The AI Why with Liam Lawson

In this episode, Swati Trehan, co-founder of Ema, breaks down what AI agents actually are, how “AI employees” work inside Fortune 500 companies, and why the future of enterprise software may look nothing like today’s SaaS tools.

Swati explains how Ema’s platform orchestrates teams of AI agents that can autonomously handle HR, IT, finance, onboarding, payroll, employee support, and customer service workflows across massive organizations. She also reveals how companies like Hitachi are already deploying AI employees at scale, why traditional automation failed, and how enterprise AI is evolving beyond simple copilots into fully agentic systems.

The conversation dives deep into the technical infrastructure behind agents, including memory, orchestration layers, knowledge graphs, model routing, and why Ema uses multiple LLMs simultaneously to optimize for cost, latency, and accuracy. Swati also shares why Excel remains one of AI’s hardest unsolved problems, why video is the next frontier for agents, and how the “SaaS apocalypse” is reshaping software businesses.

If you’ve been hearing terms like agents, autonomous workflows, AI employees, copilots, or agentic AI, this is one of the clearest explanations of where the technology is heading and what it means for the future of work.

Key Topics Covered:

What AI agents actually are (explained simply)

The difference between copilots, agents, and AI employees

Why traditional automation and RPA failed

How Fortune 500 companies are deploying AI employees today

Why HR is becoming the entry point for enterprise AI adoption

How Ema orchestrates teams of agents across workflows

The technical stack behind enterprise AI agents

Why memory, context, and permissions are critical for agents

The “mixture of experts” approach using multiple LLMs at once

Why Excel remains surprisingly difficult for AI systems

The next frontier: AI-generated video workflows

The rise of the “SaaS apocalypse”

Why solving business problems matters more than building features

How AI is changing the way founders and engineers think

Episode Timestamps:

00:00 - Intro

00:34 - What AI agents actually are

03:01 - The difference between agents and AI employees

03:25 - Liam’s “light bulb” moment using agents

04:06 - Swati’s realization that HR work could be automated

05:57 - The founding story behind Ema

08:20 - Why AI unlocks human creativity

09:20 - The technical infrastructure behind AI agents

12:12 - How Ema routes tasks across multiple LLMs

13:49 - Memory, context, and knowledge graphs for agents

16:35 - The biggest unsolved problems in AI agents

18:32 - Why video is the next frontier for AI

20:05 - Why Excel is still difficult for AI systems

21:00 - Who Ema’s ideal customers are

23:27 - Why HR teams are leading enterprise AI adoption

24:25 - How enterprise AI implementation actually works

26:13 - Why modular agents matter

28:35 - What the employee experience looks like with AI agents

30:24 - Live demo of Ema’s AI employee system

36:58 - How companies roll out AI agents internally

39:31 - Building AI employees in real time

44:01 - Ema’s competitive moat in the AI race

47:46 - The “SaaS apocalypse” and future-proofing AI businesses

49:16 - Why Ema focused on product over hype

52:12 - How AI changed the way Swati thinks

55:07 - Why rapid problem-solving matters more than ever

57:27 - Living in London while building a global AI company

59:16 - Why Swati does what she does

Swati Trehan’s Socials:

LinkedIn: https://www.linkedin.com/in/swati-trehan/

Ema: https://www.ema.co

Partner Links

Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass

Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe

Join the community: https://www.theaireport.ai/leaders-launch-guide

Learn more about your ad choices. Visit megaphone.fm/adchoices

Anthropic says fictional portrayals of AI may have influenced Claude’s recent blackmail behavior during internal testing. OpenAI officially launches a $4B enterprise deployment company. And Elon Musk’s xAI faces lawsuits over controversial power infrastructure at its data center.

This week, Anthropic explains why Claude Opus 4 attempted to blackmail engineers during safety tests, OpenAI expands aggressively into enterprise AI services with engineers embedded directly inside companies, OpenAI launches a new cybersecurity platform called Daybreak, Anthropic officially surpasses OpenAI in B2B adoption according to new Ramp data, and xAI faces growing scrutiny over gas turbines powering its AI infrastructure.

If you are a founder, operator, or executive trying to stay ahead of AI, this is your weekly AI news briefing every Tuesday.

Stories Covered This Week:

Anthropic says internet culture and fictional AI portrayals may have influenced Claude’s blackmail behavior during testing

OpenAI launches “The Deployment Company” with more than $4B in backing to help enterprises rebuild workflows around AI

OpenAI unveils Daybreak, a GPT-5.5 powered cybersecurity platform competing with Anthropic’s Mythos

Anthropic officially passes OpenAI in B2B adoption according to new Ramp data

Elon Musk’s xAI faces lawsuits over gas turbines powering its Mississippi data center site

Timestamps:

00:00 Intro

00:18 Claude’s blackmail behavior explained by Anthropic

01:15 OpenAI launches The Deployment Company

02:30 OpenAI enters cybersecurity with Daybreak

03:29 Anthropic surpasses OpenAI in B2B adoption

04:47 xAI faces environmental backlash over AI power demands

05:55 Outro

Partner Links

Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass

Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe

Join the community: https://www.theaireport.ai/leaders-launch-guide

Learn more about your ad choices. Visit megaphone.fm/adchoices

In this episode, Jason Li, CTO of Laurel, reveals how the company is turning timesheets into the AI playbook for the entire knowledge-work economy. Jason breaks down why $2,000/hour lawyers still spend Saturdays manually filling out time in six-minute increments, how Laurel's AI platform automatically captures every click, email, and meeting, and why data (not code) is the only real moat left in the age of the SaaSpocalypse.

Jason shares how Ernst & Young is using Laurel to identify high-leverage work, why Laurel deliberately integrates with "decades-old" software like Classic Outlook that most startups ignore, and the counter-intuitive reason your best rainmakers should never be forced into cookie-cutter roles again. He also explains why Laurel doesn't train its own LLM, how they run AI feedback loops that self-iterate prompts, and the frameworks leaders can use to actually measure AI ROI instead of just surveying "did it help?"

Key Topics Covered:

Why "what gets measured gets managed" is the most important rule in AI adoption

The Moneyball insight that changed how Jason thinks about metrics

How Laurel auto-generates timesheets for lawyers and accountants

Why Ernst & Young chose Laurel for their tax group

The hidden cost of manual timesheets for $2K/hour professionals

How Laurel maps knowledge work to a company's "work ontology"

Why decades-old software (Classic Outlook) is a competitive moat, not a liability

The SaaSpocalypse: what survives when AI eats applications

How to measure if an AI tool actually delivers ROI

Why data, not models, is the real defensible asset in AI

Episode Timestamps:

00:00 - Intro

00:25 - The Peter Drucker quote that shaped Jason's career

02:49 - A Moneyball analogy for AI adoption

03:25 - What Laurel actually does: the AI platform that maps time to outcomes

07:19 - Why every business (not just law firms) needs time visibility

09:17 - Inside the Ernst & Young deployment

12:27 - Jason's journey to becoming CTO at Laurel

14:21 - Live product demo: Laurel's work ontology engine

17:49 - How AI shifts the line between high and low leverage work

21:15 - What onboarding a 2,000-person firm actually looks like

23:06 - The technical architecture behind Laurel's desktop client

28:35 - Why Laurel doesn't train its own LLM

29:39 - How Laurel handles AI models "getting worse" overnight

33:35 - Capturing time for work that doesn't happen on a computer

37:17 - AI adoption meets employee behavior change

41:54 - The SaaSpocalypse and why Laurel's moat is data, not software

48:00 - Why Jason left Ironclad to join Laurel

51:16 - Jason's answer to The AI Why's signature closing question

Jason Li's Socials:

LinkedIn: https://www.linkedin.com/in/jasonhli/

Laurel: https://www.laurel.ai

Partner Links

Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass

Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe

Join the community: www.theaireport.ai/leaders-launch-guide

Learn more about your ad choices. Visit megaphone.fm/adchoices

The Pentagon just signed eight AI deals and left Anthropic out. Wall Street is helping both Claude and ChatGPT move deeper into companies. And Elon Musk went from calling Anthropic evil to powering it with his own data center.

This week, the Pentagon clears eight major tech companies for classified military AI while Anthropic fights back in court, Anthropic and OpenAI both launch enterprise AI services businesses backed by Wall Street giants, GPT-5.5 Instant becomes ChatGPT's default model with significantly fewer hallucinations, Greg Brockman takes the witness stand in Elon Musk's lawsuit against OpenAI, and Anthropic signs a deal with SpaceX to use the Colossus 1 data center that Musk built for his own AI company.

If you are a founder, operator or executive trying to keep up with AI, this is your weekly news briefing every Tuesday.

Stories Covered This Week:

The Pentagon signs AI agreements with Google, OpenAI, Microsoft, Nvidia, AWS, SpaceX, Oracle and Reflection while Anthropic fights to stay on the list

Anthropic launches a $1.5B enterprise AI services company with Blackstone and Goldman Sachs while OpenAI finalizes a $10B joint venture called The Deployment Co

GPT-5.5 Instant is now ChatGPT's default model, making 52.5% fewer errors on high stakes questions

Greg Brockman testifies in Elon Musk's OpenAI lawsuit, revealing tense scenes from the company's early days

Anthropic signs a deal with SpaceX to use Colossus 1, the data center Musk built for xAI, with talks of orbital AI infrastructure to follow

Timestamps:

00:00 Intro
00:20 Pentagon signs eight AI deals, Anthropic left out
01:53 Anthropic and OpenAI launch enterprise AI services businesses
03:30 GPT-5.5 Instant becomes ChatGPT's default model
04:53 Greg Brockman testifies in Elon Musk's OpenAI lawsuit
06:23 Elon Musk goes from calling Anthropic evil to powering Claude

08:10: Outro

Partner Links:

Upgrade your AI toolkit: https://www.theaireport.ai/ai-executive-pass

Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe

Join the community: www.theaireport.ai/leaders-launch-guide

Learn more about your ad choices. Visit megaphone.fm/adchoices

22,000 LinkedIn followers to over 210,000 in just over a year. Charlie Hills did not get there by luck. He got there by reverse engineering what already works, posting with intention every single day, and building systems with AI that most people have not even heard of yet.

In this second appearance on The AI Why, Liam Lawson sits down with Charlie Hills to break down exactly how he goes viral across LinkedIn, Instagram, and TikTok, what his full content creation process looks like with Claude, and why he just launched a SaaS tool built entirely from his own workflows inside Claude Code.

Stories Covered This Week:

The pre-validated content strategy that got Charlie 11 million views on his first Instagram reel

The CHEF framework for creating content that does not sound like AI wrote it

How Charlie uses Appify, Claude Code and Gemini together to reverse engineer viral content

Why organic reach on LinkedIn is dying and what to do about it

Claude Code for infographics, carousels, and eventually video editing from the terminal

Vislo, the branded infographic SaaS tool Charlie just launched in public beta

Pattern interrupts, B-roll clarity and what actually stops the scroll on Instagram

Going from 9 to 5 employee to full time creator in nine months

Episode timestamps:

00:00 Intro and follower count update
01:03 How Charlie goes viral on any platform
04:01 Steal Like an Artist and pre-validated content
04:53 What structure works on Instagram reels
07:36 Speaking coach, scripting and sounding human
09:06 The full LinkedIn content creation process
10:03 Posting cadence and why twice a day backfired
12:19 The CHEF framework explained
15:07 Appify, Claude Code and scraping for content research
17:04 Claude Co-Work scheduled tasks and workflows
20:22 Going deep into Claude Code for infographics
22:06 How Claude Code memory and the MD file works
24:41 Combining Gemini API inside Claude Code for Instagram scripts
26:53 Where AI ends and humans begin in the content process
30:36 Remotion and the future of AI video editing
33:33 Over-engineering with AI and why it is addictive
35:02 The story behind Vislo, the branded infographic SaaS
40:22 Beta feedback and pricing
42:01 Done For You vs SaaS customers
43:57 Expanding beyond LinkedIn to Instagram, TikTok and YouTube
47:12 What the ideal future of work looks like for Charlie
49:12 The Mexico LinkedIn creator house with Mischa and Corey
53:03 University, failed business and the nine month grind
01:00:01 Did not leave the nine to five to earn more, but to live more
01:00:55 Why do you do what you do?

Partner Links
Subscribe to our free newsletter: https://newsletter.theaireport.ai/subscribe

Join the community: www.theaireport.ai/leaders-launch-guide

Learn more about your ad choices. Visit megaphone.fm/adchoices