How our AI actually works

AI you can actually trust

Not a bigger feature list - a different design. Forge's AI is deterministic-first, tags every output with where it came from, never acts alone on money or legal matters, and logs what it does.

Why this page exists

Buildium and NARPM's 2026 Property Management Industry Report found AI adoption among property management firms jumped from 20% to 58% in a single year.¹ That's adoption, not trust - a feature a firm switched on isn't the same as a feature its team actually relies on for a money or legal decision. Closing that gap isn't about a longer feature list. It's about showing your work.

The four rules

Deterministic-first

If a rule can answer it - a statute-cited deadline, a late-fee cap, a fee-cap lookup - Forge uses the rule. The model only fills in what a rule genuinely can't reach, like drafting a sentence or reading unstructured text.

Source-tagged

Every AI-assisted output carries where it came from - rule-based or model-generated - so you're never guessing whether you're looking at a computed fact or a drafted suggestion.

Human-approve on money & legal

Screening decisions are approve/deny by a person, never an automated adverse action. Repair costs above an owner's threshold route to the owner. AI-drafted renewal offers and compliance notices are reviewed before they ever reach a resident.

Logged, not silent

Actions AI takes that touch money or legal compliance - fair-housing review, tax filings, repair approvals - are recorded to an audit trail, not left to a chat transcript no one reviews.

Where this shows up today

  • Applicant screening surfaces AI fraud signals and a state-aware fee quote today - bureau reports (credit/criminal/eviction) arrive with our screening-provider rollout. A person approves or denies every applicant; no automated adverse action.
  • Owner repair approvals above an owner-set threshold route to the owner for a real decision, not an AI auto-approval.
  • Lease-renewal drafts ground themselves in your lease's real rent, term, and payment history, run through an advisory fair-housing check, and wait for a PM to review before anything sends to a resident.
  • Every draft falls back to a deterministic template if the model is unavailable or times out - a resident or owner never sees an error where a plain, correct answer should be.

See it on your own portfolio

30 days of full access, no credit card.

1 Buildium/NARPM 2026 Property Management Industry Report, as reported on buildium.com, read live 2026-07-12. We have not independently re-run this survey.

Common questions

Does Forge's AI ever act on money or legal matters by itself?
No. Screening decisions are human approve/deny only - the AI surfaces fraud signals, it never adverse-actions an applicant on its own. Owner repair approvals above a threshold route to the owner, not to an AI auto-approval. Lease renewal drafts, resident replies, and compliance notices are AI-drafted and human-reviewed before anything sends.
What does "deterministic-first" mean?
Wherever a rule can answer the question - a statute-cited deposit deadline, a late-fee cap, a lease-renewal template - Forge uses the rule, not a model guess. AI fills the gap the rule can't reach (drafting the actual sentence, reading a messy inbound message), and every output is tagged with whether it came from a rule or a model, so you always know which one you're looking at.
Why does this matter more than a bigger AI feature list?
Because adoption and trust aren't the same thing. Buildium and NARPM's 2026 industry report found AI use among property management firms jumped from 20% to 58% in a single year - but a feature a firm has turned on isn't a feature its team actually trusts with money or legal decisions. A longer feature list doesn't close that gap; showing your work does.