AI doesn’t have a trust problem. It has a listening problem.

AI tools are becoming more capable, yet many businesses still can’t use them to do the work that actually matters. Why? Because the people building AI often don’t understand how messy real-world jobs, industries and workflows are.

In this episode, Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Galina Fendikevich, founder of Artificial Reality Media, about the gap between the AI story coming out of Silicon Valley and the experiences of the people expected to use it.

They discuss:

→ Why many AI products only get businesses 80% of the way there
→ What builders miss when they don’t speak to customers
→ Why domain specific tools matter
→ How incorrect AI advice cost one business owner $30,000
→ Why authentic customer stories will beat mass-produced AI marketing
→ What AI adoption might really look like over the next two years

If you’re building, buying or implementing AI, this is a conversation about listening before scaling, and designing technology around people rather than expecting people to work around technology.

Find Galina and Artificial Reality Media:
https://artificialrealitymedia.com/

Subscribe to PRception on YouTube and join the mailing list at:
https://prceptionmedia.com

Yariv Barsheshat, an independent AI safety researcher, discusses the potential dangers of AI training data and the unintended consequences of fine-tuning models. The conversation delves into the Hugging Face incident, the potential implications of AI, and the impact of harmless data on AI models. The conversation delves into the process of fine-tuning language models, its impact on model understanding, unintended consequences, lack of guardrails, and the decision to sunset fine-tuning platforms. It highlights the potential biases and behaviors that can result from fine-tuning, raising concerns about the ethical implications of this practice.

Takeaways

• AI Safety Research
• Fine-Tuning Models
• Unintended Consequences of AI Training Data Fine-tuning of language models involves adjusting the model's weights to integrate new data and update its responses.
• Fine-tuning language models can lead to unintended biases and behaviors, impacting how the models understand and respond to information.

Chapters

• 00:00 The Impact of Harmless Data on AI Models
• 31:14 Fine-Tuning Language Models
• 34:28 Impact of Fine-Tuning on Model Understanding
• 36:58 Company's Intentions in Fine-Tuning Models
• 40:18 Unintended Consequences of Fine-Tuning
• 43:34 Lack of Guardrails in Fine-Tuned Models
• 50:46 Sunsetting Fine-Tuning Platforms

Never miss a beat with BONUS episode content! Sign up for PRception briefing - https://prceptionmedia.com/subscribe

AI predicts patterns. Big Tech sells that capability using the language of intelligence, rationality and reasoning.

Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Stephanie Antonian, founder of Aestora, whose career has included helping to establish DeepMind’s ethics team and working at Google X.

Stephanie challenges the idea that today’s AI systems think like humans. She explains why describing prediction as reasoning is a powerful marketing decision, and why the language used around AI shapes what society is willing to trust machines to do.

They explore the difference between what is probable and what is possible, the danger of allowing machines to make judgements about people, and why the current “AI revolution” may need to become a human renaissance instead.

This is a provocative conversation about intelligence, trust, power and what technology cannot decide for us.

PRception is a podcast about how authority actually forms.

Before anyone makes a decision, they decide what feels credible, what they trust and what they believe. Now AI is changing how that happens, before anyone speaks to you.

Hosted by Sarah Kemp, narrative architect and founder of SJK Labs.

Subscribe for conversations about credibility, authority, AI discoverability and the evidence that shapes who gets trusted, shortlisted and chosen.

CONNECT: [email protected] / www.linkedin.com/in/sarahjkemp

Never miss a beat with BONUS episode content! Sign up for PRception briefing - https://prceptionmedia.com/subscribe

A neglected review profile is not merely a customer service problem. To an AI system researching your company, it can become evidence.

Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Katya Rozenoer, co-founder and CEO of Blastra, about the overlooked sources shaping AI’s opinion of software companies.

They explore why directories, reviews, Reddit discussions and other independent signals matter; how unanswered complaints can become a reputational time bomb; and why outdated product profiles undermine the story a company tells about itself.

Katya also explains why influencing AI requires more than publishing polished content. If your external evidence contradicts your brand narrative, AI can find the contradiction.

This episode is for B2B marketers, software founders and reputation leaders trying to understand how their company is being judged outside its own website.

PRception is a podcast about how authority actually forms.

Before anyone makes a decision, they decide what feels credible, what they trust and what they believe. Now AI is changing how that happens, before anyone speaks to you.

Hosted by Sarah Kemp, narrative architect and founder of SJK Labs.

Subscribe for conversations about credibility, authority, AI discoverability and the evidence that shapes who gets trusted, shortlisted and chosen.

CONNECT: [email protected] / www.linkedin.com/in/sarahjkemp

Never miss a beat with BONUS episode content! Sign up for PRception briefing - https://prceptionmedia.com/subscribe

A polished pitch can win attention. It cannot survive a second, third and fourth question if the underlying story does not hold together.

Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Tingting Peng, who has sat on both sides of the table as an operator, fundraiser and investor.

They examine the behaviours that build, or destroy, credibility during a pitch: redirecting questions instead of answering them, performing confidence, hiding what has not worked and making claims without sufficient evidence.

Tingting explains why intellectual honesty is more persuasive than pretending to have every answer, how customer stories allow other people to sell on your behalf, and why credibility is ultimately operational. It lives in what you consistently do, including when nobody important appears to be watching.

This episode is for founders, executives and advisers who need people to believe the business behind the presentation.

PRception is a podcast about how authority actually forms.

Before anyone makes a decision, they decide what feels credible, what they trust and what they believe. Now AI is changing how that happens, before anyone speaks to you.

Hosted by Sarah Kemp, narrative architect and founder of SJK Labs.

Subscribe for conversations about credibility, authority, AI discoverability and the evidence that shapes who gets trusted, shortlisted and chosen.

CONNECT: [email protected] / www.linkedin.com/in/sarahjkemp

Never miss a beat with BONUS episode content! Sign up for PRception briefing - https://prceptionmedia.com/subscribe

PR teams are pitching one set of sources. AI is citing another.

Muck Rack research found just 2% overlap between the journalists PR teams pitch most and those AI engines cite when discussing their brands.

Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Iskren Lilov about what this means for media strategy, AI visibility and PR measurement. They explore why recent, niche and context rich coverage matters; why publishers that restrict AI crawling still matter to human audiences; and why AI must be treated as an additional stakeholder, not the only one.

They also confront a bigger measurement problem - if AI visibility becomes another vanity metric, has the industry learned anything from impressions and AVEs?

This is a practical conversation for PR leaders, communications teams and founders trying to build authority with both humans and machines.

PRception is a podcast about how authority actually forms.

Before anyone makes a decision, they decide what feels credible, what they trust and what they believe. Now AI is changing how that happens, before anyone speaks to you.

Hosted by Sarah Kemp, narrative architect and founder of SJK Labs.

Subscribe for conversations about credibility, authority, AI discoverability and the evidence that shapes who gets trusted, shortlisted and chosen.

CONNECT: [email protected] / www.linkedin.com/in/sarahjkemp

Never miss a beat with BONUS episode content! Sign up for PRception briefing - https://prceptionmedia.com/subscribe

AI can produce generic content at a scale that once required entire teams. That does not make the content valuable.

Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Rick Martin, Global Content Director at PageGroup, about what happens when everyone has access to the same content-production cheat code.

Rick’s career began in a “content sweatshop,” producing and editing 30 SEO articles a day. Now he leads content strategy in a world where an LLM can generate that volume almost instantly.

They discuss why more content no longer creates an advantage, why businesses should be cautious about placing every bet on GEO, and what makes original research involving 60,000 people fundamentally different from something generated in an afternoon.

The challenge is no longer producing content. It is producing something credible, useful and difficult for a competitor to recreate with the same prompt.

PRception is a podcast about how authority actually forms.

Before anyone makes a decision, they decide what feels credible, what they trust and what they believe. Now AI is changing how that happens, before anyone speaks to you.

Hosted by Sarah Kemp, narrative architect and founder of SJK Labs.

Subscribe for conversations about credibility, authority, AI discoverability and the evidence that shapes who gets trusted, shortlisted and chosen.

CONNECT: [email protected] / www.linkedin.com/in/sarahjkemp

Never miss a beat with BONUS episode content! Sign up for PRception briefing - https://prceptionmedia.com/subscribe

Most organisations do not lack AI ideas. They lack the operating model, ownership and behaviour change required to turn those ideas into value.

Sarah Kemp, Narrative Architect and Founder of SJK Labs, speaks with Yemi Olagbaiye about what he calls “AI theatre”: borrowing the language of transformation without doing the difficult work needed to change how the business operates.

They discuss why “What is our AI strategy?” is often the wrong question; why a successful pilot proves very little without changes to workflows and decisions; and why leadership requires taste, focus and courage—not merely faster adoption.

Yemi also examines what humans contribute in an agentic workplace, how apprenticeships may need to change, why company websites are losing control of the corporate narrative and what it would take to build a genuinely connected, thinking organisation.

This episode is for leaders who want AI to change business performance—not simply improve the story told in the boardroom.

PRception is a podcast about how authority actually forms.

Before anyone makes a decision, they decide what feels credible, what they trust and what they believe. Now AI is changing how that happens, before anyone speaks to you.

Hosted by Sarah Kemp, narrative architect and founder of SJK Labs.

Subscribe for conversations about credibility, authority, AI discoverability and the evidence that shapes who gets trusted, shortlisted and chosen.

CONNECT: [email protected] / www.linkedin.com/in/sarahjkemp