From proof of concept to intelligent automation strategy: A generative AI journey

10/03/2024

by George Hyde

You’re a forward-thinking company looking to leverage the potential of generative AI, but are grappling with how to effectively integrate this cutting-edge technology into your operations. Sound familiar?

A leading CPG client of Sendero faced the same dilemma when they embarked on a transformative journey to deliver several generative AI Proof-of-Concepts (POCs). Inspired by the rapid rise of OpenAI’s ChatGPT, which hit 100 million users in record time, the client aimed to tap into this technology while safeguarding proprietary data. They approached it methodically, assessing value, feasibility, and risks. Later that same year, three generative AI POCs were delivered. From a pool of more than 50 potential applications identified by the company’s strategy team, two use cases were prioritized for rapid prototyping, with the aim to either “fail fast and learn” or move forward to production. However, the most valuable use case happened to be the most common: deploy generative AI to support the contact center.

Implementing an Intelligent Automation Strategy

As with most contact centers, this client was seeking a solution to improve the employee experience while reducing the average handling time of cases. Leveraging generative AI capabilities like call transcriptions or personalized content generation seemed like a promising strategy to address some of those goals. After the project team gathered detailed requirements and conducted a comprehensive data architecture review, they were able to ensure the AI solution was built on a robust, reliable framework. From there, the project was ready to begin in earnest.

The Parameters

Ensuring Data Privacy: The team conducted an exhaustive data privacy review, collaborating with the client’s privacy office to anonymize sensitive data and safeguard proprietary information. This approach ensured full compliance with data protection regulations and prevented exposure to public-facing large language models (LLMs) while allowing safe use within protected environments.
Strengthening Security: A detailed security architecture review was carried out to identify vulnerabilities and implement robust safeguards. By adhering to cybersecurity standards like ISO/IEC 27001, the team fortified the AI solution against threats, preserving stakeholder trust and data integrity.

The Process

Engage Stakeholders: Regular updates, feedback sessions, and alignment with business objectives kept stakeholders informed and invested, fostering ownership and smoother implementation.
Provide Training and Change Management: Workshops and training sessions ensured employees were comfortable with the new technology, supported by comprehensive user manuals for ongoing learning.
Monitor Performance: Key performance indicators (KPIs) were defined to track the AI solution’s effectiveness and enable continuous optimization based on real-time data.
Plan for Scalability: Meticulous planning for increased demand ensured the solution could scale across the organization without performance issues, including infrastructure and resource allocation.
Incorporate User Feedback: A feedback loop was established to gather input from end-users, allowing the team to continuously improve the solution based on user needs and organizational goals.
Ensure Compliance: Compliance and legal reviews secured necessary approvals, ensuring the solution adhered to all relevant legal and regulatory requirements.

The Lessons Learned

Design Phase Integrity: While speed to market is important, an accelerated timeline during the design phase can lead to unexpected challenges during implementation. Ensuring thorough design reviews and meticulous planning can mitigate these risks and enhance project outcomes.
Architectural Oversight: Having an architect who can certify design patterns and ensure alignment with architectural strategy and best practices is crucial. This oversight helps maintain consistency, quality, and adherence to strategic objectives across the solution.
Regular Demos: Regular demos are essential to demonstrate incremental value and maintain broader organizational awareness. Without these demos, organizations risk missing opportunities due to a lack of visibility into the solutions being delivered. These demos also highlight the reusable components of many assets built, such as knowledge retrieval engines, which can be leveraged in other projects.
Human-AI Collaboration: Emphasizing the role of human-AI collaboration can enhance productivity and innovation. AI is a tool to augment human capabilities, not replace them.

The implementation of the contact center generative AI use case demonstrated the ability to navigate complex challenges and deliver impactful results. The generative AI solution reduced the Average Handling Time (AHT) for the contact center, allowing agents to handle a higher volume of work more efficiently while maintaining high-quality interactions with customers.

Data Strategy and Business Transformation

A robust data strategy is essential for successful AI implementation. Retrieval Generated Architecture patterns, for example, depend on high-quality data. Organizations that create data products can accelerate value creation across the business. However, without a robust data strategy, inefficiencies and unnecessary costs can arise.

Data privacy and security are also critical. Enterprises must ensure compliance with data privacy and security standards when using third-party tools leveraging LLMs. Even for internal applications like the contact center, it is vital to protect proprietary information from being used to train publicly available tools.

While starting with a POC is necessary to prove feasibility, journey maps should be created early to understand the full value potential of intelligent automation solutions. For instance, generative AI can reduce Average Handling Time (AHT) in contact centers, but further automation, such as voice-to-text transcription and automated note-taking, can drive additional value.

By leveraging generative AI and other technologies, organizations can operate more efficiently and effectively. However, while it’s a beneficial investment, turning POCs into mature intelligence automation solutions can be challenging. Looking to transform your organization through generative AI? Connect with one of our consultants today.