Malcolm Bastien Enterprise Agile Coach
search

Conversations All the Way Down

Agile practices get their value from the conversations they produce. Standards and automation remove those conversations, decreasing it's resilience and agility.

Malcolm Bastien
Malcolm Bastien
• 8 min read
edit_note Draft Post — This post is a draft and is unlisted.

What makes agile practices useful are the conversations they produce. A story gives people something to argue about. A retro creates room for what nobody said all week. A standup reveals what changed instead of reciting a status field.

format_quote Quote

“Lead with conversation, and trail with documentation.” — Lane Halley

In agile, conversation moves work forward. The framework, roles, and tools can change; the quality of the conversations still determines how well the system responds.

When those conversations thin out the system loses its early warning signals about delays, workarounds, and dependencies. Those issues only surface later as missed dates and blame aimed at the people closest to the work.

tips_and_updates Tip

Every agile practice is a token to have a conversation When people stop challenging stories, naming awkward truths, or reporting changes, the format serves compliance instead of discovery.

The Container, Not the Content

Ask what any practice actually produces.

A Kanban board makes the value stream visible. Its value is a shared object the team can examine: where the pile is, why an item has stalled, and what to do. Without conversation, it is a picture nobody acts on.

Planning gives rough size and capacity as a starting point. Numbers miss dependencies, constraints, and private uncertainty. Disagreement is the useful output; numbers alone record intent, not shared understanding.

A retrospective gets credit for its format. The honesty does the work. A tidy list of action items and no uncomfortable sentence means the team spent an hour for little value.

Swarming, pairing, and demos make useful conversations routine. Swarming gets people unstuck. Pairing joins knowledge to a keyboard. Demos let customers contradict expectations.

A team I worked with had a healthy-looking board for months. A hallway conversation revealed a test environment shared by three teams. It wasn’t on the board because the board tracked their work; the environment belonged to someone else.

The Problem Hiding in “How Do I Write Better Stories?”

The question I hear most often arrives with a draft: how do I write better stories?

A user story is easy to judge: A clear user and goal, a succinct description of what needs to be built, and a enough acceptance criteria and scenarios that describe the intended system behaviors and edge cases. But too many people want to judge the user story as a text document, and not enough people are interested in how it was created and where it came from. If a single person wrote the user story alone and wants to show-and-tell their story to the rest of their team, then the whole process is wrong.

The user story as a card is a measurable, comparable artifact people can judge. What’s more important is the process that start from problem identification and goes until to team alignment. It’s a bundle of practices that are all fueled by conversation.

A collaborator might ask the question that takes three seconds, or say, “That’s not what I’d expect,” and stop a wrong assumption. Written alone, it ships and gets discovered downstream, where the correction costs a sprint instead of a sentence.

Ask who is writing the story. Pair up. Read it to the person who will use it and watch where their face changes. The template cannot supply the conversation.

format_quote Quote

“People over process. Why? Because systems can’t fix problems with people, but people can fix problems with systems.” — Steve (Big Red) Bishop

Problems Surface in Conversation, Not on Dashboards

Lean calls the real place where work happens genba. Toyota has used the idea far longer than agile has. Leaders go there, watch the work long enough to understand it, and ask questions about the process and the people doing it.

An afternoon at the genba shows what a week of status reporting will not: a code review that takes six days and gets read in five minutes; a two-hour database change waiting three weeks for another team; one person who understands the nightly job and is booked through the quarter.

None of it appears in the burndown. Queues, handoffs, approvals, and knowledge held in someone’s head surround the work.

The same visits build resilience. Teams know who to call and which assumptions their neighbours depend on. Leads hear about single points of failure while they are still complaints.

Improving service delivery depends on keeping this exchange going. Go where the work is, ask, listen, then say what changed. People keep talking when doing so changes something before.

The Same Promise, the Same Trap

Two responses dominate improvement plans. One is a standard: an explicit process, defined once, that everyone follows. The other is automation, increasingly AI-driven.

Both can be right. Repetitive work with stable demand and few exceptions benefits from a standard. A model can do work it performs consistently. The pitch breaks when nobody has watched the variation: a process run four times, three workflow variants with a fourth coming, or intake split between email and tickets.

A standard written before the variation is mapped freezes a shape the work does not have. People route around it, and the exceptions return to conversations the standard meant to replace. The process hides them.

format_quote Quote

“Automating an existing process without redesigning it speeds up dysfunction, not performance.”

AI and standards promise to replace useful friction with procedural efficiency. But friction carries information. “Hang on, that doesn’t match what we talked about yesterday” is the system correcting itself. Remove it and the system reaches the wrong answer faster, unnoticed until the customer sees it.

What You Lose When People Stop Talking

The first loss is the warning system. Conversations name delay while it is still small enough to fix. Once delay becomes a metric, the dashboard arrives without context.

The second is the ability to change. A standard stores a decision outside the room. Changing it means finding the owner, making the case, and waiting. Teams lose that capacity one process at a time.

The third is practical knowledge. When everyone follows the designed path, it becomes the only route they know. Shortcuts and workarounds stop being passed on. The organization depends on the design being right, with fewer people who remember its failures.

Any practice that removes conversation needs something measurable in its place. Follow the process. Keep utilization up. Reduce manual hours. None describes what the customer receives, so a team can meet them and still deliver the wrong thing slowly.

Three Questions Before You Standardize or Automate

Before replacing a conversation with a process or a model, ask:

  1. What conversation does this replace, and who is having it now? If the team argues about this every planning session, the argument is building their understanding. Your replacement has to build that understanding somewhere else.

  2. What is the current variation? Low variation and few exceptions can support a standard. High variation means encoding something you do not understand; the exceptions will fight you.

  3. What will tell you when the work no longer fits? A deviation should start an investigation, not a compliance report.

Standards and AI are not the problem. The order matters, and so does what you leave open. Fix what you understand. Automate work where variation is gone. Keep judgment calls open.

Three Things Worth Taking Away

  1. The practice is a container, not the outcome. User stories, planning, standups, and reviews get their value from the conversation they trigger. Once the conversation stops, keeping the format running is admin.

  2. Apply anything that removes conversation where variation is already low. Standards and automation work on work you understand. Apply them early and they hide the exception path you needed to see.

  3. Watch for practices that stop people from talking. When a team gets quieter, the system loses its early warning, its ability to change, and its grip on what it is actually for. That is harder to reverse than slow delivery.

Conversations are cheap, and they are the only place the system tells you what it is doing. Standards and automation can carry the parts that repeat. They cannot carry the part that is still being learned.

Malcolm Bastien

Malcolm Bastien

Enterprise Agile Coach

Enterprise Agile Coach helping leaders in Fintech, Banking, and HealthTech build fast flow through systems thinking and aligned autonomy, empowering teams to deliver high-performance results at scale.