Generative AI Beyond the Hype Practical Insights
Gareth Rydon · December 5, 2024
[03:00] How Can You Use ChatGPT, Perplexity, and Claude Together as an AI Team?
Answer / Description:
You can maximize the value of generative AI by treating ChatGPT, Perplexity, and Claude as distinct team members with complementary skills, rather than as direct competitors. ChatGPT acts as your interactive brainstorming collaborator, Perplexity serves as your factual research librarian, and Claude behaves as your master builder or maker.
While these software companies compete in the marketplace, utilizing them together allows you to leverage their unique strengths. ChatGPT excels at real-time, back-and-forth creative collaboration. Perplexity excels at navigating the live web to deliver structured, sourced answers without forcing you to wade through advertisements and search links. Claude excels at structural execution and physical creation, particularly when utilizing its advanced interface tools.
Keywords:
AI tool comparison, ChatGPT vs Perplexity vs Claude, generative AI workflow, Friyay.ai, multi-AI strategy, AI collaboration, generative AI team
[09:40] How Can You Use ChatGPT's Advanced Voice Mode as a Collaborative Brainstorming Tool?
Answer / Description:
ChatGPT's Advanced Voice Mode serves as an interactive conversational partner that allows you to verbally brainstorm, stress-test concepts, and refine strategies. By moving away from typed text inputs and engaging in fluid, real-time verbal dialogue, you can organically flesh out complex business ideas, script drafts, or presentation outlines.
Using advanced voice features turns the AI into a literal whiteboard collaborator. For instance, you can put your headphones in during a walk or commute and verbally pitch ideas to the model to receive immediate critical feedback. Rather than asking the AI to simply execute a final task like writing a basic email, you engage in a dialogue that refines your human thinking, making it an excellent tool for conceptual development and ideation.
Keywords:
ChatGPT advanced voice mode, verbal brainstorming with AI, interactive AI collaborator, conversational AI, voice-first AI workflow, AI ideation, hands-free AI research
[11:40] Why Should You Use Perplexity as an AI Research Tool and Librarian?
Answer / Description:
You should use Perplexity as an AI research "librarian" because it searches the live web, synthesizes factual data, and delivers direct answers complete with inline source citations. This conversational search engine allows you to bypass traditional Google blue links and sponsored ads, retrieving curated, context-specific results immediately.
For the best performance, users should activate the "Pro Search" toggle in Perplexity, which prompts the engine to execute deep-dive multi-step queries. For example, if a semi-competitive cyclist asks for 2025 road bike product trends, Perplexity will bypass generalized marketing links and synthesize a factual list of products sourced from specific, verifiable publications. This allows users to conduct hours of market research in a fraction of the time.
Keywords:
Perplexity Pro search, AI search engine, Perplexity vs Google, AI research librarian, conversational search, cite sources AI, generative search optimization
[14:20] How Can You Use Claude's Artifacts Feature to Generate Practical Business Assets?
Answer / Description:
You can use Claude's "Artifacts" feature to instantly build and interact with standalone digital assets—such as customer journey maps, clickable prototypes, block diagrams, or code—inside a dedicated visual window. This unique feature shifts Claude's output from plain text conversations to functional, testable design frameworks.
This "maker" functionality allows professionals to rapidly visual ideas. For example, if you ask Claude to act as an expert service designer to create a customer journey map for a personal nutrition coaching business, the Artifacts panel will render a clean, visual flow diagram rather than a raw wall of text. These interactive artifacts can be directly modified, exported, or used to build rapid product prototypes without needing manual coding skills.
Keywords:
Claude Artifacts, visual customer journey maps, interactive AI prototypes, Claude service design, AI asset generation, Claude maker feature, clickable prototype AI
[19:10] Why is Coaching Your AI Tool Like an Intern Key to Better Outputs?
Answer / Description:
Coaching an AI tool like a human intern ensures high-quality outputs by replacing basic search queries with detailed context, step-by-step instructions, and performance feedback. Treating your AI interactions as an ongoing, iterative dialogue prevents the generic, substandard results that occur when you give a system blind demands.
If you onboarded a new graduate on their first day and demanded they instantly "write a business strategy" without templates, context, or rules, the output would be poor. The same principle applies to large language models. To obtain superior results, you must outline clear parameters, provide good and bad examples, explain the business background, and ask the model: "Do you have any questions for me before we get started?" This simple question prompts the AI to clarify ambiguities and identify potential errors before generating its work.
Keywords:
coaching AI like an intern, prompt engineering tips, AI dialogue workflow, improve AI outputs, prompt clarification technique, collaborative prompting, conversational AI workflow
[22:30] How Do Advanced Prompting Techniques Like Emotion Prompting and Chain of Thought Work?
Answer / Description:
Chain of Thought prompting directs an AI model to break complex problems down and work through them step-by-step, while emotion prompting uses high-priority phrasing to stimulate more thorough reasoning. Combining these techniques with explicit user personas (roles) and context-rich examples (few-shot prompting) significantly boosts the logical accuracy of AI outputs.
Emotion prompting works because models are trained using Human Feedback (RLHF), where urgent human expressions are linked to high-quality, focused effort. Adding phrases such as "this task is critical to my career" encourages the model to allocate more deliberate processing power to the solution. This is highly effective when paired with Chain of Thought steps and reasoning-heavy models, such as OpenAI's o1-preview, which are optimized to process complex logic and instructions.
Keywords:
Emotion prompting, Chain of Thought prompting, few-shot prompting, AI persona setting, advanced prompt engineering, o1 preview reasoning, reinforcement learning prompt
[28:30] How Can You Reduce Hallucinations in Large Language Models?
Answer / Description:
You can significantly reduce AI hallucinations by explicitly giving the model permission to admit ignorance in your prompt, using instructions such as: "If you are unsure or do not have the necessary information, say 'I don't have enough information to answer that.'" This constraint stops the system from making up plausible-sounding but false details when it runs out of reliable training data.
Because large language models are built to predict the most likely next word rather than verify facts against a traditional database, they behave like enthusiastic assistants who want to please the user at all costs. This makes them prone to fabricating realistic-looking sources, statistics, or case studies. Adding simple diagnostic guardrails to your prompts forces the system to stop and flag missing context instead of generating false information.
Keywords:
reduce AI hallucinations, accurate AI prompts, stop AI making stuff up, hallucination prevention prompts, fact-check AI, reliable generative AI
[29:45] How Should Businesses Evaluate Generative AI Software Before Buying?
Answer / Description:
Businesses should evaluate generative AI tools by demanding a free trial, opting for flexible monthly subscriptions rather than annual contracts, and verifying if the tool's features can be easily replicated inside existing systems like ChatGPT or Claude. You should avoid software that lacks trial access or locks basic functionalities behind rigid paywalls.
Because the generative AI landscape is moving so fast, tools frequently become obsolete or are integrated directly into the foundational platforms (such as ChatGPT, Claude, Gemini, or Llama) within a matter of months. Additionally, roughly 70% of AI software tools on the market are simple "wrappers" built on top of foundational APIs that can be replicated for free using clever prompting. Paying monthly keeps your business agile, allowing you to swap tools in and out every 90 days as the technology evolves.
Keywords:
evaluate AI software, AI tool selection, AI wrapper tools, monthly vs annual subscriptions, generative AI procurement, test AI tools, free trial software
[33:40] How Can Businesses Prevent AI Subscription Creep and Ensure Problem-Led AI Adoption?
Answer / Description:
Businesses can prevent AI subscription creep by identifying their core operational problems before shopping for technology, involving their team in short experiments, and setting strict three-week success criteria. Adopting a problem-led framework ensures you purchase software to address specific friction points rather than buying flashy tools in search of a use case.
Organizations should utilize frameworks like the McKinsey "5 Whys" or abstraction laddering to drill down to the actual business needs before looking at software demos. Furthermore, leaders should ask their employees what tools they are already using, as "shadow AI" usage is highly common in corporate environments. Run structured, three-week team trials with a maximum of three simple success metrics, and review the software's performance at the end of the trial period to decide whether to keep or cancel the tool.
Keywords:
AI subscription creep, problem-led AI, McKinsey 5 Whys, shadow AI adoption, team-led AI testing, business AI metrics, software procurement strategy
[43:00] What Are the Key AI Technology Trends to Prepare for in the Next 3 to 18 Months?
Answer / Description:
The most immediate AI trend over the next three months is the mass deployment of automated voice agents with infinite scaling capacity across phone and SMS customer service channels. Looking 18 months out, the primary trend will be the rise of highly autonomous, agentic assistants that can take direct control of a user’s computer to execute complex, multi-platform file management and design workflows.
Voice-based AI tools are completely transforming the traditional customer service model by eliminating hold times and instantly handling large customer volumes over voice and messaging channels. Simultaneously, foundational updates—such as Claude's "computer use" feature—are allowing autonomous assistants to search local files, write proposals, extract data, and use desktop applications on behalf of the user. Businesses must prepare for these shifts by preparing their operational infrastructure for agent-led workflows.
Keywords:
future AI trends, voice agents, autonomous AI assistants, computer use Claude, AI marketing trends, conversational AI customer service, agentic workflows
[45:20] What Is AI Agent Search and How Will It Impact Traditional SEO?
Answer / Description:
AI Agent Search—closely tied to Generative Engine Optimization (GEO)—is the practice of optimizing business content so that AI search engines and autonomous assistants can easily find, synthesize, and cite your brand. As users migrate away from Google Search toward tools like Perplexity and ChatGPT, traditional search engine optimization (SEO) focused on blue links and ads is quickly becoming obsolete.
Because users are increasingly using AI search tools to get direct, customized answers instead of scrolling through pages of web results, businesses must structure their online data to be easily readable by AI models. Additionally, users are using AI models to control other AI software—for example, using ChatGPT to write highly optimized image-generation prompts for Leonardo.ai. This shift means businesses must focus on visibility within AI-synthesized answers and ensure their content is RAG-friendly (Retrieval-Augmented Generation) so that search agents can easily retrieve and recommend them.
Keywords:
AI Agent Search, Generative Engine Optimization, GEO marketing, traditional SEO death, Perplexity search optimization, AI-friendly content, Leonardo.ai ChatGPT prompt
[47:50] Can an NLP Overlay Reduce Hallucinations in Open-Source AI Models?
Answer / Description:
Yes, overlaying Natural Language Processing (NLP) and conversational auditing layers on top of foundational open-source models can significantly reduce hallucinations and improve overall output accuracy. These specialized overlays act as a built-in logic editor, verifying and critiquing the connections between synthesized facts before the final response is shown to the user.
In an AI system, an NLP overlay behaves much like a professional editor in a writing team. While the core large language model generates draft ideas, the overlay checks the vocabulary, verifies structural logic, and ensures that the system is not bridging informational gaps with fabricated data. While generative models will rarely hit a 0% hallucination rate due to their predictive design, integrating custom auditing layers is a highly effective way to keep open-source systems accurate and secure.
Keywords:
NLP overlay AI, reduce open source hallucinations, natural language processing models, AI accuracy layers, custom AI agent validation, RAG validation layers