FAQ
Common questions
Straight answers on AI Centers of Excellence, governance, and how we work.
About AI Centers of Excellence
What is an AI Center of Excellence?
An AI Center of Excellence is the team and operating model that decides which AI work gets done, sets the standards it must meet, and supports it in production. It owns intake, prioritization, validation, and the path from pilot to production. It does not have to build every use case itself. Its job is to make AI delivery repeatable and governed across the business.
Do we need a central CoE, or can business units run AI on their own?
Most regulated firms need both. A small central team owns standards, the governed platform, risk review, and the portfolio view. Business units own their use cases and outcomes. Fully decentralized AI produces duplicate tools and untracked risk. Fully centralized AI becomes a bottleneck. We design the split around your size, talent, and risk appetite.
How do we know how mature our AI capability is?
Map it. Our free AI Capability Maturity Framework assesses capabilities across three tiers, Foundation, Value, and Advanced, over five levels. Most organizations get most of their value from solid mid-level capability. Full autonomy is a choice, not a requirement. The framework shows which gaps to close next.
Governance and validation
What should AI governance at an insurer cover?
At minimum: an inventory of every AI system in use, including AI inside SaaS tools; due diligence on foundation models and vendors; a risk review and approval workflow; validation and monitoring standards; and documentation an auditor can follow.
How do you validate generative and agentic AI before it reaches production?
Nothing ships without three checks. The output must be grounded in an approved source. It must pass a golden evaluation set built from real cases with known answers. And an independent model, acting as judge, reviews it for hallucination and risk. Agentic systems also need human review thresholds, audit logs, and clear limits on what an agent can do without approval.
We have AI pilots but no governance. Where do we start?
Start with an inventory. You cannot govern what you cannot see. Map every AI system in use, triage each by risk, and put a lightweight approval path in place for new work. Then stand up the operating model that turns scattered pilots into a governed intake pipeline. Build governance and the CoE together, not one after the other.
Working with us
Which service should we start with?
Start from your situation. If pilots are scattered and no one owns AI work, start with a CoE Launch. If an audit or regulator is asking about AI, start with a Governance Assessment. If leadership wants an AI plan and you have more ideas than capacity, start with Use Case Discovery. If AI work is underway and you need senior oversight, an Advisory Retainer fits. If you are unsure, a 15-minute call will settle it.
Who will actually do the work?
Every engagement is led and overseen by Abhishek Agrawal. Depending on scope, delivery includes vetted partners, onshore and offshore, in architecture, data, security, and engineering. You will know who is on the team, and what each person does, before work starts, and every deliverable is reviewed by Abhishek before it reaches you.
How is pricing structured?
Fixed-scope services (CoE Launch, Governance Assessment, Use Case Discovery) are priced as a fixed fee for the agreed scope and deliverables. The Advisory Retainer is a monthly fee for an agreed scope and cadence. Both are set out in writing in the proposal before work begins.
How do you handle our confidential information?
We sign your NDA before any detailed discussion, and every partner on the engagement works under the same confidentiality terms. We agree with you in writing who can access your systems and data, and from where, before work begins. We never use your data to train AI models, and we do not name clients or share their work without written permission.
Why work with a boutique instead of a large consultancy?
Senior oversight on every part of the work, not only in the pitch. We are vendor-neutral, scope every engagement in writing, and build governance into delivery rather than adding it at the end. We also work alongside large firms and delivery partners when you already have them.
Which industries do you work with?
Insurers, banks, wealth and asset managers, and other regulated financial services firms.
How do you decide which AI use cases to fund first?
We score each candidate on value, feasibility, data readiness, and risk, then separate quick wins from longer-term bets. We also check whether fixing the process would deliver most of the gain without AI. Often it does. The result is a ranked portfolio tied to measurable outcomes, not a list of ideas.
Do you replace our internal teams or delivery partners?
No. We design the operating model, set the standards, and oversee delivery. Your teams and partners do the building. We are vendor-neutral.
Terms we use
- AI Center of Excellence (CoE)
- The team and operating model that decides which AI work gets done, sets its standards, and supports it in production.
- Agentic AI
- AI systems that take actions, such as updating records or calling other systems, not only generate text.
- Golden evaluation set
- A fixed set of real cases with known correct answers, used to test AI output before release.
- Model-as-judge
- A second, independent AI model that reviews output for errors, unsupported claims, or risk.
- Grounding
- Requiring AI output to come from an approved source, such as policy documents, rather than the model's general knowledge.
- AI inventory
- A list of every AI system in use, including AI inside purchased software, with its owner, data, and risk level.
- NIST AI RMF
- The U.S. National Institute of Standards and Technology's AI Risk Management Framework.
- RAG (retrieval-augmented generation)
- A design where the AI retrieves approved documents and answers from them.
Let's begin