We gave everyone access and almost nobody uses it. Why?
Access was never the thing holding it up. Technical roles don’t predict who picks up AI. Willingness to try new things does, and the gap between the best and worst person on the same team runs about 100 to 1. Buying more licenses doesn’t move that.
Half the team doesn’t trust the output. How do you deal with that?
Usually they got burned on a task it was bad at. Most people use AI to ask a question, get an answer, and go back to doing the work by hand. That’s an advisor, and it’s maybe five percent of what’s available to you. Being specific about the jobs it’s good at works better than asking anyone to trust it in general.
Nobody knows what they’re allowed to put into these tools.
That’s the most common answer on the baseline survey, and it does more damage than distrust does. People who aren’t sure either do nothing, or do it in a personal account where nobody can see it. Say out loud what individuals are free to use it for and what has to come through ops, and both of those go away.
Is this about replacing people?
No. Everything built gets handed to the people who use it, with the training to change it. Plenty of vendors build something you have to keep paying them to maintain. We give you the keys back. The audit also tells you what to leave alone, because a bot sitting in a meeting changes how people talk, and a reply that reads as automated costs more customer trust than the minutes it saved.
For a lot of this it’s faster to just do it myself.
For something you do once, probably. Two questions sort any job: how often does this happen, and does it have to come out identical every time. Automating an inefficient process just makes you inefficient faster, so the workflow gets fixed before anything gets built on top of it.
How much of this is really an AI problem versus a systems problem?
Half of what gets called AI implementation turns out to be RevOps with a different name on it. If the data is scattered across five tools with no thread tying it together, nothing built on top of it will hold. It’s not a distraction from the AI implementation project, it’s exactly the type of work we get into.