August 7, 2026
Expert Series features guest perspectives from leaders across Nova Scotia’s tech community, sharing practical insights, lessons learned and real-world expertise through the Digital Nova Scotia platform.
In this edition, John Cody, Co-founder of Trailblaze AI, explores a common mistake many organizations make when adopting AI: automating processes before they’ve been proven. He argues that successful AI adoption starts with understanding and refining the work itself, not simply applying technology to inefficient systems.
A 3-question test before you build anything
A plumber buys an automation platform on a Tuesday. By Thursday he has a 14-step follow-up sequence, an AI chatbot on his website, and automated review requests firing after every job. By the following Tuesday he is manually apologizing to a customer who got a “how was your service?” text before the job was finished.
He does not have a technology problem. He has a sequence problem. He automated before he had a process worth automating.
This happens constantly. Business owners buy the platform expecting it to build the sales process or the retention process for them. It will not do that. It scales what you already have. If what you already have is confusion, you get faster confusion.
Three Questions Before You Build Anything
Before you put any workflow into any platform, answer these honestly.
1. Does this task change based on who the customer is, or is it the same motion every time? An appointment reminder is the same for everyone. A proposal is not. That line matters.
2. If this goes out wrong, is it annoying or is it a problem you cannot fix? A reminder sent on the wrong day is a quick correction. A robotic reply to a one-star review lives on your Google profile forever.
3. Have you already run this manually and confirmed it works, or are you hoping the software figures it out? This is the trap. Most businesses skip it. They have never tested the timing, never confirmed the message actually converts. They just build it in the platform and hope. Automation does not invent your process. It repeats your process. If the process does not exist yet, there is nothing to repeat.
What Belongs in the Machine
Missed-call text-back and speed-to-lead follow-up. A Hatch analysis of over 132,000 speed-to-lead campaigns found that 88% of home service businesses take longer than five minutes to respond to a new lead. The InsideSales and MIT Lead Response Management Study found that responding within five minutes makes you 21 times more likely to qualify that lead. The text-back requires zero judgment. The lead called, nobody picked up, the system sends a message. Same motion every time.
Appointment reminders and no-show reduction. The message does not need to know anything about the customer. It just needs to go out on time.
Review requests. The ask does not change client to client. After the job is done, send the link. That part is mechanical. Keep this separate from review responses, though. Automate the ask. Never automate the reply.
Invoice and payment reminders. One of the fastest automation wins. The trigger is a date. The message is standard. The outcome is fewer overdue invoices without anyone on the team chasing payments by hand.
Intake data capture. Collecting structured fields (name, address, service needed, preferred time) conversationally instead of making someone fill out a form. Works when the questions are the same for every lead. If the questions change per lead, you are back in judgment territory.
What Does Not Belong in the Machine
Responding to reviews, especially negative ones. This is the other half of the review category, and it is the opposite call. A canned reply to a one-star review reads exactly like what it is. It makes the situation worse. Every response needs a person who reads the specific complaint and replies in a tone that fits.
Proposals without a locked customer profile. A templated proposal for a customer you have not clearly defined does one of two things. It says nothing specific, so it looks generic and kills trust. Or it says something specific that happens to be wrong, and that kills trust faster. Automate the scheduling, the contract terms, the payment setup. Do not automate the recommendation until you know exactly who you are talking to.
Original social content. Scheduling posts and distributing them across platforms? Fine, automate that. But generating the actual message before your brand voice exists in any real form is how you get a feed that sounds like nobody wrote it. The voice has to exist before a machine can copy it.
Sales conversations on anything with real deal size. An AI agent can qualify a lead and book the call. That is a repeatable motion. But the actual sales conversation, the part that requires trust, should not be handed to a bot. Qualify and route, yes. Persuade and close, no.
The Platform Is Not the Strategy
The best automation tool in the world will not fix a process you have not built yet. If the manual version works, automate it. If the manual version has not been tested, test it first.
Run the three questions before you build anything. If the task is the same motion every time, the cost of a mistake is low, and you have already proven the process by hand, build it. Everything else stays with a person until you have earned the right to hand it off.
Not sure which parts of your business are ready? Book a free 30-minute Automation Power Session with John and Colin and we will walk through it with you.