Ep 107 – 🧣Wrapping your head around Responsible AI🧣 (Ferry Grijpink)

Responsible AI refers to deploying artificial intelligence in ways that are ethical, safe, transparent, and compliant with regulations. For telcos, this is particularly critical because they operate in a regulated industry handling sensitive customer data. According to McKinsey’s report Responsible AI: A Business Imperative for Telcos, AI could create $600 billion in value for the telco industry by 2040, with about half requiring the highest AI standards for compliance. Getting it wrong can lead to biased algorithms, data leaks, regulatory fines, and reputational damage.

Ferry Grijpink from McKinsey’s Center for Advanced Connectivity advises telcos to start by setting a clear vision that matches their business goals with their risk tolerance. For example, operators pursuing an “AI-first” strategy may accept more risk for speed, while those in highly regulated markets like Europe need stricter controls. Organizations should establish policies, create AI awareness through training, provide approved AI tools to prevent “Shadow AI,” and track whether use cases actually create value—stopping experiments that don’t deliver returns.

McKinsey & Company recommends five essential elements: First, set a clear vision for what you want to achieve with AI. Second, build an operating model with named individuals responsible for each AI use case. Third, implement technical controls to ensure clean data and test models for risk. Fourth, ensure third-party vendors and products comply with your AI standards. Fifth, manage change through communication and training to help employees understand proper AI usage while avoiding innovation-killing restrictions.

DR emphasizes that defining responsible AI is a C-suite decision, not something to leave to IT or legal teams alone. Your business goals should dictate your AI risk tolerance. If you’re aiming to become AI-first like SK Telecom, you might trade some risk for speed. However, if you’re managing EU compliance, you’ll need lower risk tolerance due to regulatory fines. The key is matching guardrails and rules with business objectives to enable innovation rather than stopping it, which is exactly what Totogi focuses on with telco clients.

Shadow AI occurs when employees use unauthorized AI tools outside company controls—similar to the old “Shadow IT” problem but potentially more dangerous. Because many AI tools have freemium models or cost just $20/month, employees can easily bypass policies. McKinsey & Company advises against simply blocking access, as workers will use personal devices instead. Better approaches include setting clear policies, providing approved AI options, using monitoring tools to identify usage hotspots, and training employees on appropriate use while explaining the risks of unauthorized tools.

Organizations should implement monitoring tools to see where and how employees are using AI, creating visibility into usage patterns and costs. For example, DR’s company Totogi uses API call stubs in AI applications to track usage and consolidate spending, enabling better vendor negotiations. Companies should also look for usage hotspots—if customer service teams heavily use unauthorized LLMs, that signals a need for approved tools in that area. Tracking helps optimize costs, as specialized API contracts are typically cheaper than generic LLM tools, while also identifying which use cases deliver actual business value. Ferry Grijpink has discussed AI value creation in previous Telco in 20 episodes, including Episode 83: Monetizing Network APIs and Episode 45: Drumming up ways to increase ARPU with McKinsey & Company.