AI Management Capability Matrix: AI-Ready Edition

The AI Management Capability Matrix describes the work of running AI across an organization: funding uses, approving agents, setting their limits, monitoring them and retiring them. AI offices and governance teams at banks, insurers, asset and wealth managers and payments firms use it to give that work owners, KPIs and a budget line. It covers 12 Level 1 areas, 61 Level 2 capabilities and 298 Level 3 capabilities with definitions to Level 3 and 239 KPIs. Each Level 2 capability has an AI profile stating where AI can help the people who govern AI. The content is cross-industry.

What it is

Inside the AI Management Capability Matrix

AI management is treated here as a business function with its own work, separate from building models. The matrix sets that work out in 12 Level 1 areas, 61 Level 2 capabilities and 298 Level 3 capabilities, with a definition for each and 239 KPIs at Level 2. Ownership of a use, the limits on an agent, the review of a change and the retirement of an agent each have a place in it.

No vendor or tool appears in the structure. Agent authority management is AM-7.2 whichever platform the agents run on, so the matrix holds when the technology underneath changes.

The AI ratings apply to the AI office's own work. Each Level 2 profile states where AI can prepare what the governance team reviews, in what mode and with what oversight, along with agents, data, prerequisites, risks and role changes. Each Level 3 carries a work pattern, mode, AI fit and risk tier. For AM-7.2 the sample rates oversight as Human decides: an assistant drafts an agent's written authority limits, and the Head of AI Governance sets them.

The deliverables

What the Download Contains

The download contains the matrix and the AI ratings, with the tools to rank and present them:

The model

• 12 L1, 61 L2 and 298 L3 AI management capabilities, defined through L3, with 239 KPIs set at L2

AI and governance

• For each L2 capability, an AI profile covering opportunities, what stays human, agents, data, KPIs, AI measures, prerequisites, risks and workforce impact
• L3 capabilities rated for work pattern, target AI mode, AI fit and risk tier
• Risk-adjusted scoring and ranking of each L2 capability, driven by your own ratings and weights
• An AI heatmap across the entire model
• A weighted AI readiness assessment with 40 questions

Files and terms

• To begin, a Welcome document; then the capability map deck, standalone capability definitions and KPIs documents, the Excel workbook, the user guide
• Bonus files included: an AI management glossary, A Practical Guide to Business Architecture, a capability modeling overview, an AI-ready capability profile template and capability relationship mapping templates
• Corrections free of charge, and new versions free for 180 days after purchase

Your organization's name and license ID are printed in every file. Ratings reflect a typical large enterprise; treat them as a first draft of your own.

Who It Is For

Chief AI officer. Defines the AI office's mandate and budget by capability and decides where AI supports its own work first.

Head of AI governance. Uses the profiles to fix which decisions stay with a named person, starting with agent authority limits.

Model risk and validation leads. Compare their current review scope with the capabilities and note what agents add to it.

Enterprise architects. Link the AI management capabilities to the rest of the firm's architecture and keep the IDs consistent.

Security and privacy partners. Find the capabilities where their review is required and confirm the oversight levels for each.

The Problem Behind It

Many financial firms already review statistical models before use. That review covers the model, and it does not always cover what an agent built on the model may read, change or send. Agents are approved in one forum, configured by another team and watched by a third, and nobody holds a list of the work that managing them involves.

The matrix is that list. It gives AI management the footing finance and security already have: defined work, owners and measures. The profiles add a practical point for a governance team short of time. They show which of its own tasks AI can prepare safely and which decisions must stay with a named person.

Putting it to work

Use Cases

Setting the scope of an AI office. Walk the 12 Level 1 areas with the executive sponsor and mark which ones the AI office owns, which it shares and which stay elsewhere. The result is a written mandate tied to capability IDs.

Dividing duties among AI, risk and technology teams. Assign an owner to each Level 2 capability. Capabilities claimed by two teams, or by none, are settled before the first agent goes live.

Writing authority limits for agents. Use the AM-7.2 profile as the pattern for deciding what each agent may read, propose, change and send. The profile names the role that sets the limits and the roles it hears first.

Choosing where AI helps the governance team. Score the 61 Level 2 capabilities with your own weights. The ranking shows which review and monitoring tasks AI can prepare first at an acceptable risk tier.

Evaluating AI governance tools. List the Level 3 capabilities a governance or monitoring product should support. Products are then compared against your definitions.

Measuring readiness across the firm. Run the 40-question assessment with the AI office, model risk, security, privacy and business owners. Repeating it later shows whether gaps in data, skills or oversight are closing.

Getting Started

An AI office can go from the files to an agreed mandate in five steps:

1. Read the Welcome document and the user guide.
2. Review the 12 Level 1 areas with the executive sponsor and mark which the AI office owns.
3. Assign an owner to every Level 2 capability and settle any capability two teams claim.
4. Adjust the AI ratings and set your weights in the Excel workbook.
5. Run the readiness assessment and turn the top-ranked capabilities into proposals with an owner and KPIs.

The economics

The Alternatives

Few firms have described AI management as a function before, so the first draft is where most of the effort goes.

• Building it from a blank slate. The AI office collects activities from risk, technology, data and business teams, drafts definitions, agrees boundaries with each and repeats the review. Rating the capabilities for AI and building a scoring model is extra work on top.
• Hiring consultants to build it. Advisers bring a method and examples from elsewhere, and they still begin by drafting the generic structure. The firm's governance staff spend their time in workshops instead of on the decisions only they can make.
• Starting from this model. Definitions, KPIs, profiles, ratings, a scoring model and an assessment are in place and checked against each other. Workshops start by removing and adjusting, and the mandate can be drafted in the first weeks.

Deciding the AI office's scope, its risk appetite and its weights is work for people who know the firm, whether staff or advisers. The ratings in the files describe a typical large enterprise.

Public sample

A Sample Profile

The sample is 0.65% of the package and traces one thread through it: agent authority management. The fields below come from the Agent Authority Management (AM-7.2) AI profile:

• AI mode: Augment
• Oversight level: Human decides
• Risk rating: High
• Wave: Wave 2
• Fit, value, feasibility and data readiness scores: 3, 5, 3, 2 on a 1–5 scale
• Patterns of work: Rules-based decision, Judgment-based decision, Analysis & insight
• What AI could do: Draft an agent's written authority limits from its approved design and the agent authority standard: what it may read, propose, change and send, and what it must route to a person
• What people decide: The Head of AI Governance sets or changes the authority limits of an enterprise agent on what the Agent Authority Review Assistant prepares, after hearing the Enterprise AI Product Owner, the Agent Platform Lead, the Security Architect (cybersecurity) and the Privacy Lead

Upgrading Later

Buy this edition now and, if you upgrade to the AI Management Business Architecture Package: AI Transformation Blueprint within 90 days, its full price counts against the upgrade. Email your order number to support@finantrix.com for a one-time checkout code.

Read before you buy

Common Questions

Which edition fits my team?

Choose the AI-Ready Capability Model for the capability map, an AI profile for every L2 capability, prioritization, the heatmap and the readiness assessment. Choose the AI Transformation Blueprint for all of that plus value streams redesigned for AI, the business data model, agent specifications, use cases, responsible AI principles and controls, the operating model and the roadmap, a leadership presentation template, and bonus files on roles, operating model archetypes and transformation.

Can our internal AI tools use the content?

Yes, inside your organization. The Enterprise license covers internal tools, including assistants and agents you run yourselves. No license permits using the content to train, fine-tune or prompt AI systems on behalf of third parties.

Does it address AI laws and regulations?

No. The content is business architecture by intent: capabilities, value streams, data and the AI layer describe what AI management does and where AI fits, which holds under whatever laws apply to you. The risk tiers, principles and controls are lasting practices for your compliance team to map to the jurisdictions you operate in.

What happens when the 180 days end?

The license continues with no end date, and corrections stay free for as long as you use the files. New versions released within 180 days of purchase are free, and we email a download link for each. After that nothing renews, and new versions are offered at 25% off.

Can I move from the Capability Model to the full blueprint?

Yes. If you upgrade within 90 days, the full price paid for the Capability Model is credited. Email your order number to support@finantrix.com for a one-time checkout code.

Could an AI assistant write this instead?

An AI assistant can sketch what an AI office does, and that sketch is a fair place to start thinking. It changes with each prompt, and it cannot tell you which governance decisions are safe to hand to an agent and which must stay with a person. A board, an audit or a risk committee will ask for the record behind those choices. The matrix gives stable IDs for 298 Level 3 capabilities, cross-references validated before release, an AI profile with target mode, oversight level and risk tier for every Level 2 capability, and a scoring model you weight yourself.