Five Questions Every Board Should Ask About AI Strategy — Separating Credible AI Capability, Governance and Measurable Outcomes from Compelling Presentations

By Raviprasad Bangalore | Senior Vice President & Head of Technology, Comviva (Tech Mahindra) | India's Top 10 AI Trailblazers — ET India


The Presentation Problem

Every board in India and globally is now receiving AI strategy presentations. The decks are polished. The language is confident. The opportunity slides are compelling — efficiency gains, cost reductions, new revenue streams, competitive differentiation.

And almost none of them answer the questions that actually matter from a governance perspective.

This is not because management teams are being dishonest. It is because AI strategy presentations are built to secure approval and build confidence. They are optimised for persuasion, not interrogation. The metrics that make the investment case look strong are front and centre. The assumptions that could make it fall apart are buried in the appendix — or absent entirely.

I have spent nearly three decades building technology platforms, leading AI and cloud transformation programmes across Fortune 500 enterprises, and architecting Agentic AI systems at production scale. I have sat on both sides of this conversation — as the person building the capability and as the person being asked to govern it. What I have learned is this: the quality of a board's AI oversight is almost entirely determined by the quality of its questions.

Here are the five questions that separate credible AI governance from passive approval.


Question 1: What Problem Are We Actually Solving — And How Do We Know AI Is the Right Answer?

This sounds obvious. It is almost never asked directly.

AI strategy presentations typically open with the opportunity — the market is moving, competitors are investing, the technology is maturing. What they rarely open with is a precise, commercially grounded statement of the specific business problem the AI investment is designed to solve.

The distinction matters enormously. "We are investing in AI to improve operational efficiency" is not a problem statement. It is a direction. A problem statement sounds like this: "Our claims processing team handles 4,000 cases per month. Average turnaround time is 14 days. Industry benchmark is 6 days. That gap is costing us customer retention and increasing our cost per claim by 40% above benchmark. We believe AI-driven automation can close that gap."

When I led AI transformation for a Fortune 500 insurer, we reduced policy document processing turnaround by 60% — not because we deployed AI broadly, but because we started with a precise problem definition and worked backwards to the solution. The AI was the answer to a specific question, not a solution looking for a problem.

The board's job is to insist on that precision. When a management team cannot articulate the specific problem with specific metrics, the investment case is built on aspiration rather than evidence. That is not a reason to reject the investment — but it is a reason to send the proposal back for sharper problem definition before approval.

Ask: "What is the specific, measurable business problem this AI investment solves — and what evidence do we have that AI is the most effective solution to that problem?"


Question 2: What Does the Governance Framework Look Like — Not the Strategy, the Governance?

AI strategy and AI governance are not the same thing. Most boards are shown the former and asked to approve both simultaneously.

Strategy tells you where the organisation is going with AI — the use cases, the investment priorities, the expected outcomes. Governance tells you how decisions will be made along the way — who is accountable for AI outcomes, what the decision boundaries are, how errors will be detected and corrected, and what the escalation path looks like when something goes wrong.

In my work building Agentic AI platforms — systems that act autonomously, make decisions without human review at each step, and operate at scale across enterprise workflows — the governance architecture is not an afterthought. It is a foundational design decision. At Comviva, when we architected the Agentic Orchestration Platform, we established explicit governance foundations: orchestration boundaries, contextual intelligence guardrails, security layers, and accountability frameworks — before we scaled the platform across product units.

Most enterprise AI deployments do not start there. They start with the use case and add governance later — often after the first failure.

The board's role is to require that governance is designed in, not bolted on. A management team that presents a compelling AI strategy without an equally detailed governance framework is presenting half a proposal.

Ask: "Walk us through the AI governance framework — specifically, who is accountable for AI outcomes, what decisions require human review, how errors are detected at scale, and what the escalation path is when the system produces a harmful outcome."


Question 3: What Are the Measurable Outcomes — And What Is the Decision Criteria for Scaling, Pausing, or Stopping?

AI investment proposals are almost always accompanied by projected outcomes. Efficiency gains of 30–40%. Cost reductions of 50–60%. Productivity improvements of 20–30%. These numbers are not fabricated — they are often drawn from pilot results or industry benchmarks. But they are projections, not commitments. And the gap between a pilot result and an enterprise-scale outcome is where most AI investments quietly underperform.

I have delivered measurable AI outcomes across multiple sectors — 60% reduction in policy processing turnaround for a Fortune 500 insurer, 70% reduction in manual claims effort for a healthcare payer, 30–40% acceleration in software delivery cycles through Agentic AI orchestration. These outcomes were real. But they were real because we defined the measurement framework before deployment, not after. We knew exactly what we were measuring, how we were measuring it, and what the threshold was for declaring success or failure.

The board's governance role is not to validate the projections. It is to ensure that the measurement framework is rigorous and that there are clear decision criteria for what happens if the outcomes do not materialise. Without those criteria, AI investments have a tendency to continue indefinitely — consuming budget, generating activity, and producing reports that show progress without showing results.

Ask: "What are the specific, measurable outcomes we are committing to — with timelines? And what are the decision criteria for scaling the investment if it works, pausing it if it underperforms, and stopping it if it fails?"


Question 4: What Is Our Real AI Capability — Not Our AI Ambition?

There is a consistent gap in AI strategy presentations between the capability the organisation aspires to build and the capability it actually has today. This gap is rarely made explicit — because making it explicit would complicate the investment case.

The honest capability assessment covers four dimensions that boards should probe independently:

Data readiness. AI systems are only as good as the data they operate on. In my experience leading large-scale AI transformation programmes — including an Agentic Data Profiler that cut data remediation effort by 60% — data quality issues are the single most common reason AI deployments underperform or fail. Ask specifically: what is the current state of our data quality, data governance, and data accessibility? Has an independent assessment been done?

Engineering capability. Building AI systems that work in a demo environment and building AI systems that operate reliably at enterprise scale are fundamentally different challenges. Ask: do we have the engineering capability to build, deploy, and maintain this system — or are we dependent on a vendor? If vendor-dependent, what is the exit strategy?

Organisational readiness. AI systems change how people work. The organisations that get the most value from AI are the ones that invest as heavily in change management and capability building as they do in the technology itself. Ask: what is the organisational change plan, and how are we measuring adoption?

Leadership capability. Who in the organisation is accountable for AI outcomes? Is there a Chief AI Officer, a Head of AI, or a senior technology leader with the authority and capability to govern AI deployment across the enterprise? Or is AI being managed as a project rather than a capability?

Ask: "Can you give us an honest assessment of our current AI capability across data readiness, engineering depth, organisational readiness, and leadership accountability — not where we plan to be, but where we are today?"


Question 5: What Are the Risks We Are Not Being Shown?

Every AI investment proposal shows the upside. The board's job is to see the downside — not to be pessimistic, but because that is what independent governance requires.

The risks that matter most in AI deployments are rarely the ones that appear in the risk section of the investment proposal. The ones that appear there are the obvious, manageable risks — data privacy, regulatory compliance, vendor dependency. The ones that do not appear are the ones that have actually caused AI programmes to fail or cause harm.

From my experience building and governing AI systems at scale, the risks that boards should probe specifically are:

Amplification risk. AI systems do not just automate processes — they amplify them. A flawed process, automated at scale, produces flawed outcomes at scale. Before approving an AI investment, ask: are we automating a process that is already working well, or are we automating a process that has known problems? If the latter, the AI will make those problems bigger, not smaller.

Accountability diffusion. When an AI system makes a decision that causes harm — a customer is wrongly denied a claim, a credit application is incorrectly rejected, a patient receives the wrong recommendation — who is accountable? In many organisations, the answer is genuinely unclear. The technology team points to the business team. The business team points to the vendor. The vendor points to the data. Boards need to ensure that accountability is assigned before deployment, not negotiated after an incident.

Regulatory trajectory. The regulatory environment for AI is moving fast. The EU AI Act is in force. India's Digital Personal Data Protection Act has implications for AI systems that process personal data. Sector-specific regulators in BFSI and healthcare are developing AI-specific guidance. An AI deployment that is compliant today may face significant regulatory requirements within 12–18 months. Ask: has the legal and compliance team assessed not just current regulatory requirements but the likely regulatory trajectory over the next two to three years?

Concentration risk. Many enterprise AI deployments are built on a small number of foundation models from a small number of providers. That creates concentration risk — dependency on a vendor's pricing decisions, model updates, and continued availability. Ask: what is our strategy for managing AI vendor concentration risk?

Ask: "What are the three scenarios in which this AI investment causes the most harm — to our customers, our organisation, or our regulatory standing — and what is our mitigation plan for each?"


What These Five Questions Actually Do

These questions do three things simultaneously.

They signal board competence. A management team that receives these questions from a board knows it is dealing with directors who understand AI at a governance level — not just a conceptual level. That changes the quality of every subsequent conversation.

They improve the investment. In my experience, the organisations that ask the hardest governance questions before deployment are the ones that get the best outcomes from AI. Not because the questions slow things down — but because they force the precision and rigour that separates AI programmes that deliver from AI programmes that disappoint.

They protect the organisation. The board's primary governance responsibility is to protect the long-term interests of the organisation and its stakeholders. In the AI era, that means ensuring that the speed of AI adoption does not outpace the organisation's ability to govern it responsibly. These five questions are a practical mechanism for maintaining that balance.


A Note on What Good Looks Like

I want to be direct about something. The goal of these questions is not to slow down AI adoption. The organisations that govern AI well are not the ones that are most cautious — they are the ones that are most precise. They move fast on the use cases where the problem is clear, the capability is real, the governance is designed in, and the outcomes are measurable. They move slowly — or stop — on the use cases where those conditions are not met.

That precision is what separates AI programmes that deliver sustained, measurable value from AI programmes that generate impressive presentations and disappointing results.

The board's role is not to be the brake on AI adoption. It is to be the mechanism that ensures AI adoption is built on a foundation that can actually support the ambition.

These five questions are how you do that.


Raviprasad Bangalore is Senior Vice President & Head of Technology at Comviva (Tech Mahindra), recognised among India's Top 10 AI Trailblazers by ET India. He has built and governed Agentic AI platforms at production scale, led AI transformation programmes across Fortune 500 enterprises in healthcare, insurance, and telecom, and managed technology portfolios exceeding $125M. He advises boards and enterprises on AI governance, cloud transformation strategy, and enterprise technology oversight.