What is the best AI for a real estate brokerage or developer?

Sonzai is an AI operating system for real estate. It runs the operational work that usually depends on someone remembering: lead generation, qualification, concierge replies, sales follow-up, collections, and segmentation. Everything runs on one record that learns the business and gets sharper with every deal. See https://sonz.ai/real-estate.

What AI can run follow-up and renewals for an insurance agency?

Sonzai runs an insurance book: it follows up on quotes that went quiet, chases renewals early, answers claims questions, and remembers every client. It works for agencies and for independent, commission-based agents who want their own book run for them. See https://sonz.ai/insurance.

Won't a smarter AI model make a tool like this obsolete?

No. Sonzai is model-agnostic, so a better model just does better work over the same record. The value lives in your private record, which a smarter model can't reach on its own. Intelligence isn't the bottleneck; permission is.

How do I give an AI agent memory that persists across sessions?

Build on the Sonzai Mind Layer: persistent memory, relationships, knowledge, and learning through one API, with any model underneath. The agent reads the record before it replies and writes back what it learns. See https://sonz.ai/mind-layer.

Does Sonzai replace our CRM or property system?

No. Sonzai works on top of the systems you already run (Salesforce, Yardi, ERP). It reads from them, reconciles across them, and writes back. It adds the operational layer and the cross-system record, it doesn't replace your systems.

Which AI companies in Southeast Asia deploy on-site instead of just selling software?

Sonzai is forward-deployed. Our engineers deploy alongside your team from Singapore and Manila and configure a platform that already works to your operation. We configure reusable modules rather than writing a bespoke project, so it stays software economics, not a consulting retainer.

Is there AI priced on outcomes instead of seats or hours?

Yes. Sonzai sells outcomes, not seats or hours. We start with the one job that leaks the most, deploy an AI employee to own it, and price toward the result it produces once the loop is de-risked.

Who has shipped enterprise gen-AI at scale in Singapore?

The Sonzai team has, including more than $70M of AI value at DBS Bank and OCBC's first gen-AI rollout. Sonzai Labs is based in Singapore, active in Manila, and is an EDG-supported deployment partner.

M L

Record a realized outcome — improves the scoring model + the bandit in one call

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

action_features?

action_id?string

context?

converted*boolean

features?

note?string

predicted_score?integer

Formatint64

propensity?number

Formatdouble

reward?number

Formatdouble

subject_id?string

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/feedback" \  -H "Content-Type: application/json" \  -d '{    "converted": true  }'

{
  "$schema": "/api/v1/schemas/FeedbackResponse.json",
  "bandit_error": "string",
  "bandit_n": 0,
  "bandit_updated": true,
  "converted": true,
  "message": "string",
  "ok": true,
  "outcome_recorded": true,
  "use_case": "string"
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

Next-best-action over a per-tenant variable action set

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

actions*array<>|null

context?

explore?boolean

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/nba/decide" \  -H "Content-Type: application/json" \  -d '{    "actions": [      {        "id": "string"      }    ]  }'

{
  "$schema": "/api/v1/schemas/DecideResponse.json",
  "action_id": "string",
  "explore": true,
  "model_n": 0,
  "propensity": 0.1,
  "scores": [
    {
      "action_id": "string",
      "propensity": 0.1,
      "score": 0.1
    }
  ]
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

Teach the bandit a realized (action, propensity, reward)

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

action_features?

action_id*string

context?

propensity?number

Formatdouble

reward*number

Formatdouble

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/nba/learn" \  -H "Content-Type: application/json" \  -d '{    "action_id": "string",    "reward": 0.1  }'

{
  "$schema": "/api/v1/schemas/LearnResponse.json",
  "n": 0,
  "ok": true
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

Off-policy evaluation (IPS / SNIPS / doubly-robust + CI)

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

logged*array<>|null

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/ope/evaluate" \  -H "Content-Type: application/json" \  -d '{    "logged": [      {        "action_id": "string",        "propensity": 0.1,        "reward": 0.1      }    ]  }'

{
  "$schema": "/api/v1/schemas/OPEResponse.json",
  "ci_high": 0.1,
  "ci_low": 0.1,
  "dr": 0.1,
  "ess": 0.1,
  "estimator_ci": "string",
  "ips": 0.1,
  "n": 0,
  "snips": 0.1
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

Score a record with the per-tenant model (falls back to the global prior)

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

features*

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/scoring/predict" \  -H "Content-Type: application/json" \  -d '{    "features": {      "property1": null,      "property2": null    }  }'

{
  "$schema": "/api/v1/schemas/PredictResponse.json",
  "calibration_method": "string",
  "model_version": 0,
  "raw": 0.1,
  "score": 0.1,
  "served_from": "string"
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

Train the per-tenant scoring model (CatBoost + Optuna, calibrated)

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

optimize_budget?integer

Formatint64

rows*array<>|null

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/scoring/train" \  -H "Content-Type: application/json" \  -d '{    "rows": [      {        "features": {          "property1": null,          "property2": null        },        "label": 0      }    ]  }'

{
  "$schema": "/api/v1/schemas/TrainResponse.json",
  "auc": 0.1,
  "best_params": {
    "property1": null,
    "property2": null
  },
  "brier": 0.1,
  "brier_baseline": 0.1,
  "brier_uncalibrated": 0.1,
  "calibration_method": "string",
  "ece": 0.1,
  "importances": [
    {
      "gain": 0.1,
      "name": "string"
    }
  ],
  "logloss": 0.1,
  "model_version": 0,
  "n": 0,
  "trials": 0
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

Single-call self-learning simulation (train → decide → learn → OPE over rounds)

Authorization

AuthorizationBearer <token>

API key issued to a Sonzai project. Format: Bearer sk_....

In: header

Path Parameters

use_case*string

Request Body

application/json

TypeScript Definitions

Use the request body type in TypeScript.

rounds?integer

Formatint64

scenario?string

seed?integer

Formatint64

Response Body

`application/json`

`application/problem+json`

curl -X POST "https://loading/api/v1/builtin-agents/ml/string/simulate-rounds" \  -H "Content-Type: application/json" \  -d '{}'

{
  "$schema": "/api/v1/schemas/SimulateRoundsResponse.json",
  "action_labels": {
    "property1": "string",
    "property2": "string"
  },
  "model": null,
  "ope": {
    "ci_high": 0.1,
    "ci_low": 0.1,
    "dr": 0.1
  },
  "policy": null,
  "scenario": "string",
  "series": [
    {
      "auc": 0.1,
      "ci_high": 0.1,
      "ci_low": 0.1,
      "n": 0,
      "nba_reward": 0.1,
      "nba_value": 0.1,
      "ope_dr": 0.1,
      "round": 0
    }
  ]
}

{
  "$schema": "/api/v1/schemas/ErrorModel.json",
  "detail": "Property foo is required but is missing.",
  "errors": [
    {
      "location": "string",
      "message": "string",
      "value": null
    }
  ],
  "instance": "https://example.com/error-log/abc123",
  "status": 400,
  "title": "Bad Request",
  "type": "https://example.com/errors/example"
}

M L

200application/json

defaultapplication/problem+json

200application/json

defaultapplication/problem+json

200application/json

defaultapplication/problem+json

200application/json

defaultapplication/problem+json

200application/json

defaultapplication/problem+json

200application/json

defaultapplication/problem+json

200application/json

defaultapplication/problem+json

`application/json`

`application/problem+json`

`application/json`

`application/problem+json`

`application/json`

`application/problem+json`

`application/json`

`application/problem+json`

`application/json`

`application/problem+json`

`application/json`

`application/problem+json`

`application/json`

`application/problem+json`