Most companies have now run an AI pilot. Fewer have changed how they work. The difference is rarely the model. It is that the pilot was set up as a project, with a start, an end and a report, when what was needed was a new habit.
Here are three ways to start that we have seen lead to working tools in weeks rather than a strategy deck in months.
1. Start where the hours go
Pick one team and ask a boring question: where does the time actually go? In sales it is often research, proposals and follow-ups. In marketing and media it is producing and adapting content across channels. In operations and support it is the same documents and questions, every day.
Choose the one task that is frequent, tedious and has a clear definition of "good". Build an assistant or workflow for that task only, on your own knowledge and templates, and measure it against the manual way.
You will know within two weeks whether it helps. If it does, the team will tell everyone. That is worth more than any internal launch campaign.
2. Give people a safe sandbox before you give them rules
Policies written before anyone has used the tools tend to ban things nobody wanted to do and miss the things people actually do. Instead, set three plain rules on day one (which data may go where, what needs a human check, who is accountable), put the tools in the systems people already use, and let a small group try them on real work.
After a month you will have concrete examples of what worked, what went wrong and what people wish they could do. That is the material for a policy people will follow.
3. Treat your developers as part of the transition
Engineering teams are going through their own shift: AI-assisted coding, agents that carry out tasks, new ways of testing and reviewing. Done well, this raises quality and speed. Done badly, it produces code nobody understands.
Include developers in the transition from the start. Agree how AI-generated code is reviewed, how it is tested, and where it must not be used. Pair experienced engineers with the tools and let them set the standard for the rest of the team.
What ties the three together
- Small scope, real work. Every starting point is one team, one task, one measure.
- Working tools, not demos. If it is not connected to your data and your systems, it is a toy.
- Habits over initiatives. The goal is that in six months nobody calls it "the AI project" any more. It is just how the work gets done.
Start with one. The second one is much easier.