AI TCO & Usage Configuration Guide

Overview

The AI TCO & Usage solution is designed to solve key business problems and provide valuable insights.

It supports the following use cases

Prerequisites

The prerequisites for the AI TCO & Usage solution are:

IBM Apptio Costing Standard license
IBM Apptio Server Version: R12.11.14 (or higher)
Components Version v120 in Project Settings

Configuration Steps

The following overview outlines the configuration steps required to achieve AI TCO & Usage reporting, with detailed documentation provided in subsequent sections.:

Project Architecture - See the architecture section here
Component Install - See the component install section here
Bill vs. General Ledger based. These 2 options are described at a high level in the architecture section here
Load AI related data sources - See the data requirements section here
Light up new Model Metrics - See the model section here
AI TCO & Usage Reporting - See the reporting section here

Component Install

The core of the AI TCO & Usage solution is introduced via 4 components. 2 new components, specifically created for this solution + 2 existing components that have been extended and hence are expected to get upgraded. For new customers, the other "typical" components of Labor, Vendor, Cloud etc. would need to be installed, as per the overall intended architecture. Existing customers can re-use these previously installed components without any changes. Before installing these components, it is essential to carefully review the architecture of your environment to determine the best approach for implementation. You will need to decide whether to:

Install the components in an existing project, or
Create a new project specifically for the AI TCO & Usage solution

Following is the overview of the 4 components to consider and install in prioritized order.

Note: To install required components, temporarily set Components Version to Version 120 in Project Settings. Install the necessary components, then revert the version change to avoid triggering unnecessary upgrade indicators.

A green square with a head with a headset

AI TCO & Usage component: This component provides clear view into AI expenditures and utilization across AI models, AI solutions, and business units. This component establishes the underlying framework through a set of new, AI specific master datasets, workbenches, models, and metrics, facilitating the modelling and calculation of AI TCO and usage outcomes.
Install as the first component.
CTF – Cost Source component: A new version of the CTF - Cost Source component is available, which adds the following new fields to the Cost Source Master Data to support the AI TCO & Usage solution.:
1. AI Cost Percentage: Brings in the AI Cost Percentage as set and defined by the user in the AI Tablematch table.
2. AI Amount: Multiplies the existing Cost Source amount with the AI Cost Percentage to create a new AI Amount that looks to feed the new AI Cost model. To be installed as second component. If you do not install this component, you will not see the "% of Total IT Cost YTD" metric or the "AI Cost by Cost Pool" breakdown in your reports.
  
  Note: For existing customers (that intend to populate this solution using the Cost Source – General Ledger) it’s required to upgrade the following component:
CT Apps – Services component: A new version of the CT Apps - Services component is available, which adds new fields to All Business Services to support the AI TCO & Usage solution.

Note: Existing customers must upgrade the following component to ensure compatibility and access to new features.
1. IsAISolution: Needs to be set to “Yes” for any Service Offering that is considered directly or indirectly an AI Solution.
2. AI Solution Type: Metadata that is used as a slicer in the AI TCO & Usage solution. Suggested values are:
  1. Standalone: A dedicated AI solution that operates independently as its own product or service.
  2. Embedded: AI built deeply into an existing product, becoming a core native feature.
  3. Augmented: AI bolted on or lightly added to improve a feature/process, but not essential.
3. AI Solution Users: Sums up the nr. of unique users of AI Solutions.
4. AI Solution Count: Counts the nr. of AI Solutions.
To be installed as 3rd component. Note: Not upgrading this component will break the AI TCO & Usage reporting, since most of the reporting is created at the level of AI Solutions = Business Services with the filter set on IsAISolution = “Yes”.
AI TCO & Usage Reportingcomponent: This component delivers the reporting for the AI TCO & Usage solution. It consists of 2 reports, introduced in a new AI report collection. The main AI TCO & Usage report enables organizations to monitor the full lifecycle of AI investments, proactively addressing AI sprawl by tracking Total Cost of Ownership (TCO), usage trends, breakdowns, and anomalies. The main target personas are C-suite executives, Business & Solution Leaders, and AI & Data Science teams. In addition, there’s a secondary AI Cost Model report which allows users to dynamically trace the AI Cost & Usage allocations through interactive Sankey diagram visualizations.
To be installed as the final component.

Architecture

The first step is to decide where to install the components for the AI TCO & Usage solution. You have multiple options, and we've outlined two possible approaches below.

Option A

Integrate the AI TCO & Usage solution into an existing project and opt for a GL driven and Resource Tower approach.

This approach is recommended for existing customers looking to leverage/re-use allocation logic that already exist in the main Cost model metric. If you think your target users would benefit from a General Ledger-driven approach with detailed views of Resource Towers and Cost Pools, this solution is a good fit. The diagram below shows how to integrate the new AI Platforms object into your standard model.

A diagram of a business

AI Cost model metric blueprint – GL Based with Resource Towers

The AI TCO & Usage solution introduces new metrics that work independently, even when added to your existing project. The new AI Cost model is highlighted in the diagram, with blue showing existing parts and green showing new additions, such as AI Platforms and new fields in Cost Source and Business Services. This allows you to use your existing allocation lines with the new AI Cost model on an as-needed basis, giving you more flexibility and control.

The driver of the Cost Source object in the AI Cost model metric relies on a new column called "AI amount", which is populated from the "Cost Source AI Tablematch" dataset. This dataset is designed to identify General Ledger transactions related to Artificial Intelligence (AI).

It is essential to identify and populate the table with relevant accounts, cost centers, vendors, and journal line descriptions that you associate with AI Costs & Usage, as demonstrated in the sample above. Additionally, the AI Cost % column is required and must contain numeric values.

Note: In case of doubt regarding the AI Cost %, set it to 100. The allocation rules and mappings will further refine the transaction amounts to only include AI-related costs.

The Table Match Logic will retrieve additional column details from the Cost Source data, matching entries from the uploaded extract and populating relevant columns, as long as they exist in the Cost Source data.

Option B

Create a new project for the AI TCO & Usage solution + opt for a Bill driven approach, excluding Resource Towers.

Note: Opting for the Bill driven approach and excluding Resource Towers at this point would be considered a custom approach. Technically feasible but likely to require support from Apptio Professional Services.

This approach is recommended for existing or new customers who first want to completely reference this new solution and set of use cases related to AI. Also, it is suited for customers that believe that the target personas for this solution would benefit from a Bill driven source of truth with more direct allocation lines and without a need to report on cost breakdowns of Resource Towers and Cost Pools. The below picture gives the blueprint on the architecture .

A diagram of a company

AI Cost model metric blueprint – Bill Based without Resource Towers

The above framework steps away from the more traditional General Ledger – Cost Source driven approach while also removing Resource Towers from the architecture. Where this might introduce faster time-to-deploy it would rely on more custom allocations to be built. If adopting this approach, the following out-of-the-box reporting views will no longer work:

KPI of “% of Total IT Cost YTD”
Breakdown of AI Cost by “Cost Pool”
Breakdown of AI Cost by “Resource Towers”

Crawl-Walk-Run

As you get started with the AI TCO & Usage solution, you don't need to have everything set up from the beginning to start seeing value. The solution is designed to be flexible, allowing you to start small and build up to a full understanding of your AI costs over time. This means you can begin with the basics and add more details as you go, without needing to have all the pieces in place right away. You can think of it as a Crawl-Walk-Run approach, where you start by taking small steps, then gradually increase your pace as you become more comfortable with the solution. This approach lets you deliver value quickly and build momentum over time, making it easier to get the most out of your AI investments.

AI TCO & Usage – Crawl/Walk/Run

Crawl

In the crawl phase certain AI Solutions can get configured relatively quickly by only having to bring in datasets related to certain areas. Focus here is on AI Solutions predominantly being purchased.

Sample AI Solution: Microsoft Copilot
Main model objects:
- Vendor: Vendor License
- Labor: Training & Change Management
- Other: Security & Integration
- Service Allocations Direct: Consumption/Usage file
Value: Understand the TCO of Microsoft Copilot as an AI Solution, including views into Labor contributions. Report and track the adoption of Microsoft Copilot across the organization by bringing in and leveraging the associated consumption/usage data.

Walk

In the walk phase the focus would shift towards AI Solutions that start to bring in the usage of AI platforms and/or AI models, underpinning the TCO; alongside further advanced usages of skilled labour. These AI Solutions would typically be a combination of buy & build.

Sample AI Solution: AskIT
Main model objects (in addition to the ones from the crawl phase):
- Cloud: AI Model cost & usage
- AI Platforms: Data representing the actual AI Platforms & AI Models
- Data: In case of proprietary data, ensure to capture and represent cost related to Data preparation and training
Value: Understand the TCO of an internal built AI Solution such as AskIT by starting to bring in data related to the AI platforms and/or AI models the solution is consuming. Report and track the cost drivers as well as obtaining insights into the AI model usage.

Run

In the most advanced run phase, it’s expected that most, if not all, typical modelled objects would become applicable to light up the AI TCO & Usage outcomes & insights. The AI Solutions represented in this phase are expected to be mostly in-house built AI solutions.

Sample AI Solution: IndexGPT
Main model objects (in addition to the ones from the crawl & run phase):
- Infrastructure: Dedicated AI Compute (GPU), Storage, Database etc.
- Time Tracking: More consumptive way to ensure proper activity tracking is in place for high-end labour activities
Value: Obtain end-to-end insights into the AI TCO & Usage across all your AI solutions, including the most complex ones. Some of which are housed and built internally and therefore contain a significant infrastructure footprint that needs to be accounted for.

Data Requirements

Core Data

As shown by the section focused on the Architecture, the AI TCO & Usage solution introduces a new modelled object called AI Platforms. This object is intended to be sourced with data focused on the cost, usage and adoption associated with AI platforms and AI models.

The embedded file focuses on:

The data required (listing all the columns with their respective description, whether they’re required or not and the effect if missing)
Column clarifications (describes some sample values and their description)
Sample data (an example of what the data would look like in AI Platforms)
Source systems (sample references of potential source systems for the AI Platform data)

In addition to the main new object of AI Platforms, 2 master datasets have been updated to reflect some of the necessary changes required for lighting up the main AI TCO & Usage reports.

Several new columns have been added to Cost Source Master Data and All Business Services, which are documented in the same, embedded file.

Note: Data required to populate the other modelled objects that are part of this solution’s blueprint are not covered in here. Please reach out to your Apptio representative to obtain insights into the other, traditional data requirements.

For the data requirement details highlighted above, click here:

Workbench

As data gets sourced for AI Platforms, it goes through the following architecture:

As one can see from the above, it’s expected that your raw data gets appended into AI Platforms Feed firstly. From there, it gets exposed into 2 editable tables (“AI Platforms Metadata Enriched” and “AI Platforms Mapping Enriched) which make up the AI Workbench (which gets installed as part of the AI TCO & Usage component). The AI Platforms Master dataset ultimately gets used to expose the data into the AI model metrics via the AI Platforms modelled object. Please find a brief description for each of the 2 editable tables below:

AI Platforms Data Enrichment
The table “AI Platforms Data Enrichment” in the Workbench enables the enrichment of metadata, used for reporting, across the AI Platforms and AI Models. Any updates in this table get directly reflected in the “AI Platforms Metadata Enriched” editable table, enhancing the data related to AI platforms to support accurate cost tracking and reporting. The Enhanced data is further mapped to the “AI Platforms” passing through the “AI Platforms Mapping ET Transform”.
AI Platforms Data Mapping
This table in the workbench enables the mapping of the consumers of the AI Platforms and AI Models. It expects the mapping of an AI Platform or AI Model to 1 or many Solution Offerings. Alternatively, it can be mapped to a Business Unit or User ID in case the AI Platforms or AI Models are directly being consumed by end users. The Cost Weighting column is important since it allows for a weighted cost allocation/distribution across the consumers. Any updates in this table get directly reflected in the “AI Platforms Mapping Enrichment” editable table, which gets further mapped to the AI Platforms via the” AI Platforms Mapping ET Transform”.
AI Token Type Normalization
This editable table enables the categorization and normalization of different token types (coming in through the raw data) into the Simplified Token Types of Input and Output. This categorization helps with simplified analysis and reporting views as it relates to exposing AI Usage details. Any updates in this table get directly reflected in the “AI Token Type Normalization” editable table and onto the “AI Token Type Normalization ET Transform” table. The latter gets looked up into the AI Platforms Master table via a formula.

Models

Once the data requirements are clear it is important to understand how this data is used in the different model metrics, especially in terms of lighting up the end user reports.

The AI TCO & Usage component installs 13 model metrics, which may seem extensive, but actually provides maximum flexibility for modeling and allocations, as well as enables easy reporting views. As previously outlined in the architecture section, which introduced the new AI Cost model and the two possible approaches, this section will delve into the details of all 13 models, explaining their purpose and rationale. The models are categorized into three groups:

5 AI Cost-related models
5 AI Usage-related models
3 AI Count-related models

The following section provides a detailed summary of the modelled metrics, along with their corresponding driver logic

Modelled Metric	Modelled Object Source	Driver Logic
AI Cost	Cost	Formula: If(Cost Source Master Data.Cost Source Model Driver IN (“OPEX – ACTUALS – FIXED”, “OPEX – ACTUALS – VARIABLE”), Cost Source Master Data. AI Amount),0) Note: Filtered for AI expenditure, via AI Tablematch
AI Cloud Cost	Cloud Service Provider	Metric: AI Cost
AI Vendor Cost	Vendors	Formula: If(Is CSP!="Yes", AI Cost, 0)
AI Labor Cost	Labors	Formula: AI Cost - AI Vendor Cost
AI Other Cost	Other Cost Pools	Metric: AI Cost
AI Usage	AI Platforms	Column: "Usage Quantity" from AI Platforms
AI Token Consumption	AI Platforms	Formula: If(AI Platforms.Billing Type="Token Usage",AI Platforms.Usage Quantity,0)
AI Input Tokens	AI Platforms	Formula: If(AI Platforms. Normalized Token Type = "Input", AI Usage, 0)
AI Output Tokens	AI Platforms	Formula: If (AI Platforms. Normalized Token Type = "Output", AI Usage, 0)
AI Solution Users	AI Platforms	Column: "AI Solution Total Users" from AI Platforms
	Business Services	Formula: If({AI Solution Users to Business Services (AI Platforms- AI Solution Users driver) }=0,{All Business Services.AI Solution Total Users Prep},0)
AI Count
AI Vendor Count	Vendors	Formula: If({AI Cost}!=0,(1/SumIF(Vendor Master Data.Vendor ID,Vendor Master Data.Vendor ID,1)),0)
AI Model Count	AI Platforms	Column: "Num of AI Models" from AI Platforms
AI Solution Count	AI Platforms	Column: " AI Solution Count" from AI Platforms
AI Solution Count	Business Services	Formula: If(AI Soution Count to Business Services(AI Platforms-AI Solution Count Driver))!=0,0,{All Business Services.AI Solution Count}

AI Cost models (5)

AI Cost

Main cost model metric that powers all the AI TCO & Usage reporting. In the view below (approach driven by the Cost Source - General Ledger), it’s expected that AI Expenditures are identified (at the Cost Source level), existing allocations (from the Cost model) are re-used by tagging them to this new AI Cost model metric and new allocations are established as it relates to allocations in & out from the new AI Platforms object.

This metric was created keeping the following in mind:

Not to disturb the existing cost model.
Allowing for a linkage to be made between the IT Towers and the AI Platforms object (containing the AI expenditures)
Allowing for any extra drivers to be set (related to AI costs specifically) on an as-need basis by the customer (who wishes to further expand this modelled metric)

AI Cloud Cost

This metric has been created to allow the end user to quickly and easily understand the AI Cloud Cost driver on a monthly and year to date basis, as part of the TCO breakdown analysis. It simply looks to mimic the main AI Cost model, focusing and starting from the Cloud Service Provider object. Tag the upwards allocation lines that should be considered part of the AI Cloud Cost.

AI Vendor Cost

This metric has been created to allow the end user to quickly and easily understand the AI Vendor Cost driver on a monthly and year to date basis, as part of the TCO breakdown analysis. It simply looks to mimic the main AI Cost model, focusing and starting from the Vendor object. Note: It removes the Cloud Service Providers from this metric to ensure no double counting happens with the AI Cloud Cost metric.

Tag the upwards allocation lines that should be considered part of the AI Vendor Cost.

AI Labor Cost

This metric has been created to allow the end user to quickly and easily understand the AI Labor Cost driver on a monthly and year to date basis, as part of the TCO breakdown analysis. It simply looks to mimic the main AI Cost model, focusing and starting from the Labor object. Note: It removes the AI Vendor Cost Service from this metric to ensure no double counting happens with the AI Vendor and Cloud Cost metrics.

Tag the upwards allocation lines that should be considered part of the AI Labor Cost.

AI Other Cost

This metric has been created to allow the end user to quickly and easily understand any Other AI Cost drivers on a monthly and year to date basis, as part of the TCO breakdown analysis. It simply looks to mimic the main AI Cost model, focusing and starting from the Other Cost Pools object.

Tag the upwards allocation lines that should be considered part of the AI Other Cost.

AI Usage models (5)

AI Usage

Together with the AI cost model metric, this AI Usage metric is the most important metric, since it also powers all the AI TCO & Usage reporting. It is however a more simplified metric, given it originates immediately at the AI Platform modelled object and therefore traverses less layers. The driver for this modelled metric comes directly from the Usage Quantity column from the AI Platforms master dataset. It points to the usage of the AI Platforms and/or AI Models. If the relationship data is in place, it’s expected to link onwards to Business Services (based on AI Solution = Service Offering) and it’s expected for the data to get re-used in Service Allocations Direct to potentially highlight any direct AI Platforms and/or AI Model consumption by the Business Units (based on User ID belonging to a certain Business Unit).

AI Token Consumption

This metric has been created as a subset of AI Usage, focusing specifically on where the Billing Type = Token Usage. It is used on the reporting as a Top KPI and for various Token Consumption views. While we only highlight the driver in the below view, it’s expected to follow the same allocation logic/lines as the AI Usage model metric.

AI Input Tokens

This metric has been created as a subset of AI Usage and a further subset of AI Token Consumption, since it specifically focuses on AI Input Tokens. It predominantly is used on the Usage Detail tabs in the reporting. Note: It has a dependency on the AI Token Normalization table, where different Token Types look to get normalized into “Input” in this instance. It’s expected to follow the same allocation logic/lines as the AI Token Consumption model metric.

AI Output Tokens

This metric has been created as a subset of AI Usage and a further subset of AI Token Consumption, since it specifically focuses on AI Output Tokens. It predominantly is used on the Usage Detail tabs in the reporting. Note: It has a dependency on the AI Token Normalization table, where different Token Types look to get normalized into “Output” in this instance. It’s expected to follow the same allocation logic/lines as the AI Token Consumption model metric.

AI Solution Users

This metric points to a different type of AI Usage, namely AI Adoption. It looks to report on the number of users that are using/adopting the AI Solutions. It’s a key reporting KPI and instrumental to the insights shown on the Business Units tab in the reporting. To ensure this metric works at all levels, including slicing on AI Model related data such as AI Model Type, it gets sourced at 2 levels. First at the AI Platform modelled object (facilitated through a lookup in AI Platforms master), in a second instance at the Business Services modelled object. The driver for the latter ensures no double counting takes place by subtracting the former driver, set at the AI Platform modelled object. Tag the upwards allocation lines to ensure the metric works when slicers from Business Services or Business Units are being applied.

AI Count models (3)

AI Vendor Count

This metric has been created to ensure accurate reporting can happen on the number of distinct Vendors that are in play as it relates to AI Costs. It is shown as a Top KPI on the reporting. It only picks up Vendors with an associated AI Cost to ensure not all Vendors are represented. Tag the upwards allocation lines to ensure the metric works when slicers from Business Services or Business Units are being applied.

AI Model Count

This metric has been created to ensure accurate reporting can happen on the number of distinct AI Models that are in play. It is shown as a Top KPI on the reporting. It is sourced directly from a formula created in the AI Platforms Master dataset. Tag the upwards allocation lines to ensure the metric works when slicers from Business Services or Business Units are being applied.

AI Solution Count

This metric has been created to ensure accurate reporting can happen on the number of distinct AI Solutions that are in play. It is shown as a Top KPI on the reporting. To ensure this metric works at all levels, including slicing on AI Model related data such as AI Model Type, it gets sourced at 2 levels. First at the AI Platform modelled object (facilitated through a lookup in AI Platforms master), in a second instance at the Business Services modelled object. The driver for the latter ensures no double counting takes place by subtracting the former driver, set at the AI Platform modelled object. Tag the upwards allocation lines to ensure the metric works when slicers from Business Services or Business Units are being applied.

Reporting

The AI TCO & Usage Reporting component installs the reporting features for AI TCO & Usage solution. Two reports are installed in a new AI report collection.

The main AI TCO & Usage report enables organizations to monitor the full lifecycle of AI investments, proactively addressing AI sprawl by tracking Total Cost of Ownership (TCO), usage trends, breakdowns, and anomalies. The report is designed for C-suite executives, Business & Solution Leaders, and AI & Data Science teams and features 4 tabs with insights aligned for each of these target personas:

Summary – C-suite executives
AI Models - AI & Data Science teams
AI Solutions – Solution Leaders (Service Owners)
Business Units – Business Leaders

Additionally, a secondary AI Cost Model report is available, enabling users to dynamically track AI Cost & Usage allocations through interactive Sankey diagram visualizations.

For visualizations and key benefits delivered with the above reports, click here.

Call Outs

The components AI TCO & Usage and AI TCO & Usage Reporting are released with Component Template v120. Therefore, the Template Settings must be switched to view and download these two components.
“isAISolution” is a crucial new column, created in the All Business Services table (via the upgrade to the respective component). Ensure the value is set to “Yes” for the Service Offerings that you identify AI Solutions. This is a pre-requisite to activate the AI TCO & Usage Reporting.
Existing customers must upgrade the "CTF – Cost Source" component, if using the General Ledger-based method, and the "CT – App Services" component, as outlined above.