How to fix the hidden problem of team handoffs
Team handoffs fail because we can't see them fail. Coupled with good process observability, AI can help you see what to fix. Read: "Everything Starts Out Looking Like a Toy" #261

Hi, I’m Greg 👋! I write weekly product essays, including system “handshakes”, the expectations for workflow, and the jobs to be done for data. What is Data Operations? was the first post in the series.
This week’s toy: an open source repo to build a robot hand. We’re going to need standard templates for physical things just like we do for digital items, and this is a good place to start.
Edition 261 of this newsletter is here - it’s July 28, 2025.
Thanks for reading! Let me know if there’s a topic you’d like me to cover.
The Big Idea
A short long-form essay about data things
⚙️ How to fix the hidden problem of team handoffs
It's the problem that everyone knows. A customer request gets lost between sales and support. A bug report sits in the queue because engineering never got the full context. A project deadline slips because the handoff between design and development missed a critical detail.
These aren't just annoying moments. They're process failures that create rework, missed deadlines, and frustrated teams. Work transitions between teams are our biggest bottleneck, and most of the time, we don't even know they're failing until we're already dealing with the consequences.
The invisible intelligence gap
We have better visibility into our code than we do into our processes. Engineering teams track every commit, every deployment, every bug fix. When a build breaks, they might not know exactly what went wrong, but they get a signal that they need to investigate before they can move on to other tasks.
We're missing an analog to the handoffs that happen between teams and describe the work of getting things done. What if we designed our handoffs to be measurable and observable from day one, just like we create continuous delivery processes for code?
In a perfect world, we'd be able to watch, learn, and catch errors before they become problems.
It turns out that AI is pretty good at identifying true/false conditions from ambiguous input and many handoff conditions look just like that. Imagine knowing that handoffs between engineering and ops are failing 30% of the time, and having AI suggest specific improvements based on what works.
This might mean something as simple as tracking an engineering request for wait time, or it might mean reengineering a bug reporting process from end to end. It's about treating our operations like we treat our code: measurable, observable, and continuously improvable.
The promise of incremental improvement
The future of team handshakes involves giving teams the intelligence they need to catch errors before they become problems, with AI as a silent partner that helps prevent the small failures that add up to big headaches.
The fundamental problem with team handoffs isn't that they fail. The issue is that we don't know about problems until after they become big problems.
Here are some typical causes for our process blindness:
Context gets lost in transitions. When work moves from one team to another, critical information often gets filtered out. The receiving team gets a sanitized version that misses the nuance, the edge cases, or the historical context that could prevent problems.
We have no baseline for what "good" looks like. Without observability built into our processes, we can't measure handoff effectiveness. We don't know if a 2-day handoff is fast or slow. We can't tell if the information being passed is complete or missing key details.
Teams work in silos, missing patterns. When handoffs fail, each team blames the other. Sales thinks support dropped the ball. Engineering thinks QA is being unreasonable. Without data, we're stuck in a cycle of finger-pointing instead of process improvement.
Human memory is fallible. We forget what we learned from last month's handoff failure. We don't remember which combinations of factors led to success. Every handoff becomes a fresh experiment instead of building on past learnings.
We treat handoff failures as inevitable rather than preventable.
Some parallels from engineering
Good development teams build observability into their processes to know when they hit an error. The goal is to see exactly when something breaks and why.
When a deployment fails, engineering teams investigate. They look at the logs, trace the execution path, identify the root cause, and fix the underlying issue. They treat every failure as a learning opportunity.
But when a process handoff fails, what do we often do?
We blame the other team. We assume it was a one-off problem. We move on without learning anything. We don't have the data to do anything else.
What if we treated our operations like engineering teams treat their code?
What if we built observability into our handoffs from the start, so we could see exactly when and why they fail? That sounds like a better recipe for compounding success.
Building observability into team handoffs
The key insight is that we need to design our handoffs to be observable from day one. Instead of trying to add monitoring to existing broken processes, we need to build processes that are measurable and improvable from the start.
What we need to track
Expected vs actual sequence: What should happen vs what actually happened
Time delays and bottlenecks: Where work gets stuck and why
Context preservation: How much information makes it through the handoff
Success/failure patterns: What combinations of factors lead to good outcomes
How to implement it
Event logging: Capture every handoff moment with structured data
Real-time monitoring: Watch handoffs as they happen, not after they fail
Pattern recognition: Use AI to identify what works and what doesn't
Continuous improvement: Feed insights back into process design
We want to create better processes that are transparent, measurable, and continuously improvable.
How can AI help with process observability?
Without the context of the process, AI will deliver a middling result when you ask it to analyze why something failed during a team handoff. AI systems (like humans) need to see a pattern to start to know what to look for.
But AI is particularly good once there is a template.
Context awareness: AI can understand the full context of what's being handed off, including historical patterns and team preferences.
Risk assessment: AI can identify potential handoff failures before they happen, based on patterns it's learned from historical data.
Process optimization: AI can suggest improvements based on what's worked in the past, helping teams iterate toward better handoffs.
Real-time feedback: AI can provide immediate insights during handoffs, helping teams catch issues before they become problems.
AI provides the insights gleaned from lots of data crunching so that teams do a better job with the next steps after they know what's going on.
How do we make this real?
Let's walk through how observability could transform a real ops scenario.
The Problem
Customer issues get escalated from frontline support to engineering, but critical context gets lost in the handoff. The engineering team gets a vague description like "customer can't log in" without the specific error message, browser version, or steps to reproduce. The issue bounces back and forth while the customer waits, and the support team has no visibility into what engineering actually needs.
Building Observability
The team starts logging every escalation handoff with structured data:
When the escalation happened and how long the customer had been waiting
What information was included in the initial handoff (error messages, browser details, reproduction steps)
How many back-and-forth cycles occurred before resolution
Whether the engineering team had to ask for additional information
Total time from escalation to resolution
AI Pattern Recognition
Once there is enough data to analyze, AI identifies clear patterns: escalations with specific error messages and browser details are resolved 4x faster than those without. Issues escalated during business hours are resolved 2x faster than after-hours escalations. Support teams that include customer impact details get faster responses from engineering.
The Team's Response
Instead of guessing what engineering needs, the support team now has data. They create a simple escalation template that requires error messages, browser details, and customer impact. They start tracking which support agents consistently provide complete information and which ones need coaching.
Measurable Outcome
Within a month, the average escalation resolution time drops from 8 hours to 3 hours. The number of back-and-forth cycles drops by 70%. Customer satisfaction scores improve because issues get resolved faster, and support agents feel more confident because they know what information to collect.
These numbers are imagined, but you can see exactly how you can make them real.
The team can see exactly what works and why, then build those patterns into their process.
Getting started … one step at a time
The good news is that you don't need to rebuild everything at once. Start with one critical handoff process and build observability into it from the start.
Pick a high-impact handoff: Choose a handoff that's causing the most problems or affecting the most people.
Design for observability: Build measurement and monitoring into the process from day one, not as an afterthought.
Start logging events: Capture the data you need to understand what's working and what's not.
Look for patterns: Use AI to identify what leads to successful handoffs and what doesn't.
Iterate and improve: Use the insights to make your processes better, then measure the improvement.
The key is to start small and build momentum. Every handoff you make observable is a step toward better team coordination.
What’s the takeaway? The teams that improve their handoff observability will have better processes, happier teams, and more reliable outcomes. They'll be the ones setting the standard for how teams work together in the age of AI.
Links for Reading and Sharing
These are links that caught my 👀
1/ How do you spend your work time? If someone asked, would you know what percentage of time is spent actually writing code (or doing deep work of another kind)? It’s pretty hard to aggregate unless you use a tool to count. You might learn some surprising things. (For me, when to avoid scheduling meetings because certain hours of the day are the most productive.)
2/ Can LLMs do accounting? - Not without some help. LLMs can’t do accounting on their own without making material mistakes but do a great job with a bit of help.
3/ Vibe coding creates new problems - If you’ve tried vibe coding or any AI-assisted task, the first problem you solve is amazing. The next one starts to feel like all other problems, because the planning matters just as much (or more) than the doing.
What to do next
Hit reply if you’ve got links to share, data stories, or want to say hello.
The next big thing always starts out being dismissed as a “toy.” - Chris Dixon