AI Automation for Teams of One to Five: What Actually Works
Enterprise automation advice does not survive contact with a five-person team. Here is the version that does, learned from two years of small-team workflows.
Most automation advice is written for companies with an ops person. A five-person team has no ops person; it has whoever drew the short straw. Two years of automating tiny teams has taught me which automations survive that reality.
What works: automations that trigger on something that already exists. A form submission becoming a Slack message. An invoice attachment getting filed and logged. A missed-email follow-up. These survive because they require no behavior change, and behavior change is the thing tiny teams cannot afford. Every automation that needed someone to 'remember to log it first' died within a month.
What fails: anything with a queue that needs tending. AI drafts waiting for approval, triage buckets that need sorting, dashboards that need reviewing. In a big company someone owns the queue. In a small team the queue outlives its champion within weeks and becomes a guilt generator. If an automation produces work, it is not an automation, it is a chore distributor.
The AI-specific layer that earns its keep in small teams is summarization of accumulated noise, applied on a schedule. A weekly digest of what happened across the tools, generated without anyone asking. This works because it creates information without demanding action, and the team can ignore it safely until the week it matters.
The scoring rule I now use before building anything: does this automation remove a step or add a monitor? Remove a step, build it. Add a monitor, and it needs an owner with allocated time, which a tiny team usually does not have. Three automations that remove steps beat thirty that watch things.