AI Landscape Quiz: Do You Know the Models That Matter?

A small law firm uses an AI document tool that stores all case materials. They are considering switching providers. What vendor-related risk should they have addressed during onboarding?

Whether the AI was trained on legal documents Whether case documents can be exported in standard formats — allowing migration without data loss if they switch providers Whether the existing provider offers referral discounts Whether the new provider offers a lower price

Which of these vendor claims is a RED FLAG suggesting a tool may not have meaningful AI capability?

The tool integrates with Zapier 'AI-powered' prominently featured in marketing but no specific explanation of what the AI does in the product The tool requires a free trial signup before showing pricing The vendor has raised venture capital funding

A small business evaluates an AI tool and discovers it will fine-tune its model using the company's customer data to improve future recommendations. What is the most important consideration before agreeing?

Whether the model will be available to the business for download Whether the model improvement will be visible to the business within 30 days Whether fine-tuning on small business data produces measurably better results Whether using customer data for model training complies with the business's privacy policy and applicable data protection regulations — and whether customers consented

What does SOC 2 certification indicate about an AI tool vendor?

The vendor stores data in the US and is subject to US law The vendor is approved by the US government for handling classified data The vendor has undergone an independent security audit verifying their security controls, data availability, processing integrity, and confidentiality The vendor's AI models are certified as bias-free

Which Gemini feature most directly reduces hallucination on questions about recent events?

Gemini's large context window Gemini's fine-tuning on Google's proprietary dataset Google Search grounding — responses can be anchored to live Search results Gemini Flash's lower latency

A healthcare company is evaluating an AI vendor's product for processing patient records. The vendor's product is built on a closed-source foundation model API. What is the primary data privacy consideration?

The application layer encrypts data before it leaves the company The foundation model will permanently memorize patient data Patient data will be transmitted to the foundation model provider's servers Closed-source models automatically comply with HIPAA

What is the biggest risk when an AI produces a confident-sounding analysis of your business?

AI may accidentally leak your business data to competitors AI analysis is too long — it always produces 10+ pages The AI may refuse to complete an analysis if the business is in a regulated industry AI doesn't know your actual organizational context — it analyzes the concept you described, not the specific realities of your situation

A company wants to connect an AI assistant to its private internal knowledge base so it can answer employee questions. Which technique is most relevant?

Switching to an open-source model Building a new foundation model Fine-tuning the foundation model from scratch Retrieval-Augmented Generation (RAG)

What is the primary benefit of marking variables in a prompt template using brackets like [CLIENT_NAME]?

The brackets are required syntax for AI models to recognize variable values It allows the template to work across different AI provider APIs Brackets encrypt the variable so the AI doesn't see the actual client name It makes the changeable elements visually obvious, so any team member can fill them in without needing to understand prompt engineering

A strategy team wants to use AI to stress-test a business plan. Which prompt is most effective for this goal?

'Act as a skeptical board member. What are the three riskiest assumptions in this plan, and what evidence would change your view?' 'Tell me if this business plan will succeed' 'What improvements can you make to this plan?' 'Summarize this business plan in three paragraphs'

A sales team writes a different prompt every time they need to generate a follow-up email after a meeting. What is the highest-leverage improvement?

Create a reusable prompt template with clearly marked variables for meeting date, client name, key discussion points, and next steps Only the most senior team members should use AI — it causes quality variance Switch to a more powerful AI model Each salesperson should develop their own unique prompting style

Which of these prompt template practices is MOST important for a large team to maintain quality over time?

Lock down templates so only administrators can use them Rotate to a new AI model every quarter to keep outputs fresh Document expected outputs alongside each template so users know what 'good' looks like and can spot when quality has degraded Require every user to rewrite 30% of the template before use

A team of 50 professionals already uses Gmail, Google Docs, and Google Sheets daily. Which AI path offers the lowest adoption friction?

Google Gemini for Workspace — AI is embedded directly into their existing tools Claude API — requires integration by a developer ChatGPT Enterprise — requires a new platform adoption Self-hosted Llama — requires technical setup

A small accounting firm uses the free version of an AI chatbot to help draft client financial summaries. The summaries include client names and income figures. What is the primary data risk?

No risk — small accounting firms are exempt from data confidentiality regulations The AI chatbot will share client financial data with the IRS Free AI tools automatically encrypt all submitted data Free-tier AI tools may use inputs for model training, potentially exposing client financial data — this likely violates professional confidentiality obligations

Why can many different AI products (e.g., a writing tool, a customer service bot, a code assistant) feel similar in capability?

They all copy ChatGPT's training data AI models can only do one type of task well They may all be using the same underlying foundation model via API Foundation models are open-source and freely copyable

A small physical therapy clinic wants to use an AI scheduling chatbot on their website. Patients will input their names, contact info, and appointment reasons (which may reference medical conditions). What is required before deploying the chatbot?

Chatbots don't count as HIPAA-covered technology — only EHR systems do A Business Associate Agreement (BAA) with the chatbot provider, since the chatbot may process Protected Health Information under HIPAA HIPAA only applies to clinics with more than 20 employees A standard terms of service acceptance by the clinic is sufficient

A small business owner is considering three AI marketing tools, all with similar feature sets. Which evaluation question is MOST important for long-term cost management?

Which tool has the best-designed user interface? What is the pricing at the volume I'll be using? (Many tools have low entry prices that become expensive at scale) Which tool was most recently founded? Which tool has the most integrations in their marketplace?

A manager asks an AI to recommend which of two job candidates to hire. What is the correct role for AI in this decision?

AI can make the final hiring decision based on the resume data provided AI can make the recommendation, but a human must type it into the HR system AI can help structure the evaluation criteria and surface considerations, but the hiring decision should be made by the human hiring manager AI is too risky to use in any hiring process

Which layer of the AI stack contains products like ChatGPT, Microsoft Copilot, and Perplexity?

The API and platform layer The foundation model layer The application layer The training data layer

A healthcare startup wants to use AI to analyze sensitive patient records. Sending data to external cloud APIs is not permitted under their compliance policy. What type of deployment is most appropriate?

Self-hosted open-source model (e.g., Meta Llama) on the organization's own infrastructure Google Gemini Pro — all Gemini data stays in healthcare-dedicated servers Whichever model has the highest benchmark score ChatGPT Enterprise — data privacy is guaranteed by contract