Most AI failures aren't technology failures - they're scoping failures. In this episode from the archive (November 2024), Alastair sits down with Michael Zipursky, CEO and co-founder of Consulting Success, who has advised organisations from startups to billion-dollar corporations across more than 75 industries and is the author of *The Elite Consulting Mind* and *Consulting Success*. The conversation centres on a pattern Zipursky sees repeatedly among consulting firm owners and business leaders: the instinct to build one large, ambitious AI solution spanning five parts of the business at once, rather than solving a single narrow use case first. He is candid that his own firm made these mistakes and learned from them - and argues that unless you're a well-funded technology company with genuine project management chops, the smarter path is to pick one small use case, build it, test it, learn, and only then expand.
The discussion splits AI adoption into two distinct categories: custom solutions you build yourself, and the fast-growing pile of pre-existing tools you can simply buy. Zipursky recounts an unsolicited pitch from an AI SDR service that makes phone calls while pretending to be human, and uses it to raise a question many buyers skip entirely: is this AI representing your brand the way you'd represent it? He and Alastair explore how volume-driven automation can quietly degrade response rates and brand equity, why so much automated outreach may collapse under its own weight, and the strange near-future where AI talks to AI - already visible in recruitment, where applicant-side tools now fire off hundreds of personalised applications into employer-side screening systems. The pair also look further ahead at agents, AI-optimised websites and API-first interfaces, and why science-fiction ideas from *The Jetsons* to *Star Trek* have a habit of becoming product roadmaps.
On the practical side, Zipursky makes a blunt case that business owners who aren't at least actively thinking about AI's implications are being reckless - not because every industry will be transformed tomorrow, but because the odds of being the next Blockbimport, Kodak or BlackBerry are non-trivial. His guidance is deliberately unglamorous: don't try to keep up with every new tool (nobody can), don't treat it as a sprint, and instead build a culture of continuous learning where every role - marketing, research, operations, leadership - is asking how AI could make their specific work better. He also offers a neat trick for the common "I don't know what my use cases are" problem: ask the AI itself, describe your role and recurring problems, request ten options, then go deeper on one. Throughout, both agree that hallucination, context, situational awareness and expert judgement mean the human expert isn't going anywhere yet - but the expert who refuses to explore might be.
**Resources mentioned:**
- *An Absolute Beginner's Guide to Using AI* - https://humanspark.ai/books/beginners-guide-to-ai/
- consultingsuccess.com and the Consulting Success podcast
- Connect with Michael Zipursky on LinkedIn
---
The introduction to this episode is read by an AI-generated voice. The conversation itself is unedited archive audio.
