Bringing AI to a small team
A small property management team had several overlapping AI tools and no rules for using them. I wrote the policy, chose one standard tool, connected it to our systems and trained everyone, then used it myself to build the company's first internal apps.
The problem
People were paying for several AI add-ons, using them inconsistently, and unsure what was allowed with tenant and owner data.
Constraints
Tight budget
Every tool had to justify its cost; consolidating was part of the point.
Sensitive data
Tenant, owner and financial data needed clear rules.
Mixed comfort
Some staff were enthusiastic, some wary.
Key decisions
Policy before tools.
I wrote an AI use policy covering what data can go where, before rolling anything out.
One standard tool.
We standardised on Claude, connected it to our core systems and cut overlapping add-ons.
Train on real work.
The company-wide session used the team's own tasks, followed by a tips channel and office hours, because one training isn't adoption.
Guardrails in everything I built.
A person approves every payment, each app has limited access, and I measure accuracy on real data before anyone relies on a tool.
How it works
Outcome
- Trained the whole company on real work, ahead of the goal date.
- Rolled out AI meeting notes and built an agent that replaced a paid task tool for accounting's recurring work.
- Built two production apps and about 25 working prototypes in roughly seven weeks.
“Great job today! … I think it was really valuable!” — CEO, after the company-wide training
What I'd do differently
- Name a sponsor in each department up front. Tools landed fastest where someone besides me owned the change.