All journal entries

The Implementation Gap

The fog around AI: separating the hype from what actually works

There's more misinformation about AI than almost any technology in a generation, and most of it keeps good operators from real wins. Here's what's actually true, what's noise, and how to tell the difference.

7 min read
A Newfoundland and Labrador headland at dawn as a lighthouse beam cuts through thick sea fog, clarity emerging from the fog
A Newfoundland and Labrador headland at dawn as a lighthouse beam cuts through thick sea fog, clarity emerging from the fog

Ask ten people what AI can do for their business and you'll get ten different answers, and most of them are wrong in some direction. Some are certain it's a magic button that runs the company while they sleep. Others are convinced it's a hallucinating toy that can't be trusted with anything real. The truth sits in a narrow band between those two extremes, and the noise on either side is expensive: it keeps capable operators from wins that are sitting right in front of them.

We build AI into real businesses for a living, so we spend a lot of time clearing fog before we write a single line of code. Here are the myths we run into most, and what is actually true underneath each one.

Myth 1: “AI is going to replace your people.”

This is the headline that sells the most clicks and causes the most fear, and in the businesses we work with it is almost always backwards. The work AI is genuinely good at is the work your best people already hate: re-keying data between systems, chasing the same status update five times, turning a pile of invoices into a report. Automating that does not shrink a team, it hands a labour-strapped team its hours back. The operators who win treat AI as a way to do more with the people they already struggle to hire, not a way to have fewer of them.

Myth 2: “There’s one AI tool that does everything, just buy it.”

There isn't. The vendor demo that looks like it solves your whole business is showing you the ten percent that demos well. Real implementation is plumbing: connecting the AI to your actual systems, your inbox, your spreadsheets, your booking software, your accounting, so it works on your data, inside your workflow, the way your team already operates. The model is the easy part. Making it reliable inside one specific business is the work, and it is why “buy a subscription” and “implement AI” are not the same sentence.

Myth 3: “AI just makes things up, so you can’t trust it for anything serious.”

It is true that a raw chatbot will confidently invent an answer. It is not true that this makes AI useless for serious work, it means you build for it. A well-made system grounds the AI in your real documents, shows its sources, flags its own confidence, and keeps a human in the loop on anything that matters. That is exactly how Ava, the assistant we built into Apex Hub for a real-estate team, handles legal closing dates: she extracts each one from the signed agreement, cites where it came from, and waits for the agent to confirm before anything is filed. Trust is not assumed. It is engineered.

Myth 4: “You need to be a big company with a data team.”

The opposite is closer to the truth. The businesses with the most to gain are small and mid-sized ones where a few repetitive workflows eat a disproportionate share of the week, and where there is no IT department to build a fix. You do not need a data lake or a machine-learning team. You need one workflow that is clearly costing you time, and someone to build the system that handles it. That is a matter of days, not a moonshot.

Myth 5: “It’s too early to bother, or too late to catch up.”

Both versions of this are paralysis dressed up as strategy. It is not too early: the tools are stable enough today to run real operations, and roughly one in eight Canadian businesses is already using them. It is not too late: only about two percent are seeing measurable ROI, which means the field is wide open for operators who actually implement instead of merely adopting. The gap between “we have AI” and “AI is making us money” is execution, and that gap is where the next few years will be won.

How to tell signal from noise

When you hear a claim about AI, run it through a few simple filters:

  • Does it name a specific workflow, or just say “AI”? Specifics are real; “AI will transform your business” is a slogan.
  • Does it show where its answers come from, or just produce them? Sourcing and confidence scores are the mark of a serious system.
  • Does it keep a human in the loop on anything that matters? Good implementation augments judgment, it does not pretend to replace it.
  • Is someone selling you a tool, or building you a solution? A subscription is not an implementation.
  • Can you measure the result in hours saved or dollars earned? If no one can tell you the ROI, treat the claim as marketing.
The hype says AI will change everything. The reality is quieter and far more useful: built well, into the right workflow, it gives a busy team its time back. That is the whole game.

If you have read a frightening headline or a too-good-to-be-true pitch and you are not sure which one you are looking at, that is a good reason to talk. Thirty minutes, and we will tell you plainly what AI can and cannot do for your specific business, no fog.

The fog around AI: separating the hype from what actually works · Kipnetic