Raft shows how to prioritize AI initiatives and build a realistic roadmap

Raft has proposed a practical way to turn a scattered list of AI ideas into a managed transformation program: first calculate the potential value of each initiative, then match it against feasibility, choose an implementation approach, and only then build a roadmap. The approach is aimed at companies that already see dozens of possible AI scenarios but do not want to drown in pilots with no measurable result.

The central thesis is this: almost anything can be implemented, but it is impossible to do everything at once. That is why the author suggests dropping the chaotic set of experiments and moving to portfolio management of AI initiatives.

The logic starts with a gap analysis. The organization gets a list of tasks showing which use cases it is ready for now, and for which it lacks data, infrastructure, skills or security controls.

But a feasibility assessment alone is not enough. To decide where money and team resources are really worth spending, the expected business value has to be assessed separately. For this, Raft offers a simple potential-value formula: the scale of the problem, how often it occurs, and the complexity of solving it manually. Each parameter is rated from 1 to 10, so the final score falls between 1 and 1000. Scale shows how many people or processes the problem affects; frequency, how often it occurs; manual complexity, how much time, money and effort it takes without AI automation.

The authors also note a practical threshold: if an initiative does not score at least 200 points, in-house development may take too long to pay off. In that case it makes more sense to look at off-the-shelf products and consciously accept their limitations, rather than launch expensive custom development for a weak effect.

The next step is a value-feasibility matrix. It helps separate initiatives that deliver maximum return at acceptable complexity from those that are still premature or economically doubtful. The proposal is to take projects from the upper-left quadrant of the 2x2: quick wins, big bets, incremental innovations and other scenarios with a noticeable effect and a clear path to implementation.

The company then assembles what Raft calls the AI Tech Gartner’s Sandwich: for each initiative it picks a buy, build or partner delivery model, then adds a layer of risk and security. The idea is to treat AI not as a standalone service but as a multilayer system of applications, platform components and protective mechanisms, including the TRiSM approach to managing trust, risk and security.

The end result of this selection is not just a list of priorities but a phased transformation roadmap. In the first phase, which the authors call the foundation, they recommend taking one or two initiatives with high value and low complexity, launching pilots with a quick payoff, and closing technical gaps in parallel: preparing data, building integrations, strengthening infrastructure, hiring missing roles and training staff. For this phase, the cloud is often preferred to lower the cost of testing hypotheses and shorten time-to-market. The recommended duration is six to twelve months.

Next comes scaling: successful pilots are extended to new teams and processes, and medium-complexity initiatives move into active implementation once the critical gaps are closed. In parallel, the layer of threat monitoring and protection for AI systems is reinforced.

The third phase is optimization, when AI tools have become part of day-to-day operations and the focus shifts to SLAs, quality control, fighting model degradation, maintaining the MLOps pipeline and reducing total cost of ownership, including a possible move from the cloud to on-prem infrastructure.

The main conclusion is that AI transformation requires discipline, not inspiration. A company that first calculates value, then honestly checks feasibility, and only then sorts initiatives into phases lowers the risk of getting stuck in endless prototypes and starts managing AI like a mature investment portfolio. For the market, this is an important signal: the winners will not be those who launched a demo fastest, but those who tied AI projects to business metrics, security and a realistic rollout plan.