Podcast
AI at Work
Hosted by Elijah Szasz, Kevin Williams · EN
AI at Work is hosted by Elijah Szasz (CEO of SPARK6) and Kevin Williams (CEO of Ascend Labs). Each week delivers actionable insights on how artificial intelligence is reshaping jobs today, and how you can use it to advance your career. We unpack real-world tools, automation workflows, and emerging roles so you can stop chasing hype and start using AI to get things done. Whether you’re an individual contributor, team lead, or business owner, we show you what to build, what to ask for, and what to look for in a workplace moving at machine-speed.
19episodes
Episodes
Newest firstAll episodes
When Models Lie and You're Still Paying for the Tokens
2d ago00:55:50Tap to summarizeWhen AI models hallucinate confidently while burning through your token budget, you need more than better prompts you need better processes.In this episode, Kevin and Eli explore the reality of working with AI that lies while you pay, the evolution from saved prompts to voice-triggered workflows, and why validation loops matter more than model updates. From enterprise governance requirements to tech stack consolidation, this conversation cuts through the hype to reveal what actually works when deploying AI in real business contexts.Key topics covered:✅ Why model reliability requires systematic validation, not perfect AI✅ The shift from skills to snippets and voice-triggered workflows ✅ Enterprise governance requirements when litigation risk is involved✅ Tech stack consolidation as AI tools mature beyond experimentation✅ Cost economics driving procurement strategy changes✅ Building quality control loops that catch hallucinations before implementationTIMESTAMPS:00:00 — Intro and Sora commercial viability discussion05:30 — Video generation landscape: Veo, SeeDance, Higgs Field15:00 — Model quality regression and hallucination experiences 25:00 — Skills vs snippets: workflow evolution deep dive35:00 — Enterprise governance and validation requirements45:00 — Tech stack consolidation and platform comparisons53:00 — Eli's transition and mental health focusShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
Why Your AI Productivity Wins Are Making You Miserable
1w ago00:56:50Tap to summarizeKevin and Eli dive deep into the hidden costs of AI productivity gains, exploring how cognitive overload and social isolation are creating a new form of workplace burnout that organizations aren't prepared to address.In this candid conversation, they share their personal experiences with AI work density accomplishing 80-100x more work daily while feeling more exhausted than ever. The discussion reveals why the promise of AI efficiency is backfiring without proper human-centered implementation.Key topics covered:✅ AI work density and the cognitive load crisis✅ Why context switching between AI tasks creates decision fatigue✅ The importance of pacing in AI-forward organizations✅ How to identify what should (and shouldn't) be automated✅ The role of human connection in an AI-augmented workplace✅ Platform comparison: OpenAI vs Claude vs Gemini for team adoption✅ The hidden costs of AI agent proliferation✅ Why most organizations miss the pedestrian productivity winsTimestamps:00:00 — Platform switching and AI adoption decisions08:00 — Microsoft integration advantages and limitations15:00 — Team adoption and accessibility considerations25:00 — AI work density and cognitive load crisis35:00 — Human connection vs AI interaction42:00 — What to automate vs what to preserve48:00 — Organizational capture of AI innovations55:00 — Enterprise data integration challengesExplore practical AI implementation: https://assessment.ascendlabs.ai/Book a conversation with Kevin: tidycal.com/kevinwilliamsShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
Is Your Organization Ready or Just Excited?
2w ago00:43:05Tap to summarizeWhat does AI implementation actually look like when you move beyond the hype? Aaron Wilt, CEO of 40-person Pulse V Holdings, shares the unfiltered reality of deploying AI in a mid-market business where technical sophistication meets organizational complexity.This conversation reveals why even leaders who deeply understand AI struggle with organizational deployment, and what it really takes to bridge the gap between AI demos and business transformation. Aaron breaks down the "trust calibration" problem, the three-way intersection most companies lack, and why winners use AI to win harder while others get left behind.Key topics covered:✅ The implementation bottleneck - why technical knowledge isn't enough✅ Trust calibration - when 85% AI accuracy is acceptable vs. dangerous ✅ Building AI teams - pairing skeptics with dreamers for maximum impact✅ Control vs. scale - centralized innovation vs. distributed adoption✅ The SaaSpocalypse - how AI enables internal capability building✅ Risk management - deterministic vs. probabilistic AI applications✅ Team dynamics - getting non-technical staff to adopt AI tools✅ Future-proofing - what to tell kids about AI and careersGet practical AI guidance: https://assessment.ascendlabs.ai/Book a conversation with Kevin: tidycal.com/kevinwilliamsTIMESTAMPS:00:00 — Introduction and Aaron's background03:00 — AI evolution: Chat bot to context to orchestration eras08:00 — The probabilistic nature problem and trust calibration15:00 — Current AI stack and team adoption at Pulse V22:00 — Risk management and process controls28:00 — The SaaSpocalypse and internal capability building31:00 — Organizational challenges and people dynamics38:00 — Developer resistance and team composition42:00 — Future education and closing thoughtsShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
The Microsoft-Claude Connection That Actually Works
3w ago00:38:06Tap to summarizeKevin Williams breaks down the Microsoft-Claude integration that's transforming productivity workflows, plus the legal risks and model competition shaping AI adoption decisions.The Claude add-on for Microsoft Office enables something most people haven't seen yet: lateral document communication. Your email can talk to Word, Word can talk to Excel, Excel can talk to PowerPoint—all seamlessly connected. Kevin explains why this integration matters more than having the smartest AI model, and how it's changing his perspective on Microsoft's productivity suite.The conversation also covers the current state of AI model competition, why different models excel at specific tasks, and the emerging legal liabilities around AI agents making autonomous decisions in organizations. How the Microsoft-Claude integration actually works Why document integration beats AI intelligence for most businesses Current state of GPT vs Claude vs Gemini capabilities Legal risks of AI transcripts and agent decisions Shadow AI use and corporate policy implications The "dead internet" problem with AI-generated content Why boring AI integration creates competitive advantage Practical next steps for Microsoft Office usersAPPROXIMATE TIMESTAMPS:00:00 — Intro and Microsoft-Claude revelation03:00 — How lateral document communication works08:00 — Current AI model comparison and capabilities15:00 — GPT agents vs Claude computer use20:00 — Team plans and organizational AI adoption25:00 — Risk tolerance for AI agent deployment30:00 — The "dead internet" and content authenticity35:00 — Legal liabilities and AI policy implications40:00 — Wrap-up and next episode preview→ Get practical AI guidance: https://assessment.ascendlabs.ai/→ Book a strategy conversation: tidycal.com/kevinwilliamsShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
When Your Team Says Claude But Means GPT
May 501:02:16Tap to summarizeThe AI industry loves throwing around the word 'agents,' but most teams are still stuck in basic prompting mode. Kevin and Eli cut through the semantic noise to reveal what actually matters: sophisticated automation is now accessible through natural language, not technical configuration.In this episode, they explore the practical reality of moving from one-off prompts to systematic workflows, why the 'agent' versus 'automation' debate misses the point, and how natural language interfaces are removing technical barriers that used to require specialized workflow knowledge.Key topics covered:✅ Why most people are still just prompting instead of building workflows✅ How natural language makes complex automation accessible✅ The practical difference between projects, automations, and agents✅ Real examples of workflow automation without technical expertise✅ Why focusing on results beats debating terminology✅ Moving from ChatGPT tabs to systematic AI integrationThis isn't about the latest AI hype – it's about practical transformation that works Monday morning.TIMESTAMPS:00:00 — Future of AI and robotics discussion08:16 — Current state of enterprise AI adoption16:30 — Job displacement and economic impact25:40 — Moving beyond basic prompting35:20 — Context and platform lock-in42:30 — Agents vs automations semantics52:00 — OpenAI agents vs Claude workflows58:30 — Real-world automation examplesShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
When Your AI Budget Hits Your Salary
Apr 2900:55:22Tap to summarizeThe "token maxing" phenomenon is reshaping how organizations think about AI budgets, but most companies are asking the wrong questions about AI spending.In this episode, Kevin and Eli explore the reality behind engineers burning through massive token budgets - sometimes exceeding their own salaries - and what it means for practical AI adoption in mid-market companies.From Stockholm engineers outspending their paychecks on Claude to Jensen Huang's $250K token requirements, we break down why most organizations need output-focused spending strategies, not ego-driven token consumption.Key topics covered:✅ The token maxing phenomenon and what's driving it✅ Why most mid-market companies don't need massive AI budgets✅ The difference between productive AI spending and token burning✅ How to build sustainable AI strategies that survive subsidy endings✅ Real-world examples of agents running amok overnight✅ Microsoft's new agentic capabilities in Office suite✅ Platform comparison: OpenAI vs Anthropic vs Google for different use casesTIMESTAMPS:00:00 — Intro and token maxing overview02:30 — What token maxing actually means05:45 — Jensen Huang's $250K token requirement08:15 — Mid-market reality vs Silicon Valley hype12:00 — Agent sprawl and overnight token burns18:30 — Microsoft's new agentic Office features25:40 — AI subsidy era and pricing reality32:45 — Platform wars: choosing your AI stack42:00 — Practical token budgeting strategies48:50 — Future of AI pricing modelsShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
Why Your Chief of Staff Dreams About You
Apr 2100:57:24Tap to summarizeWhat if you could run your entire workday through one AI conversation? Kevin and Eli explore the emerging 'monothread' format that's revolutionizing how teams operate - plus the hidden security risks that amateur AI builders are creating.In this episode, we dive deep into how the monothread approach eliminates app switching by connecting your email, calendar, tasks, and CRM into one continuous AI conversation. But we also cover the reality: it's still janky to set up, the security vulnerabilities are real, and most organizations aren't ready.We also discuss Claude Design's launch that sent Figma's stock tumbling, why Canva is positioned to survive the AI design revolution, and the critical security practices every AI experimenter needs to know.✅ Key Topics Covered:✅ The monothread revolution and how to build your AI chief of staff✅ Why Claude Design might end Figma (and what that means for designers)✅ Security nightmares: API key protection and the amateur builder problem✅ Platform comparison: Claude vs GPT vs Gemini for business use✅ The hidden costs of AI tool proliferation✅ Voice-first AI workflows and their psychological impactTimestamps:00:00 — Intro and sleep tracking with AI05:15 — Main quest vs side quest in AI adoption12:40 — The monothread format explained20:30 — Voice-first AI and dream psychology25:15 — Claude Design launch and Figma's response35:45 — Platform comparison and subscription costs45:20 — Security vulnerabilities in amateur AI apps55:00 — API protection and credential rotationShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
When Your Entire AI Team Calls in Sick at 6 AM
Apr 1600:57:31Tap to summarizeWhen Anthropic's Claude went down at 6:15 AM on tax day, it exposed a critical blind spot that most AI-adopting organizations haven't considered: vendor dependency risk.In this episode, Kevin and Eli dive deep into what happened when their entire AI-powered workflow ecosystem crashed simultaneously, revealing the hidden dangers of building business operations around single AI providers without backup plans.This conversation goes beyond the surface frustration of a service outage to explore the fundamental readiness gap that exists when organizations treat experimental AI services like established infrastructure. You'll discover why this isn't just a technology problem, but an organizational planning problem that requires immediate attention.Key topics covered:✅ The "heroin dealer problem" - what happens when AI dependency meets reality✅ Hidden costs of API pricing vs subscription models and recent Anthropic changes✅ Why AI subsidies are ending and what it means for your budget✅ Building redundancy into AI-powered business operations✅ The Mythos model leak and cybersecurity implications for all businesses✅ Practical security steps every organization must take nowApproximate timestamps (verify against recording):00:00 — Introduction and the morning Claude went down02:49 — The heroin dealer analogy and dependency realization07:15 — Understanding AI subscription vs API pricing models15:19 — Anthropic's April 4th changes and the end of workarounds28:11 — Real cost examples: $200/month to $7000/month overnight40:05 — Content creation, AI slop, and the attention economy52:32 — Mythos model leak and cybersecurity implications57:37 — Practical security steps: passwords, 2FA, and analog safeguardsShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
The First Day You Stop Opening Apps.
Apr 800:53:55Tap to summarizeKevin Williams & Elijah Szasz demonstrates the end of interface friction by consolidating his entire workflow into a single Claude conversation. No more jumping between ClickUp, HubSpot, and Slack - everything happens through natural conversation with AI connectors pulling and pushing data to the right systems.This episode explores a fundamental shift happening right now: your SaaS tools are becoming expensive databases with unnecessary user interfaces. The real productivity breakthrough isn't better tools - it's eliminating the need to context-switch between tools entirely.Kevin shares his six-day experiment of running his entire business through one chat window, including project management, CRM updates, team coordination, and strategic planning. The result? Massive time savings and the elimination of what he calls "administrative fiddliness."✅ Key Topics Covered:✅ How to build Claude connectors for seamless workflow management✅ Why most productivity problems are actually interface problems✅ The coming SaaS revolution and what it means for business software✅ Practical strategies for consolidating multiple tools into single conversations✅ The psychology of context-switching and why it kills momentum✅ Future predictions for AI-powered workflow consolidationTIMESTAMPS:00:00 — Intro04:49 — The SaaS apocalypse conversation18:15 — Creative industry disruption24:49 — The end of fiddliness breakthrough35:07 — Building the single-interface workflow41:09 — Platform connectors and automations53:33 — The future of business software interfacesShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →
The Best Productivity Hack Right Now? Dictation
Apr 100:46:54Tap to summarizeKevin and Eli dive deep into the productivity unlock everyone's missing: dictation. Kevin reveals he's crossed one million words dictated and shares why voice input isn't just faster - it generates 3-5x more context-rich data that dramatically improves AI responses.This episode explores the gap between AI hype and practical implementation, covering computer use capabilities in Claude, the challenges of brittle workflows, and why the best AI adoption strategies focus on reducing input friction rather than upgrading models.The conversation touches on platform friction, the emerging agent economy, and why successful AI implementations meet people where they already communicate best - through speech.✅ Key topics covered:✅ Why dictation beats typing for AI adoption✅ Computer use vs browser use capabilities✅ Data density as the real AI productivity unlock✅ Platform friction and workflow brittleness✅ Voice interfaces for reluctant AI users✅ The knowledge graph beyond the office✅ Practical tips for Claude Cowork and automationTimestamps:00:00 — Intro and Anthropic leak discussion03:00 — Platform evolution and agent capabilities07:00 — Computer use experiments and failures15:00 — Beehive platform friction story25:00 — Dictation productivity breakthrough35:00 — Voice input and data density insightsShow Notes & Links: https://www.spark6.com/podcast Submit listener questions: elijah@spark6.comkevin@ascendlabs.ai Check out Kevin’s stuff:Ascend Labs: https://ascendlabs.ai/Follow Kevin on LinkedIn: https://www.linkedin.com/in/kevinguywilliams/ Check out Eli’s Stuff:SPARK6 Agency: https://www.spark6.com/Sign up for FREE AI Framework Friday Newsletter: https://www.spark6.com/newsletterFollow Elijah on LinkedIn:https://www.linkedin.com/in/elijahszasz/
Transcribe →