Rolling out technology across multiple locations
Choosing systems, getting them adopted at unit level, and working out where AI earns its place. Tools applied to a specific problem are filed under that problem.
On this page
How do you choose software for multiple locations?
Start from the problem, not the category. "We need an AI tool" produces a subscription. "Applicants wait two days for a reply" produces a decision you can judge afterwards.
Then check it works at your scale. Plenty of tools are built for a single site and handle several locations by making you log in separately for each one, which is where the time goes.
Ask what it costs to leave. Who owns the data, how you get it out, and how long the contract runs. That question is unpopular in a demo, which is why it is worth asking there.
Why do rollouts fail at unit level?
Almost always at adoption, not at purchase. The system works, and the unit keeps running the old process beside it because the old one still functions.
Running both is the failure. Two sources of truth means the data is wrong everywhere, so nobody trusts the new system and everybody has a reason to keep the spreadsheet.
Pilot in one unit, but not your best one. Your best manager will make anything work, which tells you nothing about the fourteen units where it actually has to survive.
What does AI actually do well right now?
Repetitive language work at volume. Replying, drafting, summarising, translating, sorting and routing. These are jobs where being fast and consistent beats being perfect.
It also does well on things nobody currently does at all, because there is no time. Nobody reads every review across nine locations each week, so a summary of them is a gain rather than a replacement.
It is weakest where being wrong is expensive and you would not check the answer. Anything involving money, legal exposure or a decision about a person needs a human between the output and the action.
What should you not automate?
The moment a customer is already unhappy. Automation at that point reliably makes it worse, and the cost of getting it wrong is the relationship.
Do not automate the hiring decision, only the steps around it. Screening, scheduling and follow-up are process. Who joins your team is judgement, and it should stay with the person who works alongside them.
And do not automate a process nobody has defined. Automation makes whatever you currently do happen faster, including the parts that are wrong.
Playbooks and tools
Filed under the problem they solve
There are no playbooks here by design. A playbook about AI screening lives in Hiring, one about AI scheduling lives in Labor, and one about review automation lives in Marketing.