Insights from AI Definitive
AI implementation,
in practice.
Practical guidance for selecting an AI workflow, proving it works and making it part of everyday operations.
Recent perspectives
Latest analysis
Practical decisions for AI governance and adoption.
Insurance
Build an insurer AI program that matches real operations
Connect the written program to the inventory, review decisions and operating evidence behind each use.
Read the guide
Workforce adoption
Replace unmanaged AI use with a usable approved path
Understand the work employees are trying to complete, then make the approved route practical.
Read the guide
Financial services
Questions to resolve before signing with an AI vendor
Focus due diligence on the proposed use, data path, operational dependency and exit plan.
Read the guide7 field guides
Plan and deliver
Choose the right scope, evaluate a partner and define the evidence needed to move forward.
Engineering
How to review AI-generated code beyond passing tests
Functional checks, business-rule parity and human review answer different questions.
Read the guide
Customer operations
Designing service agents that know when to escalate
Define the limits of a service agent before connecting it to customers and operational systems.
Read the guide
Planning
The decisions that shape an AI implementation budget
Estimate the work around data, integration, review and operation before comparing commercial proposals.
Read the guide
Engineering
Connecting Claude to enterprise systems with MCP
A common interface simplifies connection design. Each tool still needs permissions, validation and operating ownership.
Read the guide
Buying decisions
How to evaluate a Claude services partner
Compare the proposed team, acceptance evidence and handover commitments before comparing presentation quality.
Read the guide
Planning
Scope an AI pilot around a decision that can be made
One workflow, an agreed test set and a decision owner create a useful basis for investment.
Read the guide
Engineering
Why document-grounded AI still gives unreliable answers
Locate failures in source preparation, retrieval and answer generation before changing the model.
Read the guide5 field guides
Govern and operate
Connect risk decisions to practical controls, records and responsibilities.
Financial services
A practical control framework for Claude in financial services
Separate current supervisory scope from the operating controls needed for a specific AI workflow.
Read the guide
Financial services
What SR 26-2 changes for bank AI governance
The revised model-risk guidance leaves generative AI outside its scope. Banks still need a deliberate governance decision.
Read the guide
Governance
Turning Claude administration settings into operating controls
Platform settings become useful controls when ownership, review and response are defined.
Read the guide
Governance
The records needed to explain an AI-assisted action
Connect the request, source evidence, tool action and approval into one traceable record.
Read the guide
Governance
Build an evidence file for an AI-assisted workflow
Make it possible to explain the purpose, authority and outcome of a deployment from retained records.
Read the guide8 field guides
Industry applications
Design around the information, review and operating needs of the domain.
Healthcare
Designing document search around healthcare data boundaries
How to make clinical and operational information useful without losing control of who can retrieve it.
Read the guide
Legal
Protecting matter boundaries in AI-assisted legal review
Confidentiality, source access and attorney review need to be designed into the workflow.
Read the guide
Healthcare
Where human review belongs in prior authorization AI
Separate administrative assistance from the clinical judgment that determines a patient’s next step.
Read the guide
Legal
A verification workflow for AI-assisted legal research
A citation in an AI-generated draft is a starting point for review, not proof that the authority supports the argument.
Read the guide
Healthcare
Designing the prior-authorization preparation workflow
A sequence from incoming documents to a reviewable evidence package, with exceptions visible at every stage.
Read the guide
Financial services
A controlled starting point for member-service AI
Begin with approved information and clear transfers before adding account actions.
Read the guide
Insurance
Keeping evidence and judgment separate in insurance AI
Design claims and underwriting assistance around source material, decision authority and reviewable changes.
Read the guide
Healthcare
A release plan for AI in health-system operations
Move from a controlled trial to daily use through explicit operating and clinical review decisions.
Read the guide