HS134: Dodging the AI Iceberg: Midcourse Corrections
John Attil Johnson and John Burke of Numerides discuss when and how to make mid-course corrections to IT and AI strategies. They cover triggers for strategic reassessment, the value of wargaming and scenario planning, and how business imperatives should drive technology decisions. The episode emphasizes structured, recurring strategy reviews rather than reactive responses to crises.
Summary
In this episode of Heavy Strategy, co-hosts John Attil Johnson and John Burke of Numerides explore the topic of mid-course strategy corrections, with a particular focus on AI and IT strategy. They open by identifying several non-emergency categories that should prompt strategic reassessment: things that were unclear at the outset but are now visible, developments that arrived faster than anticipated, projects that are over budget or under-delivering, and unexpected surprises—both positive and negative.
The hosts use AI deployment in help desks as a concrete example, referencing research showing roughly 20% productivity gains for entry-level help desk work but weak to negative correlation for higher-level staff. They argue this kind of differential outcome should prompt organizations to revisit where and how they are applying AI, rather than assuming uniform benefits across all use cases.
A significant portion of the discussion focuses on the importance of after-action reviews and the 'five whys' methodology—drilling down into root causes rather than simply noting that something didn't work. They recommend at least quarterly strategic check-ins at the leadership level to assess whether strategy is on track and whether external conditions have shifted. They also connect IT strategy back to business imperatives, noting that if the core corporate objective changes—say, from margin improvement to revenue growth—the technology strategy must be re-examined from the ground up.
The hosts advocate strongly for wargaming as a proactive tool, arguing that organizations should red-team their strategies to identify contingencies in advance. They cite the example of clients who had pre-built pandemic response plans and were able to activate them smoothly when COVID-19 hit. They extend this logic to current geopolitical and supply chain risks, such as helium shortages affecting MRI centers, TSMC disruption affecting hardware procurement, and rare earth mineral dependencies on China.
They also discuss how technology advances can render entire business processes obsolete, warning against over-investing in technically enabling processes that new technology will soon eliminate. This is illustrated through the anecdote of Cooper Union's original architects pre-building an elevator shaft in the 1800s, anticipating future need. The hosts close by noting that course corrections should also prompt a review of whether the IT organizational structure—particularly its degree of siloing—is still fit for purpose.
About this episode
By now most organizations have AI strategies (among their other tech strategies). But how do you know when it’s time to make a midcourse correction? Better still: How can you predict when, and what kind of corrections you might need? John and Johna discuss, and tell the story of how a university prepared for technology<a class="excerpt-read-more" href="https://packetpushers.net/podcasts/heavy-strategy/hs134-dodging-the-ai-iceberg-midcourse-corrections/" title="ReadHS134: Dodging the AI Iceberg: Midcourse Corrections">... Read more »</a><img height="1" src="https://feeds.packetpushers.net/link/22503/17352804.gif" width="1" />
Key Insights
- The hosts argue that research shows AI produces roughly 20% productivity gains for entry-level help desk work but has a weak to negative correlation with productivity at second and third support tiers, suggesting AI benefits are not uniform across organizational levels.
- Johnson contends that many IT projects are launched without a clear definition of what they are supposed to accomplish, making it difficult to assess whether a course correction is needed.
- Burke argues that wargaming IT and business strategies—where someone red-teams the plan with simulated external events—can convert reactive course corrections into the activation of pre-planned contingencies, reducing the cost and chaos of mid-stream pivots.
- The hosts assert that if senior executives are repeatedly asking about a technology or risk that is not currently on the IT strategy roadmap, that is a reliable signal that a strategic reassessment is overdue.
- Johnson argues that organizations should explicitly map critical resource dependencies—such as rare earth minerals, helium, or TSMC silicon—and model how disruptions to those resources would force changes to IT and business strategy.
- Burke contends that when designing new IT systems or infrastructure, teams should proactively identify the conditions under which that investment will become obsolete, so transitions can be planned in advance rather than reverse-engineered under pressure.
- The hosts argue that IT strategy must be traced back to corporate-level business imperatives, and that a change in top-level objectives—such as shifting from margin improvement to revenue growth—should trigger a full re-evaluation of the technology strategy stack.
- Johnson suggests that when making a course correction, organizations should also evaluate whether their IT organizational structure, particularly the degree of siloing versus horizontal integration, is still suited to the new strategic direction.
Topics
Transcript
Hello, I'm John Attil Johnson, CEO of Numerides, and I'm here with my co-host, John Burke, CTO of Numerides. And you're listening to Heavy Strategy, the show that tries to ask the right questions, not give the right answers. This episode is sponsored by Meter, which is the company building networks from the ground up. Meter delivers a complete networking stack, wired, wireless, and cellular in one integrated solution. With METER, you get fast, secure, scalable connectivity without the burden of managing multiple providers or tools. METER scales from branch offices, warehouses, and large campuses to data centers. And its integrated network stack is designed to give you deep control and visibility. METER covers everything, from first site survey to…
Full transcript available for MurmurCast members
Sign Up to AccessMore from The Everything Feed - All Packet Pushers Pods
D2DO310: Developing Efficient AI Workflows
The podcast discusses the evolution and best practices of using AI coding assistants within DevOps workflows, emphasizing token efficiency, agent memory management using Obsidian, and the importance of sandboxing for security. Tyler Lynch shares insights from his experience with AI agents in coding and automation at IBM.
HW085: Designing Wi-Fi for High-Density Events
The podcast discusses strategies for designing Wi-Fi networks for high-density events, emphasizing the importance of thorough planning, questioning, and on-site validation. Key tools and techniques for optimizing performance under these conditions are also highlighted.
PP121: How CYBR.SEC.CON Builds Community for Learning and Professional Development
The episode features a conversation about CyberSecCon, a community-driven security conference in Houston, and explores the origins and growth of both the conference and the Packet Protector podcast. Key insights include how both platforms aim to foster community and collaboration within the cybersecurity and networking fields.
NB586: AI Angst Abounds; Arista Revs Up Q2 Revenue
The episode covers significant cybersecurity updates, including Cisco's patch for its SD-WAN software and the formation of the Open Secure AI Alliance, as well as quarterly financial results from several tech companies. Notable acquisitions and the potential impacts of regulation on the telecom industry are also discussed, alongside Starlink's plans to establish a terrestrial mobile network.
HN837: Agentic AI to Reduce MTTR
Edouard Douleroux discusses practical agentic AI implementation for reducing Mean Time To Recovery (MTTR) in network operations. He describes a deterministic system combining agents, fine-tuned language models, MCP servers, digital twins, and a YAML-based single source of truth to automate incident detection, diagnosis, and resolution testing.