EC-301a · Module 2

Translating AI to Board Language

4 min read

Boards speak three languages: risk, governance, and fiduciary duty. They do not speak tokens, fine-tuning, or inference speed. If you use those terms in a board presentation, you have shifted the meeting from a governance discussion to a technology tutorial — and you have lost the room. Every AI concept must be translated before it appears in a board deck.

The translation is not dumbing down. It is precision. A board member asking about large language models is actually asking: whose data trains this? What does it generate? Who is liable when it generates something wrong? Those are the questions your slide needs to answer — not how transformers work.

The translation table is simple. "We are deploying a large language model" becomes "We are implementing an AI system that drafts customer responses and routes support tickets, reducing handling time by 40%." "We are fine-tuning on proprietary data" becomes "We are customizing the AI on our historical records, which means we need data governance controls to protect that information." Every technical statement has a business-risk translation. Use the translation, not the technical statement.

Do This

  • "The AI processes customer support tickets and routes them to the correct team, reducing average handle time from 8 minutes to 5 minutes"
  • "Our data governance policy requires that no customer PII is included in the training dataset — this is enforced through automated filtering before any data is processed"
  • "The vendor's model operates under contract terms that we have reviewed with outside counsel — IP ownership and liability are clearly assigned"
  • "AI output is reviewed by a human agent before any customer-facing communication is sent — there is no fully automated customer interaction"

Avoid This

  • "We are deploying GPT-based infrastructure with RAG augmentation on our knowledge base"
  • "We are fine-tuning on proprietary datasets to improve model performance on domain-specific queries"
  • "The model operates under standard API terms with token-based pricing"
  • "The system has guardrails to prevent hallucination and misuse"