Skip to main content Start Finder Patterns Experiment Use it

AI Delivery Friction Finder

Find where delivery loses its shape.

AI does not fix the delivery system. It makes the delivery system easier to see: context gaps, late technical discovery, quality fog, flow drag, and broken learning loops.

Find the drag → Skip to the sprint experiment Remind me next week We'll email you the link now, plus one reminder next week. Set reminder

This is not an AI maturity model.

It is a short working tool for a team conversation: where does work lose its shape between intent and delivery?

Rate ten friction prompts. Get a primary friction pattern and score breakdown. Download your answers as a markdown handoff.

Two-minute diagnostic

Read the statement. Rate the friction. Add the evidence.

Each card asks about one way delivery loses shape. The score is only useful if the team can point to a real example, so keep the evidence close to the rating.

Shared context keeps getting rebuilt.

Where is your team on this? (1–5)

Not true for us Very true for us 12345 What makes you say that? Where did someone need background that should already have been in the work? Technical reality shows up after planning feels settled.

Where is your team on this? (1–5)

Rarely true Constantly true 12345 What makes you say that? What did the team learn only after someone opened the codebase or dependency map? The team can produce work faster than it can prove the work is right.

Where is your team on this? (1–5)

Clear evidence Mostly vibes 12345 What makes you say that? What evidence would convince a skeptical reviewer? AI increases local speed but makes system-level flow worse.

Where is your team on this? (1–5)

Flow feels healthy Flow is dragging 12345 What makes you say that? Where does work wait, pile up, expand, or return for rework? Delivery signals do not change how the system works next time.

Where is your team on this? (1–5)

Learning changes behavior Learning evaporates 12345 What makes you say that? What did the last sprint teach you that actually changed a working agreement? AI-generated artifacts are not grounded in a shared understanding.

Where is your team on this? (1–5)

Grounded Detached 12345 What makes you say that? When AI drafts something, who can tell whether it reflects real intent and constraints? Dependencies, ownership, and constraints are under-described.

Where is your team on this? (1–5)

Well understood Mostly unknown 12345 What makes you say that? What would make refinement more honest before the team commits? Acceptance criteria are hard to observe or test.

Where is your team on this? (1–5)

Testable Ambiguous 12345 What makes you say that? Can someone prove the behavior works without reading minds? Work is too large to inspect quickly.

Where is your team on this? (1–5)

Small and inspectable Too large 12345 What makes you say that? What is the smallest useful thing you could learn within three days? The team lacks a signal for whether AI improved delivery decisions.

Where is your team on this? (1–5)

Signal is clear No signal 12345 What makes you say that? What would tell you AI helped beyond “we moved faster”? 0 of 10 rated

Rate at least five prompts to get a useful result.

View result Reset Primary friction

Context Gap

Your result will appear here.

What AI is exposing:

Team conversation:

Sprint experiment:

Use AI to Human check Signal Download markdown copy Email me my results: Send results Use the sprint planner

The markdown file includes your ratings, notes, result, and freeform planner text.

The five patterns

Most teams do not have an AI problem first.

They have a delivery-system problem that AI makes louder. The point is not to admire the diagnosis. The point is to pick the next constraint and work it.

01 · Context

People are rebuilding the same background.

Discovery, product intent, constraints, and codebase reality live in different places.

02 · Technical reality

The work looks ready until implementation starts.

Dependencies, ownership, tests, and architecture risks arrive too late.

03 · Quality

Correctness is harder to see than output.

AI can produce plausible work faster than the team can validate it.

04 · Flow

More work starts than can be inspected.

The team accelerates ticket movement but creates waiting, review piles, and rework.

05 · Learning

The system does not remember.

Delivery signals do not change refinement, working agreements, or the next experiment.

The takeaway artifact

Turn the result into one sprint-sized experiment.

No transformation theater. No tool rollout disguised as improvement. One repeated friction. One change to the delivery system. One signal.

Step 1

Name the drag.

Where does work lose clarity, quality, or momentum?

Friction Where it shows up Step 2

Choose the experiment.

What will the team do differently for one sprint?

Small experiment Step 3

Define the signal.

What evidence will tell you whether the change helped?

Signal Step 4

Place the AI.

Will AI analyze, plan, execute, summarize, test, or inspect?

AI role Step 5

Place the humans.

What must people still judge, review, approve, or challenge?

Human check Step 6

Update the system.

If it works, what changes in your working agreement?

Working agreement update

Use it with your team

One result. One conversation. One experiment.

The score is a door, not a verdict. Here is the difference between leaving with a tools list and leaving with a system change — and three ways to run this with a real team.

The usual move

“Go try AI.”

Problem Too broad to act on. Next step Tool browsing, prompt swapping, uneven adoption. Risk More output without a better delivery system.

The better move

“Find the drag.”

Problem Specific constraint in context, planning, quality, flow, or learning. Next step One team experiment in the next sprint. Signal Evidence that decisions, feedback, or quality improved. Use it solo first

Run the diagnostic yourself before raising it with the team. Your own score is a draft agenda: the top friction is the conversation worth having.

Use it in a retrospective

Have each person answer individually, compare friction patterns, then choose one shared system change for the next sprint.

Use it in refinement

Run the scorecard against one upcoming story. If technical reality, acceptance criteria, or shared context are weak, improve the work before starting it.

The point

Do not leave with a list of tools. Leave with one constraint to improve.

Run the Friction Finder with your team, then bring the result to refinement, retro, or sprint planning.

endash.us/toolkit/items/ai-delivery-fiction-finder Copy link

En Dash Consulting

Better decisions. Better flow. Better work.

AI-enabled delivery works best when agile leadership, product thinking, and software craft reinforce each other.

Explore the toolkit Talk with En Dash