Insight On

GitHub Copilot's billing model changed on June 1 — and most enterprise teams are about to see their first full bill under the new system. For many organizations, it'll be the first real look at what AI adoption actually costs at scale. 

The shift from flat-rate to usage-based billing sounds straightforward. Pay for what you use. But what your team is "using" turns out to be a lot more complicated than anyone expected — and the habits that felt like strong AI adoption six months ago are the same ones quietly running up the tab today. 

Sumner Ohye works with GitHub partners and enterprise customers across the Americas and APAC. And since June 1, he's been fielding a lot of calls about billing. Parker Johnston is Insight's Agentic Field CTO — he's in the field with engineering teams every week. Together they get into what's happening on the ground, the common — and costly — mistakes people are making with token usage, and how teams are changing habits to keep token costs in check. 

There's a reason Parker opens with "you can no longer YOLO mode it" — and it's not just about prompting. 

Get GitHub Copilot solutions, available from Insight, because having the right partner means your billing controls, usage dashboards, and workflow guidance are in place before the charges hit — not after: https://www.insight.com/en_US/shop/partner/microsoft/github-copilot.html 

Want to go deeper on GitHub? Check out these resources from Insight: 

Burning Questions: Hot Takes on Developer Tools, AI, & What's Next: https://www.insight.com/en_US/content-and-resources/webinars/burning-questions-hot-takes-on-developer-tools-ai-whats-next.html 


What's New in GitHub Advanced Security: https://www.insight.com/en_US/content-and-resources/webinars/whats-new-in-github-advanced-security-key-updates-explained.html 


Maximizing Developer Efficiency With GitHub Enterprise: https://www.insight.com/en_US/content-and-resources/webinars/maximizing-developer-efficiency-a-guide-to-github-enterprise.html 


Strategies for GitHub Actions Optimization: https://www.insight.com/en_US/content-and-resources/webinars/strategies-for-github-actions-optimization.html 


Maximizing Team Productivity With GitHub Copilot: https://www.insight.com/en_US/content-and-resources/webinars/maximizing-team-productivity-with-github-copilot.html 


Subscribe for more episodes of Insight On: new conversations every week on AI, technology strategy, and what's actually working in enterprise tech. 


#GitHubCopilot #UsageBasedBilling #AgenticCoding  

Chapters:

03:25 — Is a 24x bill increase normal? What actually changed June 1st 


05:38 — The transition period and what enterprises need to do now 


07:14 — Agentic coding: when to use AI vs. doing it yourself 


09:27 — Governing AI agents inside your codebase 


11:24 — How to set up billing thresholds and avoid access shutdowns 


13:28 — Why a third-party partner beats going direct to GitHub 


17:11 — What the GitHub billing shift means for software development's future 


18:20 — Running AI models locally vs. cloud: the real cost comparison 


20:25 — How GitHub Copilot adoption has shifted since June 1st 


21:43 — GitHub's evolution from source control to full agentic platform 


25:34 — Building trust in agentic workflows: cost, security, and code quality 


28:07 — The biggest mistakes teams are making with model selection and context 


31:53 — Best practices for developers new to the GitHub Copilot environment 


35:58 — How to measure ROI on agentic AI: DORA metrics, retention, and churn 


37:54 — Real-world agentic use cases where the ROI still holds 


40:57 — Code review automation: understanding the double billing reality

The 90-day patch cycle was never perfect. But it worked because the window between a vulnerability being disclosed and an attacker building a working exploit was sufficient time for most organizations to respond. Frontier AI models like Anthropic's Mythos have closed that window. Now we're talking hours. Possibly minutes. 

That's not a headline. That's the operating reality two of Insight's CISOs are already building security programs around. 

In this episode, Jeremy Nelson explains how frontier AI models don't just find vulnerabilities; they build the exploit in the same process. He walks through the risk tolerance slider bar, a framework for understanding why the old patch cycle used to make sense but no longer does. And he shares the five questions every security leader should be able to answer right now (the same ones your CEO and board are about to start asking). 

Jason Rader picks up where Jeremy leaves off. He's been having CISO-to-CISO conversations with clients who want to know how Insight is handling this. Why? Not because everything is going smoothly, but because the honest account of what's working and what's not is more useful than a polished pitch. His answer to mid-market organizations facing the same threat surface as Fortune 500 companies — with fractions of the resources — is direct: Leave behind security operations that aren't core to what you do and get visibility over everything else. 

Jeremy and Jason are operating with Insight as client zero, meaning Insight's own environment is where tools and programs are being proven before they're offered to anyone else. 

Get Insight's Managed Exposure Defense because it's the same program built and battle-tested on Insight's own infrastructure — and it's available to organizations from one seat up: https://www.insight.com/en_US/what-we-do/expertise/managed-exposure-defense.html 

Also worth your time: The CISO Mythos Readiness Checklist: https://www.insight.com/en_US/content-and-resources/guide/the-ciso-mythos-readiness-checklist.html 5 Questions Every CISO Will Be Asked About Mythos: https://www.insight.com/en_US/content-and-resources/blog/5-questions-every-ciso-will-be-asked-about-mythos-and-how-to-answer-them.html Don't Hit Later on That Update Notification: https://www.insight.com/en_US/content-and-resources/article/dont-hit-later-on-that-update-notification.html 

Subscribe to Insight On for more conversations with the practitioners shaping enterprise technology strategy. 

#FrontierAI #Cybersecurity #CISO  

Chapters 

 • 00:00 — Welcome and episode overview  

• 00:37 — Why the weaponization window has collapsed 

 • 03:54 — How frontier AI models discover and create exploits  

• 08:18 — What to do when the threat feels overwhelming 

 • 10:23 — Why your CMDB is your biggest security liability 

 • 12:20 — Treating patch updates as security incidents 

 • 14:22 — How AI fits into vulnerability management and defense 

 • 16:48 — Inside Insight's managed exposure defense program  

• 18:42 — Extending protection to clients of all sizes  

• 20:14 — The risk tolerance slider — patching vs. business disruption 

 • 24:16 — Why patch maintenance is now a permanent priority 

 • 26:19 — Five questions every security leader must answer now 

 • 38:12 — Jason Rader on being client zero and managed security services 

 • 40:29 — What every business leader needs to know about Mythos 

 • 44:13 — How frontier models operate inside your network 

 • 47:04 — How Insight's security priorities have shifted 

 • 48:07 — Mid-market security and the managed SOC advantage 

 • 55:25 — Phishing risk when vulnerabilities go public 

 • 57:43 — Zero days, patch ecosystems, and the hacker economy 

 • 01:02:56 — The one thing every security leader must do tomorrow

WestJet crews used to carry their entire jobs in heavy tote bags — paper manuals, paper shift trades, policy updates inserted page by page. It was tedious, inefficient, and everyone knew it had to change. 

This is the story of what happened when WestJet went all-in on Apple devices across every cockpit and cabin. It's a story about a former flight attendant from a 500-person town in Newfoundland who became the person responsible for putting an iPad in every crew member's hand.  

A story about what goes wrong when your device vendor is in another country and keeps shipping the wrong thing.  

And a story about what "ready for takeoff" actually means when thousands of iPads are expected to work seamlessly before hundreds of daily flights. 

This episode is told in a narrative format — cinematic, human, and unlike anything we've done before on Insight On. 

Get Insight Flex for Devices, an AI-enabled Device as a Service (Daas) model because it can lower your total cost of ownership, automate lifecycle workflows, and enhance your employee experience: https://www.insight.com/en_US/what-we-do/expertise/digital-workplace/device-as-a-service.html 

Learn more about effortless Apple device management for the modern workforce: https://www.insight.com/en_US/content-and-resources/brands/apple/effortless-device-management-for-the-modern-workforce.html 

#DeviceLifecycleManagement #Apple #DeviceAsAService #WestJet #InsightOn

Enterprise AI deployment is moving fast — and most organizations feel like they're already behind. This episode breaks down what separates companies making real AI progress from those stuck in strategy mode, and what you should demand from any partner before you commit. 

You'll find out what leaders across industries are asking about agentic AI, why agent sprawl is the new shadow IT, and the three criteria that matter most when evaluating an AI partner. Ferhan Zaki, SVP of Sales for the Google Cloud Solution Line at Insight, shares what he's seeing on the ground every single week in client conversations across North America. 

Get Insight's data and AI solutions because you'll see exactly how enterprises are moving from proof of concept to production without getting stuck in pilot purgatory: https://www.insight.com/en_US/what-we-do/expertise/data-and-ai.html 

Subscribe to Insight On for new episodes every week. 

#EnterpriseAI #AgenticAI #AIGovernance #GoogleCloud #InsightOn 
 

Chapters (5–12) 

00:00 — Welcome and intro 


01:25 — Common themes from enterprise AI client conversations 


03:01 — Why every organization feels months behind 


05:17 — How financial services is deploying agentic AI 


08:05 — Manufacturing gets real-time sentiment analysis with AI 



10:03 — Three layers of AI deployment in organizations 


12:12 — What separates fast movers from stuck organizations 


14:28 — How to evaluate your AI partner today 


16:49 — The agent sprawl governance challenge 


18:21 — Why data readiness is the question nobody asks 


19:16 — Top reasons organizations aren't getting AI outcomes 


23:14 — Why AI licenses become shelfware without adoption strategy 


25:01 — What every organization should do right now

AI training for creative teams fails when it ignores the real barriers: People don't know where to start, and they're afraid they'll automate themselves out of a job. Samuel Archibald, director of AI at the Sherlock Company, built an entire AI enablement program from scratch — gamified learning modules, multi-agent grading systems, prompt auditing tools — and hit a 93% completion rate without making any of it mandatory. 

In this conversation, Samuel breaks down how he moved a creative agency from AI resistance to AI fluency. You'll hear how he separated LLM capabilities from image and audio models to reduce fear, why non-coders started building Photoshop plugins before the engineering team caught up, and the simple automation rule that changed how his team thinks about repetitive work. 

See how the Sherlock Company is actively using AI to transform their creative workflow with Insight and Vertex AI (now Gemini Enterprise Agent Platform): https://www.insight.com/en_US/content-and-resources/case-studies/the-sherlock-company-automates-creative-content-at-scale-with-vertex-ai-and-sada-services.html 

Still facing AI hype? See how Insight is helping organizations build AI programs that stick — from training to production: https://www.insight.com/en_US/what-we-do/expertise/data-and-ai.html 

Subscribe to Insight On for new episodes every week. 

#AItraining #AIadoption #GenerativeAI #CreativeAgency #InsightOn 

Chapters (5–12) 

00:00 — Welcome and intro 


01:49 — What the Sherlock Company does for studios 


05:11 — Building a custom AI learning platform 


05:45 — Gamified quest mode and 93% completion 


07:23 — Overcoming fear and the empty chat box 



09:01 — Finding AI use cases for every role 


09:55 — Multi-agent grading and prompt auditing 


12:02 — Non-coders building apps and plugins 


14:06 — Keeping training current when tools change monthly 


16:28 — Blurring roles between creatives and engineers 


19:43 — Turning AI-resistant teammates into adopters 


22:50 — Reframing AI as "I can do this faster" 


23:09 — Why showing failures matters for adoption