This is an archive episode, first broadcast in 2024 under the show's previous name. It carries a short synthetic announcer introduction; the conversation underneath it is untouched human audio.
One company bought twenty thousand AI licences and could not show a return on any of them. Where does the money actually come from?
Alvaro Melendez founded Krant Creativity and Technology and comes at this from the creative side rather than the engineering side, which shapes the whole conversation. His central objection is to the word "tool".
Recorded in August 2024, when the show carried a different name.
What this episode covers
- The licence problem. A licence lands on your desk with no training, no worked examples and no clarity on whether you are even permitted to use it for the thing you have in mind. Melendez treats the resulting scepticism as an education failure rather than a technology failure.
- The dupla. His Spanish term for a thinking partner - the second brain you work a problem with. He argues that treating AI as a tool rather than a partner is what caps the value people get from it.
- Einstein serving coffee. His image for what most deployments are actually doing with the capability they have bought, and the one that tends to land hardest with executives.
- What paid, and what did not. A ten per cent saving on email drafting is the sort of result that impresses nobody and justifies no budget. Books, transcripts and data mining are where he saw returns worth reporting, and he is specific about why those differ in kind rather than degree.
- Breaking a job into tasks. He walks through doing this for a creative director, which is the practical version of the reframe.
- The PhD intern in your pocket - and its blind spots. Context windows, hallucination, and the lawyer who filed fake citations.
- Your AI council must be internal. Why an outsourced one fails, and why ethics and legal belong in the room early rather than at sign-off.
- What he will not hand over. He is specific about where he keeps a human in the loop and why, which is more useful than another list of things AI can do.
- Start with the willing ten. His adoption advice: find the ten people who actually want to try, support them properly, and let the rest follow rather than mandating it.
- Why perception lags capability. GPT-3.5 and a wave of poor email assistants soured a lot of people, and he argues many of them have not updated since.
- Implode your own company before someone else does. The exercise he puts to leadership teams, which is less comfortable than the usual roadmap workshop.
About the guest
Alvaro Melendez is the founder of Krant Creativity and Technology, and of Ink Lantern. He comes from the creative industries rather than from engineering, which is why the conversation keeps returning to how creative teams actually work - two people thinking together, rather than one person operating a tool.
Find the product at InkLantern.io, where there is a demo, and reach him directly on LinkedIn, which is where he says he is most active.
Resources mentioned
- Ink Lantern - Melendez's product
- The Titanic dataset exercise - his way of getting senior leaders hands-on in an afternoon
- Google's Gemini Olympics advert - his example of exactly what not to do
- Ethan Mollick - cited for the "AI is not a tool" argument
Quotes
"If you're thinking about AI as a tool, you're just massively undercutting this thing, because AI is not a tool."
Alvaro Melendez
"You just hired twenty thousand employees."
Alvaro Melendez, on the reframe that changes behaviour
Books - the Human-First AI Series
- Hello! I'm New Here
- An Absolute Beginner's Guide to AI
- Use AI, Stay Human
- The AI Pilot Handbook
- When AI Becomes Your Responsibility
Connect with Alastair on LinkedIn.