Introduction
Enterprises continue to invest heavily in artificial intelligence, with the expectation that it will improve efficiency, reduce risk, and unlock new value. Yet in practice, many AI initiatives fail to scale or deliver tangible business impact. The problem is not the technology; it is the approach to identifying where and how AI should be applied.
Many programmes begin with predefined use cases promoted as industry best practice. While these can be useful as inspiration, they often overlook the specific context, challenges, and performance levers of the organisation. This can lead to AI solutions that are technically impressive but misaligned with business priorities, and therefore deliver limited value.
A business journey approach takes a different path. It begins with mapping how value is actually created or lost — the end-to-end activities that turn inputs into outcomes. Within each journey lie decision points, moments where a choice affects efficiency, cost, risk, and experience. By identifying and quantifying these decisions, it becomes possible to see where better intelligence could deliver the most meaningful results.
This method anchors AI discovery in measurable business performance, ensuring that any AI initiative is evaluated against its potential to improve outcomes that truly matter. The result is a more focused, data-driven approach that aligns investment with strategic priorities and builds a stronger business case for execution.
Why Discovery Must Go Beyond the Visible
A fraud prevention system may appear to be functioning well because it operates reliably and integrates cleanly with other systems. Yet technical stability does not guarantee business impact. Without examining metrics such as actual fraud losses, false positive rates, investigation times, recovery rates, and downstream customer experience, it is impossible to know whether the system is achieving its intended objectives.
Business outcomes must be the anchor for AI discovery. The fact that a process or system runs smoothly says little about whether it is protecting revenue, reducing costs, managing risk effectively, or improving customer satisfaction. These are the measures that matter; they are often not visible from a purely technical or operational perspective.
A system can run perfectly and still allow significant value leakage if it is not making optimal decisions at critical points in the process. For example, a fraud model that is overly cautious can generate excessive false positives, flagging and even denying legitimate transactions. This drives up investigation costs, frustrates customers, and erodes trust. On the other hand, a model that is too permissive may reduce false positives but create more false negatives, allowing high-value fraud to slip through undetected. Only by tying these decision outcomes to measurable KPIs can the true performance of the system be understood.
This is why AI discovery must be anchored in business performance, not only in technical capability. The objective is to identify decisions that materially influence outcomes, measure the quality of those decisions using real business metrics, and determine where additional intelligence could produce measurable gains. This approach ensures that AI delivers tangible, defensible results that are directly linked to strategic objectives and financial performance.
The Business Journey Approach
A business journey is the complete sequence of activities that delivers a specific outcome for the organisation and its customers. Examples include authorising a credit card transaction, processing an insurance claim, approving a commercial loan, or fulfilling an online order. Each journey contains a series of decision points — moments where a choice must be made that affects efficiency, cost, risk, and customer experience.
The business journey approach to AI discovery begins by mapping these processes end-to-end. The objective is not simply to document steps, but to understand how value is created, where it is delayed, and where it is lost. By tracing the flow of work from start to finish, critical decision points naturally emerge. These are the moments that determine whether an outcome will be profitable, compliant, timely, or satisfying to the customer.
This is not a “boil the ocean” exercise. While a journey could contain many potential decision points, in practice the number of high-value ones is typically manageable — often in the single digits. This keeps the scope focused and ensures that effort is directed where it will have the greatest impact.
Once these decision points are identified, they can be assessed against relevant business KPIs. This may include measuring how frequently the decision occurs, the cost or risk of making it incorrectly, and the potential benefit of making it better. This quantitative lens ensures that opportunities are ranked by impact, not by intuition or familiarity.
By focusing on decision points within a journey, AI opportunities are identified in direct connection to the organisation’s strategic goals. This avoids the trap of chasing solutions in search of a problem and instead ensures that AI is applied where it will produce the greatest business benefit. It also provides a clear line of sight from AI investment to measurable outcomes, making it easier to build a credible business case and secure stakeholder commitment.
The business journey approach does more than find opportunities. It creates a repeatable framework for continuous improvement. Once one journey is mapped and optimised, the same method can be applied to others, building a sustainable pipeline of high-value AI initiatives across the enterprise.
Quantifying Opportunity
Identifying the right decision points is only the first step. To turn those opportunities into actionable AI initiatives, they must be quantified in terms of potential business impact. This requires establishing a clear baseline of current performance and linking it to measurable outcomes.
Without a baseline, it is impossible to determine what improvement would look like or to calculate the return on investment. Claims of large potential savings or revenue gains may sound compelling, but without supporting data they remain speculative. A defensible business case begins with facts, not estimates borrowed from other organisations or generic benchmarks.
Quantification works at two levels. At the journey level, metrics such as total annual claim volumes, transaction counts, and processing costs provide a high-level view of performance and potential value leakage. At the decision point level, the analysis becomes more granular, focusing on factors such as error rates, manual review percentages, false positives, and other drivers of inefficiency or loss. In some cases, a single decision point may carry nearly all of the journey’s volume, while in others it may account for only a fraction. This breakdown helps identify where intervention will yield the greatest benefit.
Leakage is used as a generic measure of lost value, whether it results from errors, manual handling, false positives, delays, or other inefficiencies. For each decision point, the contributing factors to leakage can be itemised, providing a clear view of both the scale and the sources of the problem.
From there, the potential uplift from applying AI can be modelled. This should be based on realistic expectations drawn from relevant data and evidence, rather than optimistic projections. By comparing the baseline to the projected improvement, it becomes possible to estimate tangible outcomes in financial terms — such as reduced losses, increased revenue, lower operating costs, or improved customer retention.
This disciplined approach to quantification ensures that AI initiatives are prioritised according to their likely business impact, rather than their novelty or technical appeal. It also gives stakeholders the confidence that each proposed project is grounded in measurable, defensible value.
Technical Feasibility
Identifying a high-value decision point is necessary but not sufficient. To apply AI, the decision must also be technically feasible. Feasibility depends on whether the right data exists for model development and training, and whether those same data — in the form of features — are available at the exact moment of inferencing to inform the decision.
The first consideration is data availability in two forms: historical data for model training and runtime data for scoring. Historical data must include the inputs and outcomes required to learn meaningful patterns. Runtime data must be available at the exact moment the decision is made, in the right format and within the required latency window.
Labelling and outcome truth matter as well. Supervised learning needs clear examples of past results — for instance confirmed fraud versus legitimate transactions, or claims paid as billed versus denied for cause. If labels are sparse or delayed, you may begin with unsupervised or semi-supervised approaches, such as anomaly detection, while a labelling pipeline is established.
Runtime feature access is critical. It is not enough for data to exist somewhere; it has to be accessible instantly. Adaptive Behavioural Models (ABMs), which detect patterns over time, require richer context at scoring time — such as recent activity, rolling aggregates, cross-channel signals, and peer-group comparisons. When these features are not instantly available, they should be precomputed or enriched in flight.
On IBM Z, on-platform inferencing enables sub-millisecond scoring inside transactional systems. This allows 100% of eligible transactions to be evaluated in real time without sampling or off-platform delays, so decision intelligence can be applied universally rather than to a subset of activity.
Governance and explainability must also be considered from the outset. Models need to meet regulatory, ethical, and organisational standards, with decision logic that can be understood, validated, and defended. This is particularly important where decisions impact customers, compliance, or financial risk.
Feasibility does not eliminate opportunities; it shapes the delivery path. If runtime features are missing, build the capability to supply them. If governance or explainability constraints apply, adapt the model choice and deployment strategy. The outcome of feasibility is a clear plan: immediate execution, execution with prerequisites, or a phased approach that builds foundations first.
Prioritisation & Roadmapping
Once opportunities have been quantified for business impact and confirmed for technical feasibility, the next step is to determine which ones to pursue first. AI resources — data engineering capacity, modelling expertise, infrastructure, and business stakeholder attention — are finite. A clear prioritisation framework ensures they are directed toward the opportunities that will deliver the greatest return in the shortest time.
Prioritisation begins by evaluating each decision point across two main dimensions: value and feasibility. Value reflects the potential uplift in key business KPIs such as revenue, cost reduction, risk mitigation, or customer satisfaction. Feasibility reflects the readiness of the data, the runtime environment, integration pathways, and compliance requirements. Plotting opportunities on a value–feasibility matrix makes it immediately clear which should move forward first: those that offer high value and are ready to implement with manageable effort.
However, sequencing is not just about picking the easy wins. Strategic dependencies matter. In some cases, a high-value decision point may require data or infrastructure capabilities that can be developed by delivering a smaller, adjacent opportunity first. For example, building a feature store for a moderately complex decision today may enable a far higher-value decision to be addressed six months later.
Roadmapping turns the prioritised list into an actionable plan. This involves aligning each opportunity with the resources, skills, and stakeholder commitments required to deliver it. It also means defining the expected delivery horizon and the checkpoints where value realisation will be measured. This plan should be dynamic, reviewed regularly as new decision points are discovered, performance data is captured, and the organisation’s strategic priorities evolve.
By combining rigorous prioritisation with a clear roadmap, AI opportunity discovery moves from a list of possibilities to a structured programme of initiatives. The result is a pipeline of projects that are both impactful and executable, ensuring that AI investments consistently produce measurable business outcomes.
Example: Discovering High-Value Decision Points in Claims Processing
A national healthcare payer processed 50 million claims annually. A conventional “fraud detection” use case would have focused on a narrow part of the process. By instead mapping the end-to-end claims adjudication journey, the team uncovered several decision points driving significant value leakage.
Rather than starting with a predefined solution, the discovery process traced the claim from submission to settlement, revealing specific moments where errors, inefficiencies, and fraud exposure combined to erode performance.
Claim validation — Needed to prescreen claims for errors. Analysis found some providers were exploiting the process, knowing that erroneous claims would need to be resubmitted and would then always be manually reviewed, bypassing straight-through processing controls.
CPT/ICD/modifier anomalies — Needed to detect mismatches between the diagnosis (ICD), procedures performed (CPT), and modifiers used. Unlike MUEs, which flag improbable quantities for a single service, these anomalies catch cases where the diagnosis and treatment do not logically align — often signalling errors or intentional miscoding.
Bundling and unbundling patterns — Needed to identify overpayments, including cases of semantic fraud where providers deliberately unbundled related procedures to claim higher reimbursements.
Cost and frequency anomalies — Needed to detect outlier claims with unusually high costs or excessive frequency of certain procedures, identifying patterns that rules-based checks had missed.
Interest and penalties — Late claim payments were triggering regulator-imposed interest and penalties, putting the payer under pressure to accelerate adjudication without increasing errors or fraud exposure.
This structured analysis provided a clear set of quantified opportunities, each tied to specific operational decisions. It also supplied the baseline volumes and metrics needed to build a credible business case, ensuring that any AI investment would directly address the areas with the greatest potential for improvement.
The Strategic Playbook
Successful AI transformation is not the result of isolated projects. It follows a deliberate sequence that moves from insight to measurable impact. The organisations that win with AI apply a consistent method, ensuring that investment is directed toward the decisions that matter most.
Map — Begin by identifying the business journeys where critical decisions are made. These are the points in your operations where outcomes directly determine revenue, cost, or risk. Mapping these journeys reveals where value is created, where it is lost, and where AI could make a measurable difference.
Quantify — For each decision point, determine the volume of decisions made and the financial impact of getting them wrong. This puts a real number on the value at stake and creates a fact-based case for change. Without quantification, AI remains a technology initiative instead of a business priority.
Prioritise — Use a dual lens of business value and technical feasibility to identify the decisions worth pursuing first. High-value, high-feasibility opportunities deliver rapid proof of impact. High-value but lower-feasibility opportunities become strategic investments for the future.
Act — Deploy AI into the highest-priority decision points, focusing on production impact rather than prolonged proof-of-concept exercises. Embed the models into operational workflows so they influence outcomes in real time.
Measure and refine — Track the business results, not just technical metrics like model accuracy. Monitor the value delivered, refine the models as conditions change, and extend the approach to other high-impact decisions across the organisation.
This disciplined sequence ensures that AI is not treated as an experiment but as a repeatable capability that compounds value over time. It aligns technology with measurable outcomes and avoids the common trap of chasing too many ideas without delivering meaningful results.
By following a playbook grounded in mapping, quantification, prioritisation, action, and measurement, AI becomes a driver of business transformation, not just a technical project.
AI Discovery Workshop
The most effective way to assess a business journey is through a focused AI Discovery Workshop. This structured session brings together the right stakeholders to map the end-to-end journey, surface critical decision points, and evaluate each for AI potential.
The Right Mix of Stakeholders
Execution starts with assembling a team that understands the targeted business outcome, the decisions that drive it, and the data and systems that enable those decisions. This typically includes an Executive Sponsor (such as a Head of Business Unit or COO) to set vision, priorities, and business alignment; a Business Owner to define pain points, goals, and success factors; a Business Analyst to walk through journey flows, decisions, and business metrics; an Application Owner to explain how decisions are implemented today; a Data Owner to share data availability, structure, and labelling knowledge; an Infrastructure or Platform Owner to describe systems architecture and deployment context; and an AI Lead or Data Scientist to advise on model suitability and AI readiness.
Why the Mix Matters
Workshops that involve only infrastructure or platform teams tend to uncover technology upgrades rather than genuine AI opportunities. While infrastructure readiness is important, it cannot substitute for understanding decision-making context, performance levers, or business priorities.
Collaborative Delivery
Workshops should be a coordinated effort between business and technical stakeholders. This collaboration ensures feasibility is validated, success measures are agreed, and deployment plans are designed to deliver both technical soundness and measurable business improvement.
The Outcome
Involving the full spectrum of stakeholders from the outset ensures AI opportunities identified in the business journey are implementable, strategically aligned, and capable of delivering sustained impact.
Conclusion
AI delivers the greatest value when it is applied where it matters most: the high-volume, high-impact decisions that shape business performance. Finding these decisions requires going beyond visible pain points and familiar examples to examine the operational journeys that produce core business outcomes.
The business journey approach offers a structured, repeatable method for identifying, sizing, and validating AI opportunities. By anchoring discovery in measurable business metrics, assessing technical feasibility, and prioritising with intent, organisations can build a focused roadmap of AI initiatives that are both executable and impactful.
This approach moves AI from theory to practice — from scattered experiments to a coordinated programme of initiatives that directly support strategic objectives. The result is not just more AI projects, but the right AI projects, delivering measurable improvements to the decisions that matter most.
