AI Needs a Strategy, Not Just a Sponsor

The pattern is consistent enough that I recognize it within the first hour of any AI engagement. An organization has approved an AI transformation agenda. Leadership is committed. Budget has been allocated. And then the diagnostic data arrives and tells a different story.

I recently led a TMForum-based digital maturity assessment for a leading insurance company in Vietnam. Overall digital maturity sat at Level 3. But the data pillar - the foundation on which every AI ambition depends - scored at Level 2. Data governance recorded a gap of -1.63 against the organization's own 2030 targets. Only one in ten employees reported data access supported by AI or dashboards. A significant number had already begun using personal, consumer-grade AI tools for work - ungoverned, unsecured, and invisible to the organization's risk function.

The AI education market has made this worse. Organizations have produced a generation of AI-literate employees within a strategically illiterate organization. Employees know how to use Copilot, yet nobody has decided which processes it should touch. Data science teams build impressive models that never reach production. Executives approve AI budgets with no clear owner, no measurable objective, and no exit criterion. AI adoption is stalling not because of the technology, but because of how the decision to invest was made in the first place.

Why Most AI Investments Fail To Deliver ROI

This gap between investment and strategy is not unique to one sector. According to McKinsey, 53% of organizations that consider themselves AI-aware have not implemented anything in production. Accenture's conclusion is sharper: organizations with a formal AI strategy generate three times the return on investment (ROI) of those treating AI as a series of ad hoc projects. The gap is not explained by technology access; both groups have it. It is explained by strategic discipline.

In my experience, the root cause is consistent: AI investment decisions are treated as technology procurement rather than strategic decisions. The question being answered is not “Does this create value?” and “Are we ready to capture it?”, but “Are we keeping up?”. That reactive dynamic produces initiatives that are technically initiated and strategically underprepared. An organization without a clear AI strategy does not become more strategic by deploying AI. It becomes more efficiently committed to the wrong priorities.

AI Strategy Vs Classical Strategy: What Stays The Same, And What Changes

The most dangerous misconception in AI strategy is that it is binary, which means either AI is just another capital investment, or it is so novel that classical tools are useless. Both are wrong. Classical strategy provides the right foundation. The goal remains unchanged, and it is sustainable competitive advantage. Every investment still requires a business case, a return estimate, a risk assessment, and a definition of success. Make-or-buy logic applies directly. Change management is essential regardless of the technology. Executive sponsorship remains the single strongest predictor of transformation success.

The listed four AI-specific demands sit on the foundation:

First, data is a prerequisite, not a project input. Quality, completeness, and legal ownership must be assessed before committing resources, not during delivery. In the life insurance engagement, the organization’s most ambitious AI use cases were Level 4 capabilities planned on a Level 2 data infrastructure. No budget closes that gap without sequencing the investment correctly.

Second, AI follows a mandatory maturity sequence that cannot be skipped. The technology does not override the constraint; it exposes it.

Third, AI introduces risk categories with no classical equivalent: algorithmic bias, interpretability failures, and data liability. The employees' self-provisioning of consumer AI tools was not a training gap; they were a live governance exposure.

Fourth, AI returns are non-linear: data advantages compound in ways that structurally widen early leads, not marginally.

Questions Every AI Investment Decision Must Answer

Translating this into practice means insisting on honest answers to four questions before any AI investment is approved - questions that mirror classical discipline but extend it into AI-specific territory.

• What specific problem are we solving, and is it worth solving with AI? “We want to use AI in customer service” is not an answer. “We want to reduce first-contact resolution time by 25% for the inquiry types that consume 40% of agent hours” is. The specificity forces three conversations: whether the problem justifies the investment; whether AI is the right instrument (not every problem needs it!); and how success will be measured before the project begins.
• Do we have the data to do this well? This question has no equivalent in classical capital allocation, and its absence from most AI governance processes is the most common cause of failure. In the life insurance engagement, data governance scored 2.25 out of 5- the weakest of 136 assessed criteria. Before any build decision, organizations need honest answers: Is the data accessible? Is it clean? Is it representative of the problem? And, is it legally ours to use? The last question is not an IT detail. It is a strategic constraint.
• Should we build, buy, or partner, and why? Apply make-or-buy logic with three AI-specific criteria: data uniqueness, strategic differentiation, and internal talent availability. In the engagement, several planned build initiatives did not survive this filter. The capabilities were available from mature vendors, and the organization’s real differentiator was its proprietary customer data, not its model-building capacity. The strategic value is in the data, not in the architecture layered on top of it.
• How will we govern, measure, and, if necessary, stop this? Every AI investment needs an owner, a measurable baseline, a review cadence, and a shutdown criterion. The life insurance organization’s culture and values scored 4.0 out of 5—the strongest dimension in the assessment. But the employee survey showed that KPI structures did not reward participation in transformation initiatives. Strong values without aligned incentives produce aspiration, not behaviour change.

AI Strategy Frameworks That Support Better Investment Decisions

Diagnosing the problem is necessary. Solving it requires concrete instruments. Several AI strategy frameworks map precisely onto the four gaps identified above.

Gartner's AI Maturity Model makes the capability sequencing problem actionable. When the life insurance company's leadership discussed deploying AI-driven personalisation, the model allowed us to be precise: that capability is executable at Level 4; the current data infrastructure sits at Level 2. The strategic recommendation becomes sequencing, not delay—build what makes the AI use case viable before approving it.

The Big Data Business Model Maturity Index (BDBMMI) extends this to data investment, mapping each phase of data capability to the business value it enables, preventing organizations from approving AI deployments that their architecture cannot yet support.

The Build-Buy-Partner framework applies the make-or-buy discipline with AI-specific decision criteria. For prioritization and risk, the six-factor AI Strategy Prioritization criteria, including strategic rationale, opportunity size, investment required, expected ROI, risk, and timeline, create the scoring rigour that mirrors a capital allocation committee. Every AI project proposal should pass through this filter before receiving budget approval.

ADKAR (Awareness, Desire, Knowledge, Ability, Reinforcement) addresses what classical governance consistently misses: organization do not change because leadership decides to. In the life insurance engagement, 35% of staff did not understand their own role in the transformation, and employees under 35—the group most expected to lead it—had the lowest strategic awareness of any cohort.

ADKAR makes it clear that this is not a communication failure. It is an architecture failure: awareness built at the executive level had not been translated into aligned training, incentives, or career pathways. That diagnosis turns a morale problem into a set of specific intervention points.

How To Build Strategic Discipline Into Every Ai Investment Decision

The bottleneck in enterprise AI adoption is not the technology. AI is more capable than most organizations are currently using it. The bottleneck is the quality of strategic thinking applied before the investment is made—and the willingness to ask the hard questions rather than approve the proposal.

For executives, the practical implication is this: evaluate the next AI proposal as a capital allocation decision, not a technology one. Ask the four questions. Use the maturity model to assess what is actually executable today. Apply build-buy-partner rigorously. Ensure governance is specified upfront. And if the proposal cannot answer those questions clearly, send it back. Not because AI is not worth investing in, but because AI investments that lack strategic clarity do not become better by being approved faster.

The organizations that lead the AI decade will not be those that moved fastest. They will be those who moved with the clearest understanding of where they were starting from.