What changed when AI could work with my tools, not just talk about them. About twelve months ago, I had a question about the future of work. Was I going to spend the rest of my career copying and pasting? Copy an email into AI. Paste the answer into a text. Bring the reply back. Find the booking information. Explain who everybody is. Then put the result somewhere useful. Apparently the future needed a clipboard operator, and I had the job. These days, I do very little of that compared with a year ago. Not zero. I still copy and paste things. I just don’t have to carry every piece of information between tools before useful work can happen. I wrote about the foundation in How AI became a force multiplier in our business. Here’s what changed in the day-to-day work. Connect the information Most of what I was feeding AI already existed somewhere. The customer had sent the email. The booking was recorded. The conversation was in the group chat. Connecting those sources meant I could ask about the work without assembling the whole packet myself. Our lead desk is one example. A scheduled sweep reads available call records, texts, emails and booking information. It creates or updates entries in our own customer tracker, our CRM, and flags uncertain matches for confirmation. Those outside sources stay read-only in this workflow. The permitted updates happen in our own lead records. People still decide and send customer replies. The result has somewhere useful to go. I’m not left with another summary that somebody needs to retype. Give AI, code and people different jobs I use a simple division of the work: AI interprets. Records supply the facts. Code checks the rules. Permissions control what can happen. In practice: • AI makes sense of a message or request and proposes what it might mean for the work. • Records hold the bookings, assignments, qualifications and other details. They still need to be current and checked. • Code runs defined checks, such as whether assignments conflict or a vehicle has enough seats. • Automation repeats predictable steps, such as running a scheduled check or routing an approved update. • The crew app shows people the assignments and details relevant to their work, drawn from the shared records. • People approve consequential staffing, customer, money and safety decisions. Not every part needs AI. A crew member opening an assignment in the app doesn’t need a new AI answer. They need the approved plan. This is where building up the code creates leverage. Once we’ve written and tested a rule, the same check can run across many trips and changes. I don’t need AI to work out the seat-count rule again every time someone changes a booking. AI can help write and debug that code, too. The changes still need review and testing before they go live. A useful improvement becomes part of how the next job gets done. It doesn’t have to end as a good answer buried in a chat. Give repeatable work a regular job The next improvement was not having to remember to start everything myself. The daily lead desk has a defined job: review these sources, maintain these approved records, and flag what needs a person. Don’t contact customers. I can review the result without explaining the task again tomorrow. Routine automation starts the repeatable work; AI helps interpret the information where needed. There’s still upkeep. Connections can fail, records can be wrong and rules can change. Somebody needs to check the results and maintain the system. Getting here took organized records, written rules, code, permissions and testing. Connecting a few logins wouldn’t have done it. Start with one annoying workflow Think about the inquiry you enter twice, the change you explain in three places or the task that goes quiet unless you remember to check it. Ask: • Where does the information live, and how do we verify it? • Which checks or repeated steps can we put into code or automation? • What needs a person before anything changes? You don’t need to solve this for the whole business at once. Pick one job and make it easier to finish correctly. I still need to understand the operation and make decisions. I’m spending less time carrying information around so those decisions can happen. If you have a workflow like that, I’d be interested to hear about it. I wanted AI to help with the work. I wasn’t particularly excited about becoming its copy-paste guy.