AI rollout: Approach
I have spent more time worrying about what can go wrong rolling out AI than what might go right.
There's no doubt that AI is powerful and it's improving daily, but a powerful tool does not a super employee make.
🔹🔹🔹
Have you ever worked at a company where everyone is shit-hot at what they do? Well, most companies are not like that, and that's okay. There are a range of skillsets. Some will flourish with AI; some will not.
Imagine we’ve built a business which runs effectively. There is room for improvement, but on the whole things do actually work.
Processes were built intentionally. Decisions were made based on company experience dating back to day zero. Some don't know exactly why we do everything the way we do, but they respect it.
Policies have been introduced on a need-to basis over time. Each has a purpose and a behaviour it's intended to influence.
We chose not to rule by policy or treat our staff as if they've committed a crime. We trust our staff, and our narrow range of policies reflects that.
Then one day, there's a new Sheriff in town. They've got big ideas - a million of them - and they're implementing them today. There's no time to pause. It's all go, go, go with this guy; standard processes are rolled out by the dozen.
If we want to be the best, then this is what we've got to do.
Who are we to challenge the Sheriff? He seems to know what he's talking about. It's why we brought him in. We should trust him.
Policies start multiplying like rabbits.
"To run a business, we've got to get a better handle on the staff” the Sheriff says.
Policy, policy, policy. Go, go, go.
Six months have passed since the Sheriff came to town, and our business is unrecognisable. Most processes now go via the Sheriff, as he's the only one who can make sense of it all.
Thank goodness we have him.
Policies have been revolutionised. No more ambiguity over what is or isn't allowed. Can someone have a day off for their dog passing away? Don't worry, there's a tidy 16-point policy for that.
We are riddled with micro-policies, and although nobody has a clue why any of them exist, it doesn't matter because we have the Sheriff, and what he says goes.
Thank goodness we have him.
🔹🔹🔹
What has really changed? We have more - policies, processes, whatever we desire. But more doesn’t make us more effective.
It’s easy to lose sight of what makes an impact when the tool in front of us is designed to dazzle. We think we’re using it to execute a vision, yet in reality we’re using it to do our thinking.
My concern with deferred thinking is that we won't know why we do anything. Why do we have that policy? What behaviour is it supposed to influence? Why is the process designed like that? Are those actually my tasks, or are they inferred?
Or maybe none of it actually matters?
This might be the future. We don't need to know anything about anything; we simply have to follow the instructions given to us.
I’m not sure that’s a future I’m particularly interested in.
I believe there's a happy medium: a way of working in unison with AI which can raise the bar for everyone. I can see what the worst-case scenario is, and I have a plan to mitigate it.
AI principles
My guiding principles on how to use AI. These are designed so that whenever someone is making a decision, they can point to them and say it aligns. The intention is for me to be able to guide from afar.
Governance
Putting structures in place to oversee how the team are using AI and the output of it. Someone has to be watching what's going on - even if it's AI doing the monitoring - so I can course-correct if necessary. If a hundred new files appear, it's good to know when it happens and not weeks down the line.
Champions
The people on the front line driving the AI rollout. These are the go-to people for the team. This can't be me for two reasons: 1) it's a fantastic growth opportunity for people in the team; 2) I need to oversee the rollout, not be in the middle of it.
I have a vision. Now how can the Sheriff help me bring it to life?
Stoirm
