FAQs/How AI pricing works
Costs & tokens

What does AI actually cost?
Less than you think.

AI models are billed by the token — a unit of text roughly equal to ¾ of a word. Here's what that means in practice for a property management agency.

The basics

What is a token?

of a word

The word "tenant" = 2 tokens. "Hi" = 1 token.

~4

characters

"Baker Street" = 3 tokens. Numbers and punctuation count too.

~260

tokens per 200 words

A typical email or chat message runs 50–150 words.

How billing works

Every AI response has two costs: input tokens (the conversation history + instructions sent to the model) and output tokens (the response it writes back). Output costs more, but responses are usually shorter than the context sent in.

Raia adds a short system prompt (~160 tokens) to every request — this tells the model your org name, the conversation context, and how to respond. This overhead is the same whether the message is two words or two paragraphs.

Example 1

Booking a viewing appointment

A tenant enquires via WhatsApp and books a slot in two replies.

Tenant

Hi, I'd like to book a viewing for the 2-bed on Baker Street. Do you have anything this week?

~25 tokens

Raia

Hi! We have slots on Thursday 24 April at 2pm or Friday 25 April at 11am. Which works better for you?

~27 tokens

Tenant

Friday at 11am please.

~5 tokens

Raia

Confirmed! Your viewing is booked for Friday 25 April at 11am at 47 Baker Street, W1U 3PR. You'll get a confirmation SMS shortly. See you then!

~34 tokens

Token breakdown

System context (org + instructions)~160
Conversation messages~91
Total input tokens~251
Total output tokens~61

Cost for this conversation

Haiku 4.5

$0.80 in · $4 out per 1M

$0.445m

<0.01p

Sonnet 4.6

$3 in · $15 out per 1M

$0.0017

0.13p

Example 2

Logging an urgent maintenance request

A tenant reports a broken boiler. Raia raises the job and notifies all parties automatically.

Tenant

The boiler has stopped working — no hot water since yesterday morning.

~16 tokens

Raia

I'm sorry to hear that. No hot water is classified as urgent. I've raised emergency job J-1042 on your property. Our engineer will call you within 4 hours to arrange access. Is there a preferred time today?

~48 tokens

Tenant

Any time before 3pm would be great.

~9 tokens

Raia

Done. J-1042 is updated — engineer to arrive before 3pm today. If you haven't heard within 2 hours, reply here and I'll escalate. Your landlord has been notified.

~36 tokens

Token breakdown

System context (org + instructions)~160
Conversation messages~109
Total input tokens~269
Total output tokens~84

Cost for this conversation

Haiku 4.5

$0.80 in · $4 out per 1M

$0.551m

<0.01p

Sonnet 4.6

$3 in · $15 out per 1M

$0.0021

0.17p

Scale

What does a month look like?

Estimates based on a mix of conversation types. Actual usage depends on message length and how much your tenants chat.

Portfolio sizeAI conversations/monthHaiku 4.5Sonnet 4.6
20 properties80£0.03£0.12
50 properties200£0.08£0.30
100 properties400£0.16£0.60
250 properties1,000£0.40£1.49
500 properties2,000£0.80£2.99

Estimates at ~5 AI conversations per property per month. Exchange rate: $1 = £0.80.

Choosing a model

Haiku vs Sonnet — which is right?

Haiku 4.5

$0.80 in · $4 out per 1M tokens

  • Simple booking confirmations
  • Maintenance acknowledgements
  • Document chase reminders
  • FAQ-style tenant questions
  • Status updates

Sonnet 4.6

Recommended

$3 in · $15 out per 1M tokens

  • Rent negotiation conversations
  • Complaint handling & escalation
  • Tenancy renewal discussions
  • Complex compliance queries
  • Anything requiring judgment

Note on legal tasks

Legal-grade tasks — eviction assessments, deposit disputes, contract reviews — always use Opus, Raia's highest-tier model. These tasks are never available through the chat panel and are not configurable. This is a governance requirement, not a cost decision.

Ready to see it in action?

Change your model any time in Settings → AI model inside the platform.