Barbara rejects the idea that a generic instruction to write an article can produce strong content. Her breakthrough came from supplying webinar transcripts, brand guidelines, and writing-style context so the model could transform genuine source material instead of fabricating a generic draft.

Generic topic prompts produce weak and undifferentiated writing.
Webinar transcripts supply original knowledge and concrete insights.
Brand guidelines and style references shape the resulting voice.
Strong AI content begins with meaningful source material rather than a blank page.

“And it was exponentially better because you had a genuine piece of content and insights and knowledge that you first gathered and then used to…”
Barbara Jovanovic · 07:00

“We never start with just like prompting AI just by itself to do anything, even a simplest social media post.”
Barbara Jovanovic · 07:30

#prompting#source material#ai writing#content quality

Hot Take· 2

Hot Take15:00

AI Makes Content Experiments Cheap Enough to Run Immediately

Traditional content experiments often require contractors, long commitments, and weeks of coordination before anything reaches an audience. Barbara argues that an AI-supported operation compresses the journey from idea to finished content, allowing startups to test channels without making a large staffing bet.

Conventional specialists can require costly six-month commitments.
Slow delegation can delay a simple channel test by months.
AI lowers the cost of turning an idea into finished content.
Faster production lets startups test ideas that might fail without risking a large budget.

“But this is a you know 400x cheaper uh way to do hey, I have an idea and I need to get that idea to…”
Host · 15:00

“And then when you actually delegate that, it's like two months after that you're actually testing it.”
Barbara Jovanovic · 16:00

#experimentation#startup marketing#content operations#cost reduction

Hot Take18:00

Prompting Does Not Replace Editorial Judgment

AI can identify topics and generate drafts, but Barbara says a human still needs to recognize novelty, personality, and genuine audience value. Some transcripts yield only one or two worthwhile ideas, and publishing more merely because generation is easy leads to low-quality content.

AI may miss personality-rich moments and memorable lines.
Editors must judge whether an idea deserves expansion.
Not every discussed topic is interesting enough to publish.
The amount of useful content should depend on source quality, not production capacity.

“You have to be present and you have to kind of look at that list and think critically.”
Barbara Jovanovic · 21:00

“Sometimes I'm like, okay, that's not really interesting.”
Barbara Jovanovic · 22:00

#editorial judgment#content quality#human in the loop#ai writing

Explainer· 1

Explainer13:00

The AI Stack Behind a Zero-Employee Content Agency

Barbara uses a collection of mainstream AI and media tools to cover work previously distributed across writers, designers, editors, analysts, and managers. Granola handles searchable meeting transcripts, while Riverside, Midjourney, Descript, Superwhisper, OpenAI, and Claude support production and analysis.

Granola provides meeting transcripts that users can query immediately.
Riverside and Descript support media recording and editing.
Midjourney covers visual production.
Superwhisper enables voice-first interaction with AI.
OpenAI and Claude handle content development and analytical tasks.

“Granola is very excellent for for meeting transcripts, and you get to chat with the meeting transcript afterwards.”
Barbara Jovanovic · 13:30

“And then yeah, Riverside, mid-journey, descript, super whisper, open AI, Claude, that basically replaces all of these roles.”
Barbara Jovanovic · 13:30

#ai tools#content stack#granola#superwhisper#automation

Story· 1

Story00:30

How ChatGPT Made a Solo Content Agency Viable

Barbara Jovanovic founded Startup Cookie after seeing how much time and money conventional content teams consumed. Early ChatGPT results convinced her that one person could deliver agency work without hiring a large team of writers and specialists.

Startup Cookie operates with Barbara and AI agents rather than employees.
Managing conventional content teams consumed time without feeling creative.
The first client demonstrated that the solo AI-powered model was workable.
Each improvement in AI capabilities increases the agency's leverage.

“I am a solo founder, so it's just me and all of the AI agents.”
Barbara Jovanovic · 00:30

“And when I saw what AI could do, even in the early stages, if you prompt it correctly, I was like, Wait, what if I…”
Barbara Jovanovic · 02:30

#ai agency#solopreneurship#content marketing#startup cookie

Tool· 3

Tool10:00

Use Transcript Archives as a Searchable Founder Persona

Long-term client relationships give the agency a growing archive of founder conversations. Barbara loads multiple transcripts into ChatGPT to check whether a founder has discussed a topic and to approximate the founder's likely position and phrasing when internal teams need timely messaging.

Accumulated transcripts become a reusable body of founder knowledge.
Teams can search past discussions before asking a busy founder again.
Projects in ChatGPT or Claude keep persona context available.
The archive supports more consistent phrasing and messaging.

“And I just asked Chat GP Dum, like, did he mention this? Did we talk about this topic at all?”
Barbara Jovanovic · 11:00

“It's like kind of creating a persona of someone and having that available in your chat GPT.”
Barbara Jovanovic · 11:00

#persona#knowledge base#chatgpt projects#brand voice

Tool20:00

Why Barbara Talks to ChatGPT Instead of Typing

Barbara uses Superwhisper to dictate most prompts because speaking naturally produces more background, constraints, and situational detail than typing. She finds that this additional context improves the model's results, particularly for substantial content such as blog posts.

Typing encourages shorter and more compressed instructions.
Speaking naturally adds context that users might otherwise omit.
More context helps ChatGPT understand the audience, product, and desired angle.
Voice prompting is especially useful for larger content tasks.

“I very rarely type into Chat GPT. I only use Super Whisper, so I use just my voice. I talk to it.”
Barbara Jovanovic · 20:00

“Chat GPT performs much better when you give it as much context as possible.”
Barbara Jovanovic · 20:30

#voice prompting#superwhisper#chatgpt#prompt engineering

Tool27:30

Client-Specific Projects Preserve Voice, Product Context, and Examples

Barbara creates separate ChatGPT projects for each client and deliverable. Project files include admired writing samples, previously approved work, company information, and product context, while persistent instructions specify the desired style and prohibited habits.

Separate projects prevent different client and content contexts from blending.
Favorite writers provide concrete voice and style references.
Approved deliverables become additional examples for future generations.
Persistent instructions encode direct language and prohibited writing patterns.
Barbara still repeats key instructions because project adherence is imperfect.

“We have basically a project for each client and for each of their deliverables.”
Barbara Jovanovic · 27:30

“And when we create a piece that he's like, that was great. Love that. Love that style that he established. I also just put it…”
Barbara Jovanovic · 28:30

#chatgpt projects#brand voice#client management#prompting

Takeaway· 5

Takeaway08:00

One Founder Interview Can Supply a Month of Authentic Content

Startup Cookie interviews founders who lack podcasts, webinars, or other reusable source material. The transcript captures their ideas, phrasing, tone, and product messaging, while the conversation itself helps founders explore thoughts they rarely have time to develop in operational meetings.

A monthly one-hour interview creates reusable source material.
Interview questions uncover personal views, product insights, and hot takes.
Transcripts preserve the founder's natural phrasing and tone.
Clients often value the interview as a space to think through ideas.

“But we always start with that transcript. So we have someone's actual insights and actual knowledge, and we also have like, you know, the tone…”
Barbara Jovanovic · 08:30

“The founders love it. They're like, oh my God, this was like a therapy session for us, like a monthly one.”
Barbara Jovanovic · 09:00

#founder content#interviews#thought leadership#transcripts

Takeaway16:30

No Interviewer? Record Yourself Talking Through the Week

Founders do not need a podcast invitation or an agency to create authentic source material. Barbara recommends recording a spoken review of recent lessons, product progress, team challenges, and solutions, then using the resulting transcript as the basis for content.

A solo voice recording can replace a formal interview.
Weekly events and lessons provide concrete topics.
Product and team challenges create useful narrative material.
The transcript becomes reusable input for multiple content formats.

“I often just turn on super whisper and I record myself talking. I don't need someone to interview me.”
Barbara Jovanovic · 16:30

“Right. So just talk about what happened during the week, talk about what you learned this month, how the team is doing, how's the product…”
Barbara Jovanovic · 17:00

#voice notes#founder marketing#content creation#superwhisper

Takeaway26:30

The Prompt Constraints That Make AI Writing Less Cringe

Barbara improves generated posts by specifying the desired energy and explicitly banning recurring AI habits. She favors words such as punchy and engaging, prohibits metaphors, jargon, clichés, and question-based transitions, and maintains a list of phrases the model should avoid.

Use precise style words such as punchy, engaging, and direct.
Ban metaphors, jargon, clichés, and unwanted stock phrases.
Prohibit questions used as transitions between topics.
Update the banned-word list as new AI writing patterns appear.

“And then I usually instruct it not to use metaphors, catchphrases, jargon, because it kind of makes it cringe.”
Barbara Jovanovic · 26:30

“Never use questions as transitions between topics.”
Barbara Jovanovic · 30:00

#prompt constraints#writing style#ai detection#editing

Takeaway33:00

Stop Hunting for Prompts and Ask AI to Build Yours

Barbara does not manually write many of the prompts used to operate her agency. Instead, she records herself describing a recurring task, its current process, and the desired outcome, then asks ChatGPT to identify efficiencies, automation opportunities, integrations, and an appropriate prompt.

Describe the actual recurring task rather than searching for a generic prompt.
Explain both the current workflow and the desired future state.
Ask AI to identify work that can be streamlined, outsourced, or automated.
Use tools such as Zapier when an integration can remove repetitive effort.
Custom prompts can outperform templates built for someone else's circumstances.

“I don't write them. I tell Chat GPT what I need and then it writes the prompt for me.”
Barbara Jovanovic · 33:30

“Like just tell it what you needed to do and then we'll create the prompt for you.”
Barbara Jovanovic · 35:00

#workflow automation#prompt generation#process improvement#zapier

Takeaway31:30

Use Your Existing AI Tools More Deeply Before Adding New Ones

The closing lesson is that leverage often comes from mastering the native capabilities of a small toolset rather than continually adding software. Barbara's use of projects, tasks, voice prompting, detailed instructions, and contextual files illustrates how deeper adoption can transform an operating model.

Document the workflow before automating it.
Learn native features such as projects and scheduled tasks.
Use persistent context and detailed instructions instead of isolated prompts.
Improve depth of use before expanding the tool stack.

“You don't need more tools. You need to use the tools that you're using much deeper”
Host · 35:30

“You're using way more of chat GPT than the average person.”
Host · 32:00

#ai adoption#tool mastery#operations#productivity

Frameworks in this episode

Authentic Insight Content Flywheel

Client Project Context Stack

Editorial Interest Gate

Narrate-to-Prompt Delegation Method

Positive-and-Negative Style Prompt

Test-Before-Invest Channel Rule

Transcript Insight Mining Pipeline

Voice-First Context Prompting

Weekly AI Friction Audit

Insights & moments

⚡Myth Buster· 1

AI Content Fails When It Starts From an Empty Prompt

◆Hot Take· 2