Jonathan Goodman appears as the featured guest and frames his work through Halyard Consulting as a practical AI transformation for businesses, especially small and mid-sized firms. He explains that Halyard’s approach is not “replace everyone with AI,” but to identify operational pain points and repetitive tasks that drain time, then build targeted AI-enabled bots and workflows that reduce friction and free staff to focus on higher-value work. He uses a concrete example from accounting operations: instead of manual receipt entry and review chains, a business can capture receipts quickly (even via photos), have AI extract the relevant fields, populate structured outputs like spreadsheets, and integrate that into accounting tools, saving hours and improving operational focus.

When the host raises concerns about AI hallucinations and accuracy, Jonathan emphasizes responsible implementation. He says the way they reduce hallucinations is by anchoring bots to a client-specific knowledge base—training the system on the organization’s real documents, history, policies, and constraints, so the AI operates within a narrower, more relevant context rather than guessing broadly. He’s clear that review and validation are still required, but the productivity gain comes from shifting humans from “create everything from scratch” to “verify and refine,” dramatically compressing turnaround time while preserving quality. He also notes that AI can struggle with math in certain contexts, which is why checks and balances, and sometimes tool selection, matter.

Jonathan then advises on building AI systems for customer service and platform support. He explains that if a business has a large product set (e.g., thousands of SKUs in a database), structured data can feed a knowledge base, enabling a chatbot to answer questions, recommend products, and respond consistently. He extends this to routing and escalation: the bot can be designed with conditional logic (“if-then” handling) to recognize support scenarios and direct customers to the right channel—such as providing a partner’s support contact (e.g., Amazon) when fulfillment or returns are handled externally. He also supports a hybrid rollout model where human support teams and AI run in tandem, using real incoming questions to continuously improve the knowledge base and the bot’s performance over time.

On AI-generated audio, Jonathan recommends using ElevenLabs to produce audiobook-style narration. He explains that voice models can learn a speaker’s cadence and tone from a short voice sample, after which chapters can be converted into audio in that voice at relatively manageable cost. Beyond one-off production, he proposes a longer-term strategy: build a dedicated, persistent knowledge-based assistant trained on an author’s prior books and content, so future writing aligns with the author’s voice, themes, and structure. In his view, this “owned bot” becomes a durable asset—more controllable and reliable than relying on third-party GPTs that can change unexpectedly.

Finally, Jonathan shares his perspective on authorship and AI. He positions AI as an assistant—not a substitute for the creator’s responsibility—arguing that meaningful work still requires human intent, oversight, and editing. Even if AI drafts a large share, the author must validate accuracy, address hallucinations, confirm sources when needed, and ensure the ideas are genuinely their own. He closes by stressing that the technology is improving rapidly and that today’s limitations represent the weakest versions businesses will ever use—making now an advantageous time to start experimenting, building workflows, and removing operational pain while maintaining human control.