We gave the team ChatGPT or Copilot and nothing changed. Why?
A ChatGPT or Copilot licence helps each person write and summarize faster. It doesn't change your operations, because the assistant works beside your tools, not inside them. Here's the difference with AI built into your operations, and where to start.
:quality(80))
The scenario is familiar. The company buys ChatGPT or Copilot licences. There's a training session over the lunch hour. A few people use it for their emails, someone else to summarize documents. Then life goes back to normal.
3 months later, quotes are still prepared by hand. The CRM is still updated after every meeting, when someone remembers. And the question “where is this at?” comes up just as often as before. Very reassuring.
It's not the team's fault, or the tool's. The licence simply solves a different problem from the one you thought you were solving.
Why don't ChatGPT or Copilot change anything in your operations?
Because the assistant works in a separate window, beside your tools. To use it, someone has to go get the information, paste it into the window, read the answer, then copy it back into the right software.
The assistant sits beside the work, not inside it. The handoffs between your tools stay human: it's still someone on the team who bridges the email, the CRM, the quoting software and the accounting.
It's a bit like hiring a very talented person and putting them in a room with no phone and no access to the files. For them to help, you have to bring them everything, then carry everything back.
The text gets written faster. The file itself moves at the same pace.
What does an assistant like ChatGPT or Copilot do well?
It helps each person with their own work, and that's already a lot. Claude, ChatGPT, Copilot or Gemini write a first draft, summarize a long document, rephrase a delicate reply or explain a topic you don't know well. For many people, it's the end of the blank page.
The gain is real, but it stays in each person's pocket. It depends on who thinks to open the window and who knows what to ask. The company's process doesn't move.
That's why training matters. Used well, with an eye on mistakes and on the confidential information that gets pasted in, an assistant saves time every day. It just won't move your files along for you.
Examples
- Writing the first draft of an email, an offer or a report
- Summarizing a long document before a meeting
- Rephrasing a delicate reply to a customer
- Explaining a regulation or a concept you don't know well
What's the difference between a licence and AI built into your operations?
A licence gives each person an assistant. AI built into your operations is connected to your tools, your data and your processes: it works where the file already travels.
Connected doesn't mean your software gets replaced. It means AI has controlled access to the tools your team already uses: email, the CRM, the quoting software, internal documents. It sees what the person is allowed to see, no more.
In practice, it reads a request where it comes in. It looks up the relevant information in your systems. It prepares the next action in the right software, updates the tools and flags anything out of the ordinary. A person approves what matters.
AI has to live in your operations, not just in a chat window. That's where hours get freed up: in the handoffs nobody has to do anymore.
What does built-in AI look like on a real day?
Take a quote request at Toitures Boréal, a fictional company. A customer writes in to have the roof of their warehouse redone.
With built-in AI, the email is read as soon as it arrives. The customer is found in the CRM, with their history. A quote is prepared in the usual software, following the company's pricing rules. The roof area is missing? It's flagged right away, and the question goes out to the customer.
The estimator opens the ready quote, adjusts a price and approves it. The CRM updates itself and the follow-up is already scheduled. Nobody opened a chat window. AI handles the handoffs. People make the important decisions.
The same principle applies elsewhere: a tenant request that comes in by email, a work order filled out on the job site, a timesheet to chase. Each time, AI makes the trip and a person keeps the decision.
Why do so many companies stop at the licence?
Because it's the easiest step. Buying licences takes a few minutes and doesn't require changing any habit. Connecting AI to a process first requires understanding it: who receives the request, where the information goes, who decides and where it gets stuck.
A process has to be understood before it's automated. Otherwise, you automate the detour instead of removing it.
It's less spectacular than a demo, but that's where the value is: in the real path of your files, with the exceptions and shortcuts only your team knows.
Should you drop your ChatGPT or Copilot licences?
No, keep them. The assistant your team already knows can become its way into your tools. Once connected, it doesn't just answer anymore: it finds a file, prepares a task or updates a record, with the access rights of the person asking.
The licence isn't the problem. The problem is stopping there. A licence is a bit like a gym membership: having it in your wallet changes nothing until you actually use it.
And training is still useful. A team that's comfortable with its assistant adopts AI connected to its tools faster. That's actually what we do during the AI integration day: train the team on its real files, then connect AI to 2 of its tasks.
Where do you start so that something actually changes?
With a single process, the one where your team bridges 2 pieces of software. A request someone retypes, a status updated by hand, a report compiled from data that already exists.
Next, count what that handoff costs. For example, if 3 people each retype 20 requests a week and each request takes 6 minutes, that's 6 hours a week spent bridging the gap. It's a hypothesis: redo the math with your real numbers. We explain how in 5 steps to automate processes in an SMB.
Then connect AI to that specific handoff, with a person who approves. Not to the whole company at once: not everything deserves to be automated, and not everything needs AI. To sort out what you hand over to AI and what stays human, see also Should I hire or automate?
After a few weeks, redo the math. If the time spent bridging the gap has melted away, you have your answer. And probably your next process to connect.
Frequently asked questions.
Why doesn't ChatGPT change anything in my company?
Because it works in a separate window, beside your tools. It helps each person write and summarize faster, but the handoffs between your software stay manual. For operations to change, AI has to be connected where your files travel.
What's the difference between Copilot and AI built into operations?
Copilot, like ChatGPT, Claude or Gemini, is an assistant each person consults. AI built into operations is connected to your tools, your data and your processes: it reads requests, prepares actions in the right software and flags exceptions. A person approves what matters.
Should we stop paying for our ChatGPT or Copilot licences?
No. They're still useful for writing, summarizing and searching. And the assistant your team already knows can be connected to your tools to go further. The problem isn't the licence: it's stopping there.
Do we have to replace our software to integrate AI?
Rarely. Your existing tools can often be connected instead of replaced. The problem isn't always that the software is bad: it's often that each tool works in its own corner.
Will built-in AI replace members of my team?
That's not the goal. AI takes over the handoffs: searching, sorting, preparing, updating. Your team keeps the judgment, the customer relationships and the important decisions, with more time to focus on them.
A licence is a tool for each person. Built-in AI is a process that changes for the whole team.
If nothing has moved since you bought the licences, it's not a failure. It's a sign that it's time to look at where your files pass from hand to hand, and to connect AI there.
Read next
An idea in mind?
Let’s chat!
:quality(80))
AI and automationCan you put confidential data into ChatGPT?
AI and automationHow do you find your company's knowledge with an AI assistant?
AI and automationWill AI replace my employees?