Stop Replacing Humans With Software. Replace the Phone Calls.

One of the fastest ways to find an automation opportunity is to listen for the sentence, “I’ll call them and find out.”

There is nothing inherently wrong with phone calls. I like humans. Some of my favorite people are humans.

But if an operational process requires one person to repeatedly call another person solely to discover the state of work, the humans are compensating for missing infrastructure.

Real estate closings are full of this. Did the notary receive the documents? Did the appointment happen? Were there exceptions? Are scanbacks available? Did the package ship? What’s the tracking number? Did somebody reschedule?

These are not relationship questions. They are state queries being executed through telephony.

Humans are expensive message queues

When one company calls another company to ask whether an event occurred, a person becomes the integration layer between two systems.

The caller has to notice that information is missing, stop what they are doing, contact someone, wait for an answer, interpret it, enter it somewhere else and potentially tell several additional people.

Congratulations. We have implemented distributed computing using humans.

Humans are extremely expensive message queues. They’re also terrible APIs.

They take lunch.

Automation should remove uncertainty, not people

This distinction is important to me because technology companies love talking about “replacing manual processes,” and somehow the conversation immediately becomes replacing the people performing them.

That is frequently the least interesting opportunity.

The escrow officer has judgment. The transaction coordinator understands context. The notary is physically present with the signer. The real-estate agent manages a relationship. Those are valuable human functions.

Having those same people spend their time copying status between systems is not.

A better architecture allows the systems to propagate state automatically and asks humans to handle the exceptions.

The appointment was completed? Publish that event. Scanbacks arrived? Make them available and notify whoever needs them. A signer did not appear? Escalate the exception. Shipping generated a tracking number? Attach it to the transaction.

Nobody needs to call three people to discover any of those facts.

Exceptions are where humans become more valuable

There is a misconception that automation should make a workflow completely touchless. In messy real-world processes, that goal can produce brittle software because the world refuses to stay on the happy path.

I would rather automate the routine path aggressively and make exceptions extremely visible.

If 95 percent of transactions proceed normally, software should quietly move those along. The remaining five percent deserve human attention precisely because something unusual happened.

That gives people more time for the work computers are bad at: judgment, negotiation, reassurance, problem solving and dealing with the person who somehow signed every page except the one containing the signature line.

The best automation can feel boring

Good operational software often isn’t dramatic. Nobody gathers around a monitor and applauds because a status changed from “scheduled” to “completed” and six interested systems learned about it automatically.

They simply stop making phone calls.

That quiet disappearance of unnecessary work is one of my favorite measures of successful integration.

It is also how I think about Coverity. The interesting problem is not creating another dashboard for people to stare at. The world does not suffer from an insufficient number of login screens.

The interesting problem is connecting the participants and propagating enough trustworthy state that people stop having to ask each other what happened.

I don’t want software to replace the humans in a closing.

I want it to replace the sentence, “Let me call them and find out.”