Tag: Technical Debt

Posted on by

TWIET Episode 43

Welcome back to This Week in Enterprise Technology, Hyoun Park and Charles Araujo analyze the latest enterprise technology announcements and how they will affect your business and your bosses’ expectations.

Join TWIET as we guide CIOs and technical managers through the strategic ramifications behind the vendor hype, product innovation, and the avalanches of money going in and out of enterprise tech. As always, this podcast is available in audio, video, and broken up into sections for your benefit.

Audio Podcast: https://www.buzzsprout.com/2319034/episodes/16252199

This Week in Enterprise Technology, Hyoun Park and Charles Araujo critically assess last week’s biggest tech news:

  1. AWS Enhances Amazon Connect with Generative AI Tools
  2. AWS Takes on AI Hallucination Challenges
  3. AWS Bedrock Adds Multi-Agent Orchestration and Model Routing
  4. AWS Centralizes AI Efforts with SageMaker
  5. Casey Newton Examines AI Skepticism’s Comforts
  6. Emergence AI Coordinates Multi-Vendor Agents
  7. Exa Redefines Generative Search Experiences
  8. MLCommons Benchmarks LLM Output Risks
  9. South Korea’s Unrest Threatens Global Memory Supply
  10. Werner Vogels on Managing “Simplexity”
  11. Broadcom Adjusts to Minimize VMware Migration Risks

AWS Upgrades Amazon Connect with New Generative AI Features


Amazon Connect has been a successful cloud contact center product and contact center has been one of the clearest areas for AI to provide productivity benefits and increase potential revenue transactions,  AWS re:invent was an opportunity to announce the latest generative AI advancements within Connect. Charles and Hyoun discuss the opportunities for contact centers to adopt AI.

Source:
Maria Deutscher from Silicon Angle: https://siliconangle.com/2024/12/01/aws-upgrades-amazon-connect-new-generative-ai-features/ 

AWS Tackles AI Hallucinations

AWS launches Automated Reasoning checks to cross reference outputs with known facts and enterprise data. Although this is not as novel as AWS was stating, it is a valuable step forward. Hyoun and Charles debate the utility of this Automated Reasoning checks and whether AI hallucinations really matter or are just a sign of AI immaturity and inexperience. 

Source:

Kyle Wiggers on TechCrunch: https://techcrunch.com/2024/12/03/aws-new-service-tackles-ai-hallucinations/ 

AWS Bedrock Updates: Multi-Agent Collaboration, Model Routing

AWS announced interesting AI management updates for Amazon Bedrock. Both multi-agent management and prompt routing across models will be useful for enterprises seeking to expand the utility and cost structure of AI. Charles and Hyoun wonder if this agent management will cover the bill given the wide variety of agents that are starting to appear in the enterprise. . 

Source:

AWS: https://aws.amazon.com/blogs/aws/introducing-multi-agent-collaboration-capability-for-amazon-bedrock/ 

AWS Wraps Everything Together Under Sagemaker

AWS create a new umbrella brand that includes data studio, data lake, analytics, and data management. Hyoun and Charles argue about whether Sagemaker, best known as a data science tool, was the right umbrella brand for these data efforts.

Source:

AWS: https://aws.amazon.com/blogs/aws/introducing-the-next-generation-of-amazon-sagemaker-the-center-for-all-your-data-analytics-and-ai/ 

Casey Newton Examines AI Skepticism’s Comforts

One of TWIET’s favorite journalists, Casey Newton, weigh in on the false comfort of AI skepticism. Newton argues that the potential harm of AI is being underestimated by those who simply think that AI is full of lies or incompetent.  Charles and Hyoun discuss a more realistic path for IT departments to consider as they deploy AI.

Source:

Casey Newton on Platformer: https://www.platformer.news/ai-skeptics-gary-marcus-curve-conference/ 

Emergence AI Coordinates Multi-Vendor Agents

Start up Emergence AI announced its autonomous multi-agent AI orchestrator. At a time on every enterprise platform seems to be coming out with its own set of agents, Hyoun and Charles think it is about time for a third-party agent orchestration solution to hit the market and get some traction.

Source

Carl Franzen on VentureBeat: https://venturebeat.com/ai/emergences-ai-orchestrator-launches-to-do-what-big-tech-offerings-cant-play-well-with-others/ 

Exa Redefines Generative Search Experiences

The MIT Technology Review covered a startup named Exa taking a novel approach to Gen AI based web searches with the goal of using the web like a database. Charles and Hyoun discuss the scale and results for this approach.

Source:

Will Douglas Heaven on MIT Technology Review: https://www.technologyreview.com/2024/12/03/1107726/the-startup-trying-to-turn-the-web-into-a-database/ 

MLCommons Benchmarks LLM Output Risks

MLCommons has released its AIluminate 1.0 benchmarks to describe several categories of harm including sex crimes, violence, and defamation risks. Hyoun and Charles discuss past challenges regarding model benchmarking and risks. 

Source:

MLCommons: https://ailuminate.mlcommons.org/benchmarks/ 

South Korea’s Unrest Threatens Global Memory Supply

South Korea saw government unrest in an attempted military coup last week. Although we are not expert political scientists, international supply chains do affect our ability to source IT. We discussed the ramifications of South Korea earning 60% of the global memory, check market and considerations for the CIO in looking at geopolitical strife.

Source:

Prasanth Aby Thomas on CIO.com: https://www.cio.com/article/3617847/south-koreas-political-unrest-threatens-the-stability-of-global-tech-supply-chains.html 

Werner Vogels On Managing “Simplexity”

At Amazon re:invent, Amazon CTO pointed out both that complexity is inevitable and that there are two types of complexity that are important for technical audiences to consider, including a new concept of “simplexity”.. Hyoun is reminded of the Nassim Taleb concept of antifragility while Charles digs deeper into the strategic issues of technical debt. 

Source:

Tom Krazit on Runtime News: https://www.runtime.news/werner-vogels-complexity-is-inevitable/ 

Broadcom Adjusts to Minimize VMware Migration Risks

Broadcom has had to call back from its initial plans of making its top 2000 customers all direct and has handed much of that business back to its channels. With help from The  Register and Canalys, Hyoun and Charles discuss repercussions for tech sourcing. 

Source:

Simon Starwood on The Register: https://www.theregister.com/2024/12/05/vmware_user_migration_plans/ 

Posted on by

Why Companies Will Be Tempted to Repeat the Southwest Airlines Debacle in 2023

The biggest US-based travel story at the end of 2022 was the absolute collapse of Southwest Airlines. The United States was hit by a sudden cold snap just as Christmas approached, leading to a massive travel delay across almost all travel modes including airlines, trains, and road-based transit. However, after a couple of days, most US-domestic airlines seemed to have recovered with the exception of Southwest, which suddenly and unexpectedly canceled nearly all of its flights in the last week of 2022, just as people were traveling from or to locations for Christmas, Hanukkah, New Year’s Eve, and other holidays. The timing was horrible and inexplicable. And with little to no official explanation, travelers stranded across the country could only guess whether this was due to an unannounced strike. Were there problems with Southwest’s airplane fleet? Were there problems with a specific airport?

It turns out that the problem was with Southwest’s internal scheduling tool, an in-house software application built in the 1990s and held together over the years as Southwest roughly doubled in size across passengers, planes, trips, employees, and number of destinations supported. This complexity ended up being especially challenging because Southwest’s model as a regional airline meant that it did not use a central hub as most other large airlines in the United States use. Rather, each plane flies from point to point leading to a combination of possibilities that grew exponentially rather than linearly. Although Southwest does not fly every plane from each location to every other location, the complexity of operations from roughly 45 locations in the late 1990s to roughly 100 domestic locations today is not a doubling of complexity but more along the lines of N*(N-1)/2, as long-time analytic advisor Neil Raden pointed out. This means the complexity increase is more akin to (45*44/2) = 990 vs. (100*99/2) = 4950. This level of complexity is multiplied by the challenges of organizing the thousands of pilots and flight attendants traveling from point to point every day.

The orders of magnitude in complexity associated with this scheduling system had already been strained in previous years but met a critical breaking point at the end of 2022 due to a lack of investment and modernization. This failure is a textbook example of the concept of “technical debt.”

Technical debt is often described as a concept that is difficult to articulate for a business audience, but the concept is actually very straightforward from a business perspective. Just as with financial debt, which must be paid back with interest or risk a default that threatens business assets, technical debt is an act of borrowing against the future. Like financial debt, technical debt either requires future investment (the “interest”) to fix the technology over time or to accept that the technology will fail (“default”) and lead to breaking down any processes dependent on the technology.

The lessons from this breakdown are straightforward but are potentially challenging to follow in 2023, a year where companies will be tempted to cut costs by any means possible.

Ensure that executive stakeholders are clear both on the concept of technical debt and the labor associated with current technical debt. It may not be possible to put an exact dollar amount on the technical debt that currently exists in the organization, but it should be possible to provide some guidance on the current labor and resources assigned to managing outdated technology as well as the potential points of failure associated with, say, being unable to find a FORTRAN developer quickly or the use of applications no longer supported by a vendor or by in-house developers.

Document every technology associated with each mission-critical process. With the cliché that “every company is a technology company” having been fully realized in today’s web, mobile, and automated world, IT’s job is to provide proactive guidance on the hardware, software, and skills that must either be supported or upgraded. The business value propositions of IT asset and service management are unlocked when assets are specifically aligned to business dependencies, projects, and processes.

Identify technologies where business growth lead to exponential technology demand. Southwest’s scheduling system needed to grow exponentially and eventually failed based on its legacy design. Look at the mathematics associated with key processes to see if growth is logarithmic, linear, exponential, or unpredictable. Simply assuming that a process grows linearly with revenue, employee growth, or business traffic can be a job-ending mistake.

Ensure that legacy technologies have the capacity to support forecasted business complexity or business growth. Any time technology growth needs to expand faster than overall IT spend or overall operational spend, it should serve as a warning sign to either change the technological approach or to invest in the necessary capacity.

We face a challenging year as inflation, foreign currency challenges, geopolitical issues, and supply chain bottlenecks still threaten the spectre of recession. But as executives seek to cut costs, Southwest serves as a reminder that businesses must still futureproof their technology approaches, evaluate the scalability of their processes, and invest in service delivery commensurate with their brand promise or risk lasting revenue and market capitalization losses.