Paris Talks Marketing

There are a handful of practical tips - the 'config mode' prompting trick, voice dictation as primary input method, Make.com assembly-line workflows for content, and the 1:1 outreach angle on the content flywheel - but the episode is heavily padded with existential AI job-displacement musings, a tangent about Patreon creators and food production costs, and repetitive agreement between host and guest. Actionable insight per minute is low.

“I put it in what I call config mode. So I just prompted to reply, you know, Reply with like 5 options where I can just reply with A, B or C to pick from each option”

“once you have a new piece of content, you need to go out to the market and present it to them. And I don't mean just posting it on LinkedIn. Think about the five people in front of which you want to put this piece of content”

The 'content conversation flywheel' framing - using content as a prompt for deliberate 1:1 outreach rather than broadcast distribution - is a modestly fresh angle on content strategy, and the rebranding of 'prompt engineering' to 'prompt tinkering' is a small but useful reframe. Beyond that, the takes on AI replacing jobs, marketers becoming orchestrators, and assembly-line content production are widely circulated in 2024-era marketing discourse.

“The thing that is ironic here is that this works because it's not scalable. Actually uh, one to one outreach for a piece of content is not at all scalable”

“prompt engineering is a pretty bad brand for what you're really doing there... I would think it would be more prompt tinkering, prompt experimenting”

Jakub is a working practitioner - head of marketing at a small healthcare AI startup, previously marketing manager at a niche agency tooling company - which gives him legitimate operator credibility, but his experience is at modest scale with no notable results cited, and he is not a recognised name in B2B marketing. The conversation confirms he is thoughtful but not unusually senior or accomplished.

“I was at an agency called Zenpilot helping marketing agencies to get success with ClickUp implementations. And now another pivot. I'm at talkie AI”

“I'm a marketer for over a decade now. My primary background is from writing public speaking. I studied English. Then I got my first marketing job”

The episode is almost entirely devoid of concrete data - no campaign metrics, no revenue figures, no conversion rates, no timelines with outcomes. Tool names (Make.com, Super Whisper, ChatGPT) and personal names (Chris Dubois, Kurt Schmidt) are dropped, but the most detailed workflow description is a secondhand, vague relay of someone else's setup with no measurable results attached.

“he showed me a, uh, make.com workflow... He uses a tool called Super Whisper... he has a whole flow where this input is taken into a variety of nodes within make. So the first node would, for example, outline the content”

“I gave it a couple images for inspiration, and then I replied with something like config mode”

The host occasionally follows up usefully (clarifying 'config mode,' probing the deep research memory claim), but frequently pivots to lengthy self-directed monologues - including a multi-minute three-phase framework speech - rather than drawing more out of the guest. There is no meaningful pushback or productive disagreement anywhere in the episode; host and guest agree on virtually every point.

“Can I ask you a little bit more about that? So you're asking ChatGPT to just summarize its answers into three simple choices, or A, B, C. Or am I getting that right?”

“I'm thinking about introducing a framework. And, uh, I'll bounce this off of you where there's three phases of it. And the end goal is that I want people to think that ultimately they will all become orchestrators”