Mapping Touch Edges: How to Find Where AI Actually Helps

Today I spent several hours mapping communication flows through an organization. Every handoff. Every bottleneck. Every place where information dies in transit.

The output wasn't "here's where to deploy AI." It was a map of touch edges — the points where humans coordinate with systems or each other. Only after mapping those edges could I identify which ones were actually breaking.

This is the part most AI implementation skips. And it's why most AI implementation fails.

What's a Touch Edge?

A touch edge is any point where information passes between people, or between people and systems. Some examples:

A customer submitting a form on your website
A salesperson asking ops about inventory status
A manager approving a purchase order
An alert triggering when stock runs low

Each touch edge has properties: Who initiates? Who receives? What's the expected response time? What happens when it fails? These properties determine whether the edge is working or broken.

Key Insight

Most organizations know their processes. Few have mapped their touch edges. The process says what should happen. The touch edges show what actually does.

The Three Questions

For each touch edge, I ask:

1. Does the information reach its destination?

This sounds obvious, but you'd be surprised. Customer inquiry forms that nobody checks. Emails sent to shared inboxes where everyone assumes someone else will respond. Slack messages lost in channel noise. The simplest failure mode is just: the message never arrives.

2. Is the intent preserved?

Even when information arrives, meaning can be lost. A customer asks for "the blue one" and by the time it reaches fulfillment, nobody knows which blue one. A priority gets marked as normal because the original urgency wasn't captured. Intent leaks at every handoff.

3. Is the response timely?

Information might arrive intact but too late to matter. By the time the stock alert reaches the production team, customers are already churning. By the time the approval comes through, the window has closed.

A touch edge fails when any of these break down. And these are the only edges worth automating.

Why This Matters for AI

The instinct with AI is to automate the most visible work. Customer service chatbots. Report generation. Email drafting. These are easy to implement and easy to demo.

But visibility doesn't correlate with impact. Often the biggest coordination failures happen in the invisible edges — internal handoffs that nobody monitors, queries that drain hours from high-value employees, escalation paths that dead-end silently.

Real Example

One organization had sales reps pinging operations 10-20 times per day asking "do we have X in stock?" Each query took 2-3 minutes to answer. The data existed in a system — sales just couldn't access it. A simple bot on their internal chat saved 30-50 hours per week. Not glamorous. High impact.

The Wrong Edges

Not every touch edge should be automated. Some edges work fine. Some are complex enough that human judgment is essential. Some are so rare that automation isn't worth the setup cost.

The failure mode I see most often: automating edges that work fine while ignoring edges that are actively failing. A company deploys an AI assistant for their website while their email inbox remains a black hole. They get the chatbot demo. They lose customers to unanswered emails.

How to Map Your Touch Edges

The process is simpler than it sounds:

List every way information enters your organization. Customer channels, vendor communications, internal requests, automated alerts.
For each input, trace where it goes. Who receives it? Who acts on it? What systems does it touch?
Ask the three questions at each handoff. Does it arrive? Is intent preserved? Is response timely?
Mark the failures. These are your candidates for AI intervention.
Prioritize by impact. What's the cost of each failure? What's the effort to fix it?

The output is a map. Not of your org chart — of your actual coordination patterns. That map tells you where AI helps and where it's a distraction.

The ALC Lens

In ALC terms, each touch edge has properties that determine whether human-system communication succeeds:

Intent specification — Can the sender express what they need?
Machine orchestration — Does a system help route and track?
Asymmetric interpretation — Does the receiver understand the same thing the sender meant?
Stratified fluency — Does the edge assume expertise that users might not have?

Broken edges usually fail on one or more of these. An AI agent can help by improving intent capture, adding orchestration, reducing interpretation gaps, or lowering fluency requirements.

Bottom Line

Don't start with "what can AI do?" Start with "where does coordination fail?" Map the touch edges. Find the breaks. Then — and only then — figure out which breaks an AI agent can fix.

This is what I do. Not selling AI solutions — mapping organizational reality first, then proposing targeted interventions. The unsexy work that makes AI actually useful.