Privacy, PwC Switzerland

+41 58 792 84 59

Email

Morgan Badoud

Director, Digital Assurance & Trust, Geneva, PwC Switzerland

+41 58 792 90 80

Email

Angelo Mathis

Senior Manager, Digital Assurance & Trust, PwC Switzerland

+41 79 795 01 11

Email