Definition
Services in which a provider operates AI on a client's behalf. The client doesn't buy software, write prompts, wire up integrations, or learn a dashboard: the provider configures the system, runs it day to day, keeps a human in the loop, and reports the outcomes it produced.
What it looks like in practice
Setup for a home care agency with eighteen caregivers is one onboarding call: which services it offers, its intake questions, the counties it covers, who to call when something's urgent. Everything after that happens on the provider's side. Inquiry calls from adult children get answered and qualified, a nurse assessment gets scheduled, caregivers get shift reminders, and families get an update after the first visit. When the agency adds overnight care in the spring, the owner sends one email and the provider updates the intake script, the scheduling rules, and the website copy. Nobody at the agency logs into anything to make that happen.
Why it matters
The word doing the work here is 'managed'. Plenty of AI products are capable and still fail in a small business, because someone has to own them; write the prompts, fix the integration that broke, notice when the tone went off. That someone is usually the owner, who already has a job. A managed service moves that ownership to the provider, so the business gets the capability without a new operating burden. The tradeoff is control and lock-in: you're trusting someone else's judgment on how the work runs, which makes it worth asking upfront what you can see, what you approve, and what happens to your data and phone numbers if you leave.