What Is An AI Fair Use Pricing Policy & How Does It Impact MiaRec's Pricing [Incl. Price Calculator]?

Fair Use Policy (FUP)

A Fair Use Policy (FUP) is a set of guidelines a service provider defines to ensure equitable access to services. This is not to be confused with the Fair Use Doctrine, which permits using copyrighted material without permission in certain circumstances.

The Fair Use Policy encompasses various restrictions on how the service can be used, including prohibitions against misuse, abuse, and activities that violate laws or service terms. It ensures that services are used in a manner that is fair to all users by preventing any single user from excessively consuming resources, which could negatively impact other users' experience.

Example: permissible number of pieces of content, images, or AI insights generated per hour.

Fair Use Pricing Policy

On the other hand, a Fair Use Pricing Policy specifically addresses how AI services are billed based on that usage. Essentially, the AI service provider determines what a fair and reasonable use of their service is and creates a pricing model to reflect that. In other words, this policy ensures that users are charged appropriately for their consumption and that pricing remains sustainable for providers and users.

Fair Use Pricing Policy often involves tiered pricing, where different levels of service usage are priced accordingly and include limitations on the service according to the normal expected usage. Some vendors may also include measures to throttle or limit usage if it exceeds what is considered fair under the user's pricing plan.

Differences

• Focus: The Fair Use Policy focuses on equitable and lawful usage of services, ensuring that all users have fair access and preventing misuse. The Fair Use Pricing Policy, however, is more concerned with the financial aspect, ensuring that users pay according to their usage levels and that excessive use is managed through pricing adjustments or service throttling.
• Implementation: The Fair Use Policy typically includes usage guidelines, restrictions, and prohibitions, while the Fair Use Pricing Policy includes pricing tiers, usage caps, and potential overage charges.

Tokens in AI Fair Usage Pricing Policies

In the context of AI, and particularly in usage-based pricing policies, a "token" is a unit of measurement used to quantify the amount of text processed by the AI model. This can be used for prompts and text generated or text processed. 

Tokens can be thought of as pieces of words or language components. They are not always complete words but can include parts of words, spaces, and punctuation marks. One token equals approximately 4 characters (in English) or 0.75 words.

To give you better context, a one-minute conversation equals about 200 tokens. To process a 15-minute conversation using AI, you would need approximately 1,700 tokens:

• 15 minutes x 150 words/minute = 2,250 words
• 2,250 words x 0.75 words/token = 1,687.5 tokens

Depending on what you are asking the AI to do, the AI might need to take multiple passes through the transcript to answer your question. For example, we recommend asking the AI to answer no more than five questions per pass-through for an automated scorecard to maximize scoring accuracy. This means that to score a 15-minute conversation with a scorecard containing 15 questions, you will need around 5,000 tokens (5 minutes x 150 words/minute x 0.75 words/token x 3 passes) = 5,062.5 tokens.

Both input and output tokens are counted towards the total number of tokens used, which impacts the cost. For example, if a prompt uses 500 tokens and the response uses 1,000 tokens, the total usage would be 1,500. In addition, tokens will have a different price based on the AI's model complexity and performance capabilities. For example, a more advanced model like OpenAI’s GPT-4o might cost more per token than a less complex model such as GPT-3.5. 

It is also important to know that you will be charged based on actual usage. You can always see how many tokens you have used, and you can set a cap on the amount of tokens you use per month, so you are in full control at all times. 

Alright, now that the theory is out of the way, let's examine how it applies to MiaRec's pricing.

MiaRec offers a tiered pricing model that provides adequate Conversation Intelligence and Contact Center AI capabilities to contact centers as they grow (capabilities in bold are under the fair use pricing policy):

• One minute audio recording in WAV dual-channel format equals approximately ~1.92 MB based on 256 kbps at 30 fps. 
• MiaRec's Fair Use Overage Charges: $0.20 per GB/month and $0.01 per transcription minute. 
  
• Storage capacity is billed by the user but allocated to the organization. In other words, the storage capacity included in the administrator and supervisor licenses is added to the storage pool of the organization and can be used by the agents. 

Example 1: Company A Fair Use Pricing Calculation (Within Allocation)

Let's look at an example. The breakdown below clearly shows how Company A's usage stays within the limits of their fair use allocation for both storage and transcription minutes:

• Advanced Package Purchased For 25 Users In Total:
  • 21 agents
  • 1 administrator
  • 3 supervisors

• Storage Allocation:
  • Per user: 25GB
  • Total: 25 users x 25GB/user = 625GB

• Transcription Minutes Allocation:
  • Per user: 2,500 minutes
  • Total: 25 users x 2,500 minutes/user = 62,500 minutes

• Average Call Volume:
  • Per agent: 42 hours or 2,520 minutes/month
  • Total: 21 agents x 2,520 minutes/agent = 52,920 minutes/month
  • Comparison: 52,920 minutes is below the 60,000 minutes allocation

• Storage Usage Calculation:
  • Audio format: WAV dual-channel
  • Disk space usage: 1 minute = 1.92MB
  • Monthly addition: 52,920 minutes x 1.92MB/minute = 101.6GB

• Data Retention Period: 6 months
  • Total storage after 6 months: 99.84GB/month x 6 months = 609.64GB
  • Comparison: 101.6GB is below the 625GB storage allocation

Company A is within allocation and no overages are charged.

Example 2: Company B Fair Use Pricing Calculation (Over Allocation)

Let's look at a second example. Company B has the exact number and types of users as Company A and purchases the same package. However, their usage exceeds their fair use allocation for both storage and transcription minutes, resulting in a total monthly overcharge of $154.34.

For its Premium plan, MiaRec uses tokens to determine its Fair Use pricing policy. MiaRec will allocate two million tokens per Premium user per month. Additionally, every one million tokens used is billed at $5.00. Note that your team can monitor usage and manually set monthly limits for the number of tokens used.

Screenshot: Token usage tracker, found in the MiaRec platform at Administration > Speech Analytics > AI Assistant > Usage

With 2,000,000 tokens per user/month, an average organization can utilize them to automatically score all user calls and generate 3-4 AI insights. These insights include call classification/call disposition and extracting crucial information such as products discussed, next actions for agents, and pricing details. 

To make it more tangible, let's examine two examples that demonstrate within and over-allocation.

Example 3: Company C Fair Use Pricing Calculation (Within Allocation With Tokens)

Companies C and D purchase the same number of user licenses on the Premium level.

Company C records about 1,800 minutes per agent per month, excluding all calls under 120 seconds from auto call scoring and AI Insights. We recommend you do this to maximize the accuracy of the AI-generated results because there isn't enough context for the AI to evaluate the agent correctly or extract good enough insights. In this example, these calls comprise about 10% of its total call volume. 

• Premium Package Purchased For 100 Users In Total:
  • 2 administrators
  • 8 supervisors
  • 90 agents
  • Total Tokens Included In License: 2,000,000 x 100 agents = 200,000,000

• Average Call Volume:
  • Per User: 1,620 minutes/month (1,800 total call recording minutes minus 10% (180 minutes per user) of the calls from Auto QA and AI Insights due to too short call length)
  • Total Call Volume: 100 users x 1,620 minutes/user = 162,000 minutes/month

• Token Calculation for Call Classification:
  • Rate: 200 tokens per minute
  • Total Tokens: 162,000 minutes x 200 tokens/minute = 32,400,000 tokens/month

• Token Calculation for Auto Scorecard (15 questions, 3 passes per scored call):
  • Tokens per Minute: 200 tokens/minute x 3 passes = 600 tokens/minute
  • Total Tokens: 162,000 minutes x 600 tokens/minute = 97,200,000 tokens/month

• Token Calculation for Extracting 2 AI Insights (2 passes):
  • Rate: 200 tokens/minute x 2 = 400 tokens/minute
  • Total Tokens: 162,000 minutes x 400 tokens/minute = 64,800,000 tokens/month

• Total Token Usage:

  • Call Classification: 46,800,000 tokens/month
  • Auto Scorecard (10 questions, 2 passes):: 93,600,000 tokens/month
  • AI Insights (2 per minute): 93,600,000 tokens/month
  • Grand Total of Tokens Used: 32,400,000 + 97,200,000 + 64,800,000 = 194,400,000 tokens/month
  • Overage: Within Allocation, no overage.

Company C has scored all relevant calls on a 15-question scorecard, extracted 2 AI Insights, and categorized the call without going over the allocated tokens. There will be no token overage charges.

Example 4: Company D Fair Use Pricing Calculation (Over Allocation With Tokens)

Company D, on the other hand, auto scores 100% of its calls, and its call volume per user and month is higher than that of Company C. It has a scorecard with 10 questions, categorizes each call, and extracts two AI Insights per call.

• Premium Package Purchased For 100 Users In Total:
  • 2 administrators
  • 8 supervisors
  • 90 agents
  • Total Tokens Included In License: 2,000,000 x 100 agents = 200,000,000

• Average Call Volume:
  • Per User: 2,600 minutes/month
  • Total Call Volume: 100 users x 2,600 minutes/user = 234,000 minutes/month

• Token Calculation for Call Classification:
  • Rate: 200 tokens per minute
  • Total Tokens: 234,000 minutes x 200 tokens/minute = 46,800,000 tokens/month

• Token Calculation for Auto Scorecard (10 questions, 2 passes):
  • Rate: 2 passes per call
  • Tokens per Minute: 200 tokens/minute x 2 passes = 400 tokens/minute
  • Total Tokens: 234,000 minutes x 400 tokens/minute = 93,600,000 tokens/month

• Token Calculation for Extracting 2 AI Insights (2 passes):
  • Rate: 200 tokens/minute x 2 = 400 tokens/minute
  • Total Tokens: 234,000 minutes x 400 tokens/minute = 93,600,000 tokens/month

• Total Token Usage:

  • Call Classification: 46,800,000 tokens/month
  • Auto Scorecard (10 questions, 2 passes): 93,600,000 tokens/month
  • AI Insights (2 per minute): 93,600,000 tokens/month
  • Grand Total of Tokens Used: 46,800,000 + 93,600,000 + 93,600,000 = 234,000,000 tokens/month
  • Overage: 234,000,000 tokens/month used - 200,000,000 tokens/month included = 34,000,000 tokens/month over (= 24 million x $5,00 = $170/month overage cost)

Company D would require 34 million tokens per month per organization in addition to the 200 million allotted, resulting in a small overcharge of $170/month. Please note, this is in total, NOT per user!