The Art of Balancing Speed and Governance in TPM: A Journey Through Process Landmines

Join me on a humorous yet insightful journey through the maze of TPM processes, from incident management to SLOs, where we navigate the landmines of bureaucracy and cargo cults while seeking the holy grail of agile and adaptive frameworks in the age of AI.

Navigating Challenges In Technical Program Management

Picture this: a bustling tech office, the sound of keyboards clattering, the aroma of artisanal coffee wafting through the air, and a group of TPMs huddled around a table, deep in discussion. It could be a scene from any tech startup’s daily grind, but today we’re all here for a single purpose: to dissect our latest incident and ensure we don’t repeat the same mistakes. Welcome to the colorful, sometimes chaotic world of Technical Program Management (TPM), where processes are both our compass and our potential quicksand.

Now, let’s be frank. There’s a lot of hype surrounding AI these days, and as a skeptical TPM, I often find myself in the middle of a hype cycle that could rival a rollercoaster. Generative models, machine learning, and AI efficiencies swirl around us like confetti after a New Year's Eve party. But beneath this shiny surface, the underlying processes that allow these technologies to flourish—or flop—are what keep me up at night.

Incident Management: The Blame Game

Let’s start with incident management. It’s a necessary evil, isn’t it? The moment an incident occurs, it's as if we're all suddenly starring in a detective film—spotlighting the culprits, examining the evidence, and, of course, pointing fingers (not the most productive approach). But here’s the kicker: we’re all adults in this room, and the only thing worse than a failed incident is the blame that often follows. Enter the blameless postmortem.

Blameless postmortems are like the therapy sessions of the tech world. They encourage open dialogue about what went wrong without fear of retribution. Picture a circle of TPMs, each with their favorite mug, sharing stories about how the AI model went rogue because someone forgot to update the training data. Instead of a blame game, we transform into a collaborative think tank seeking to understand and learn.

SLO/SLA Hygiene: The Unsung Heroes

Next up, we have SLOs (Service Level Objectives) and SLAs (Service Level Agreements). Think of these as the health metrics of your tech ecosystem. Maintaining SLO/SLA hygiene is akin to regular check-ups at the doctor’s office; you don’t wait for a crisis to roll around to monitor your vitals. Instead, you actively track performance and set expectations for your teams.

In the world of AI, where models can be fickle and unpredictable, having well-defined SLOs isn’t just a nice-to-have; it’s a necessity. They provide a roadmap, guiding teams to build and refine solutions that meet customer needs without compromising on quality. And let’s face it, no one wants to be on the receiving end of an angry customer email because a model’s performance dipped below acceptable levels.

Release Trains and Quality Gates: The Expressway to Success

As we delve deeper into processes, let’s talk about release trains and quality gates. I like to think of release trains as the expressways of our deployment process. They keep everything on track and ensure that we don’t veer off course into the chaos of uncoordinated releases. However, let’s not confuse expressways with stoplights; quality gates are the necessary checkpoints ensuring we don’t end up with a half-baked product hurtling towards customers.

In the fast-paced environment of AI development, where iterations happen at lightning speed, having these processes in place ensures that while we’re agile, we’re also maintaining quality. Imagine racing down the expressway only to hit a pothole because the QA team didn’t have a chance to inspect the ride. Ouch.

Design/PRD Review Rituals: The Craft of Collaboration

Ah, the design and Product Requirement Document (PRD) reviews—those sacred moments when creativity meets structure. It’s like a potluck dinner where everyone brings their dish, hoping it’ll be the star of the show. The challenge? Ensuring that no one’s dish is a mystery meat casserole, and that everyone’s contributions are harmoniously blending into a delightful feast.

Here’s where healthy patterns emerge. Instead of rigid, bureaucratic reviews that stifle innovation, we can adopt a more adaptive approach. A lightweight review ritual can be more effective, allowing teams to iterate quickly and refine ideas without the heavy-handedness of a formalized process that feels more like a chore than a collaboration.

Balancing Governance with Speed: The Tightrope Walk

Finally, let’s discuss the age-old balancing act of governance and speed. It’s like trying to juggle flaming torches while riding a unicycle—impressive but highly risky. In the realm of AI, where speed is often glorified, we must not forget the importance of governance. Yes, we want to push boundaries and innovate, but we also need to ensure that our processes don’t devolve into bureaucracy or cargo cult practices, where we blindly mimic behaviors without understanding their purpose.

Instead, we can embrace data-informed decision-making and adapt our practices as needed.

Striking Balance For Sustainable Success

Teams that strike this balance can operate effectively—moving quickly while ensuring that the necessary governance frameworks are in place to support sustainable growth.

Reflection: The Journey Continues

As I sit here reflecting on this journey through the processes of TPM, I’m reminded that while the landscape may evolve, the underlying principles remain the same: collaboration, learning from our mistakes, and adapting to change. Amidst the noise of AI hype, it’s the humble yet powerful processes that keep us grounded. So, the next time you find yourself in a meeting, dissecting an incident or reviewing a PRD, remember: it’s not just about the processes. It’s about the people behind them—and the stories we create together along the way.