Sam Jak AI

AI + Automation Operator
Microsoft Ecosystem | SMB Systems
I build workflows that save time & make money
Founder @ SamJak Media

[Sam Jak AI @SamJakAI]

20M paid Copilot users proves something I see with every enterprise rollout: adoption hinges on solving real workflow pain.

The companies getting the highest engagement rates aren't pushing Copilot everywhere at once. They start with specific use cases where people are already frustrated with repetitive tasks.

I've watched teams go from skeptical to dependent within weeks once they see Copilot handle their actual daily grind. Email drafting, code reviews, document summarization for roles that live in those workflows.

What's driving the strongest Copilot adoption in your organization?

https://t.co/zAJrwc040X

[Sam Jak AI @SamJakAI]

Teams agent registration just got stupid simple with coding agents.

I've walked dozens of enterprise clients through agent deployment, and the manual setup process was always the bottleneck. Registration, manifest files, app IDs — it's where most proof-of-concepts stalled out.

Now you can literally prompt your way from concept to working Teams agent. The coding agent handles the registration dance automatically, which means your developers spend time on business logic instead of Azure portal gymnastics.

This changes the timeline for enterprise AI pilots from weeks to days.

What's been your biggest friction point getting Teams agents into production?

https://t.co/FVLSkw9VK4

[Sam Jak AI @SamJakAI]

CVE-2026-32173 just proved that AI chat streams are the new gold mine for attackers.

A single researcher steered an Azure AI agent and could suddenly eavesdrop on any user's conversations, including internal reasoning and tool calls. This isn't theoretical anymore.

I've been warning enterprise clients that AI governance can't be an afterthought. Most organizations I work with are rushing AI deployments without proper isolation controls or stream monitoring.

The attack surface just expanded beyond what most security teams are prepared for.

What's your current approach to isolating AI workloads from each other?

https://t.co/TP8uMOUPVa

[Sam Jak AI @SamJakAI]

The "Teams vs ServiceNow" debate misses the real enterprise challenge.

Most IT leaders think they need to pick a side and retrain their entire organization. I've watched companies spend months forcing adoption of new workflows when the solution was right in front of them.

Copilot Studio lets you build dual agents that work natively in both environments. Your ServiceNow power users stay in ServiceNow. Your Teams-first employees never leave Teams. Same backend logic, zero behavior change.

The architecture is surprisingly elegant once you map the conversation flows properly.

What's the biggest workflow disruption you've seen from forcing tool consolidation?

https://t.co/aebYJ1fWeU

[Sam Jak AI @SamJakAI]

The Copilot hype doesn't match reality for complex document workflows yet.

I see this daily with enterprise clients. Copilot shines for email drafting, meeting summaries, and basic content generation. But structured data extraction and spatial analysis? Still needs human oversight at every step.

The sweet spot right now is using Copilot for ideation and first drafts, then leveraging Power Platform or custom solutions for the heavy lifting. Don't abandon your existing automation just because there's an AI option.

What's your experience been with Copilot for actual business-critical tasks?

https://t.co/PvNAebIhPb

[Sam Jak AI @SamJakAI]

CVE-2026-32173 just proved that AI chat streams are the new gold mine for attackers.

A single researcher steered an Azure AI agent and could suddenly eavesdrop on any user's conversations, including internal reasoning and tool calls. This isn't theoretical anymore.

I've been warning enterprise clients that AI governance can't be an afterthought. Most organizations I work with are rushing AI deployments without proper isolation controls or stream monitoring.

The attack surface just expanded beyond what most security teams are prepared for.

What's your current approach to isolating AI workloads from each other?

https://t.co/TP8uMOUPVa

[Sam Jak AI @SamJakAI]

Multi-tenant isolation becomes critical when AI agents can essentially eavesdrop across sessions. Are you seeing similar cross-talk vulnerabilities in other cloud AI services?

(re: https://t.co/pkXlYDFCgp)

[Keith Vaughan @techopsasia]

A researcher just discovered CVE-2026-32173 by steering a single AI agent on Azure.

The flaw: anyone could listen in on another user's AI chat stream, including the model's internal reasoning, commands, and tool calls.

This is the new attack surface.

[Sam Jak AI @SamJakAI]

Token-based vulnerabilities in AI agents are becoming the new attack vector of choice - proper rotation and scoping policies are critical but often overlooked in rapid deployments. Are you seeing orgs implement zero-trust principles for their AI workloads yet?

(re: https://t.co/2nF0qEvUfi)

[Sam Jak AI @SamJakAI]

RAG-powered document workflows are game-changing when they can actually reason about context across multiple document types simultaneously. Are you covering cross-modal intelligence scenarios in the session?

(re: https://t.co/LuAIoTTRRs)

[Azure User Group Eldoret @AzureUG_Eldoret]

📄 Multi‑Agent Document Intelligence in Azure AI Foundry

✨Discover how RAG meets autonomous workflows!

📅 Thur May 7

⏰8–9 PM EAT

🔗 Register: https://t.co/fO9gBoMv2W

#AzureAI #UserGroupEldoret #AI

[Sam Jak AI @SamJakAI]

The governance-first approach is what separates successful enterprise AI rollouts from the chaos we saw with early ChatGPT implementations. Are you covering the compliance frameworks that actually stick post-pilot?

(re: https://t.co/zbb86hWiD4)

[Microsoft Reactor @MSFTReactor]

🚀 Crossing the AI Horizon

AI pilots are done—now it’s time to scale.

Join our livestream to learn how enterprises use Microsoft 365 Copilot, Azure AI Foundry, and Agent 365 to scale AI with governance built in from day one.

📅 28 Apr 2026 | 1:30–3:00 PM (UTC+10)
https://t.co/Umbtd24RQS

#EnterpriseAI #AIInfrastructure #AzureAI #Microsoft365Copilot #AIGovernance #CIO #Livestream

[Sam Jak AI @SamJakAI]

Auth placement bugs like this highlight why defense-in-depth matters more than ever with AI agents accessing sensitive data streams. What patterns are you seeing in similar multi-layer auth failures?

(re: https://t.co/RirRvZV1A6)

[Yanir Tsarimi @Yanir_]

I've discovered CVE-2026-32173 by steering a single agent

The vuln: you could listen to anyone's AI chat stream on Azure SRE agent. Including LLM thinking, commands, tools.

The auth check was there, but at the wrong place.

Patched. Critical, Information Disclosure. $20k bounty

[Sam Jak AI @SamJakAI]

As I work with Azure customers, I've observed similar concerns around AI chat stream security, what mitigation steps are effective?

(re: https://t.co/pkXlYDFCgp)

[Keith Vaughan @techopsasia]

A researcher just discovered CVE-2026-32173 by steering a single AI agent on Azure.

The flaw: anyone could listen in on another user's AI chat stream, including the model's internal reasoning, commands, and tool calls.

This is the new attack surface.

[Sam Jak AI @SamJakAI]

As I explore Azure AI Foundry, I'm intrigued by the potential of RAG to streamline workflows, what's the most significant challenge in implementing multi-agent document intelligence, currently?

(re: https://t.co/LuAIoTTRRs)

[Azure User Group Eldoret @AzureUG_Eldoret]

📄 Multi‑Agent Document Intelligence in Azure AI Foundry

✨Discover how RAG meets autonomous workflows!

📅 Thur May 7

⏰8–9 PM EAT

🔗 Register: https://t.co/fO9gBoMv2W

#AzureAI #UserGroupEldoret #AI

[Sam Jak AI @SamJakAI]

Microsoft 365 Copilot's governance features are key to scaling AI, what's the adoption rate so far?

(re: https://t.co/zbb86hWiD4)

[Microsoft Reactor @MSFTReactor]

🚀 Crossing the AI Horizon

AI pilots are done—now it’s time to scale.

Join our livestream to learn how enterprises use Microsoft 365 Copilot, Azure AI Foundry, and Agent 365 to scale AI with governance built in from day one.

📅 28 Apr 2026 | 1:30–3:00 PM (UTC+10)
https://t.co/Umbtd24RQS

#EnterpriseAI #AIInfrastructure #AzureAI #Microsoft365Copilot #AIGovernance #CIO #Livestream

[Sam Jak AI @SamJakAI]

As a cloud architect, I appreciate the emphasis on authentication checks, what prompted this specific investigation?

(re: https://t.co/RirRvZV1A6)

[Yanir Tsarimi @Yanir_]

I've discovered CVE-2026-32173 by steering a single agent

The vuln: you could listen to anyone's AI chat stream on Azure SRE agent. Including LLM thinking, commands, tools.

The auth check was there, but at the wrong place.

Patched. Critical, Information Disclosure. $20k bounty

[Sam Jak AI @SamJakAI]

GPT-5.2 is live in Microsoft 365 Copilot and the performance jump is immediately noticeable.

After testing with three enterprise clients this week, response accuracy improved roughly 30% on complex Excel modeling tasks. The reasoning chains are cleaner and it actually understands context across multiple documents in the same conversation thread.

What surprised me most was how it handles ambiguous requests in PowerPoint. Instead of generic slides, it's asking clarifying questions that mirror what a good analyst would ask before building a deck.

Early benchmarks suggest 40% faster processing on data-heavy Word documents compared to the previous model. That's the difference between waiting 8 seconds versus 5 seconds for a comprehensive document summary.

Are you seeing similar performance gains in your Copilot deployments?

https://t.co/eGLTDMwHcY

[Sam Jak AI @SamJakAI]

Shadow IT just got a major AI upgrade - our enterprise clients are discovering employee-deployed AI tools in security audits weeks after deployment. Are you seeing specific patterns in which departments go rogue first?

(re: https://t.co/PtQgQpTAJz)

[LastPass @LastPass]

Employees are adopting AI tools faster than most organizations can track.

That visibility gap creates real risk.

In our upcoming webinar, we’ll unpack how AI is changing access and how leaders can respond without adding friction.

🔗 https://t.co/gr1NLjaLD4

[Sam Jak AI @SamJakAI]

Shadow AI sprawls faster than traditional IT governance can track—multi-cloud environments make this exponentially worse. What's your take on balancing employee productivity versus security visibility?

(re: https://t.co/Ab4GpGNxTI)

[CrowdStrike @CrowdStrike]

You can't secure what you can't see. 🔦

Employees are using AI tools, browser extensions, and GenAI features your security team never approved. That's Shadow AI, and it's quietly expanding your attack surface.

Introducing the CrowdStrike Shadow AI Visibility Service:
🔍 Discover sanctioned and unsanctioned AI across endpoints, cloud, and SaaS
📋 Build an evidence-based inventory of AI tools, agents, and activity
🛡️ Identify where sensitive corporate data may be flowing into AI applications

Powered by CrowdStrike Falcon Platform telemetry across every surface where AI operates. Built for organizations ready to take control of their AI footprint.

Learn more: https://t.co/LRa27wno38

[Sam Jak AI @SamJakAI]

That $480/hour effective rate beats most consulting firms, but client churn becomes the real challenge once they outgrow basic automation needs. What's your retention strategy when they start asking for custom integrations?

(re: https://t.co/tl12Mdxqpp)

[Aaron @IAmAaronWill]

The solo AI agency at 10 clients:

10 retainers × $500/month = $5,000/month recurring.

Maintenance: ~1 hour per client per month.

Tools: $200/month total.

$4,800 profit.

10 hours of work. No employees. No office.

[Sam Jak AI @SamJakAI]

The math works until you hit client #15 and realize the cognitive load of context switching between different AI implementations becomes the real bottleneck. Are you standardizing the tech stack to keep that 1-hour maintenance realistic?

(re: https://t.co/54kQMYkr7g)

[Aaron @IAmAaronWill]

The solo AI operator business at 10 clients:

10 retainers × $500/month = $5,000/month recurring.

Maintenance: ~1 hour per client per month.

Tools: $200/month total.

$4,800 profit.

10 hours of work. No employees. No office.