Business Case and Prioritisation

Guide · Enterprise AI Solution EngineeringPage 3 of 18Overview → Discovery → Business case

Executive summary

Not every opportunity should progress to delivery. Qualify use cases for suitability, score value, feasibility, risk and adoption, and build honest business cases with best / likely / downside scenarios—protecting scarce capacity from impressive systems nobody adopts.

When to use this page

• Use after discovery artefacts exist and candidates need ranking.
• Use before funding architecture or prototype work.
• Return to Discovery if problem statement or baselines are still missing.

Executive view

Demand transparent scoring, scenario ranges and benefits separated by certainty—not a single optimistic number.

Decision required: Fund, defer, or invest in foundations first?

Technical view

Complete the rules-vs-AI suitability test and evidence notes behind each score.

Label quick wins, strategic bets, foundations and deferrals so architecture work targets the right portfolio.

Owners

Owner · Business sponsor with AI Solution Engineer support

Contributors · Finance, product, architecture, risk, change, delivery lead

Not every opportunity should progress to delivery. Prioritisation protects scarce engineering capacity and reduces the risk of building impressive systems that nobody adopts.

AI use-case qualification

Evaluate candidate use cases across:

• Strategic alignment
• Business value
• Technical feasibility
• Data readiness
• User desirability
• Risk
• Time to value
• Scalability
• Organisational readiness

AI suitability test

Before proposing AI, determine whether the task involves large volumes of information, pattern recognition, unstructured data, repetitive judgement, knowledge retrieval, prediction, personalisation, optimisation, content generation or multi-step coordination.

Then ask:

• Could a deterministic rule solve this more safely?
• Could workflow redesign solve the problem?
• Could traditional automation solve it?
• Is AI materially better than the alternatives?

If the answer is “rules would do,” choose rules. AI is a means, not a goal.

Decision required

If rules, workflow redesign or traditional automation would solve the problem safely, do not fund an AI build—record the deferral rationale.

Use-case prioritisation matrix

Score each use case from 1 to 5.

Dimension	What to score
Value	Revenue, cost reduction, productivity, risk reduction, customer value, strategic relevance
Feasibility	Data readiness, technical complexity, integration effort, model availability, delivery capability
Risk	Data sensitivity, regulatory exposure, decision impact, explainability, security exposure
Adoption	User readiness, process fit, leadership support, change complexity, trust requirements

A simple ranking heuristic:

Priority score = Value + Feasibility + Adoption − Risk

Adjust weights to organisational strategy, but keep the method transparent.

Example priority categories

Category	Meaning
Quick wins	High value, low complexity, manageable risk
Strategic bets	High value with significant architecture, data or change requirements
Foundations	Capabilities that unlock many future use cases (identity, model gateway, vector platform, evaluation, governance)
Defer	Low value, high complexity or unacceptable risk

Building the business case

Include problem, baseline, target outcome, proposed intervention, investment, expected benefits, risks, dependencies, delivery approach and measurement plan.

Benefit categories

Financial — revenue growth, cost avoidance, reduced headcount growth, lower third-party spend, reduced error cost.

Operational — faster cycle time, increased throughput, reduced manual effort, improved availability, better consistency.

Customer — faster response, better personalisation, improved resolution, increased satisfaction, reduced abandonment.

Risk and compliance — fewer policy breaches, better audit evidence, improved monitoring, faster issue detection, reduced regulatory exposure.

Business case formula

A simple annual-value estimate:

Annual value = Productivity value + Revenue value + Risk reduction value − Annual operating cost

Where:

Productivity value = Hours saved × Loaded hourly cost × Adoption rate

Avoid presenting theoretical savings as guaranteed cash savings. Separate:

• Capacity released
• Avoided future hiring
• Direct cost reduction
• Revenue uplift
• Risk-adjusted benefit

Risk

Single-point estimates and counting 100% of theoretical hours as cash savings create overconfident cases that fail after launch.

Control

Require best / likely / downside scenarios and label benefit certainty before executive approval.

Scenario planning

Develop three scenarios.

Scenario	Assumptions
Best case	High adoption, strong model quality, rapid scale
Likely case	Moderate adoption, normal delivery constraints, realistic benefit capture
Downside case	Low adoption, higher operating costs or delayed integration

This prevents overconfident business cases and gives executives a decision with eyes open.

Case study: financial-services assistant

Factor	Assessment
Value	High — search time and first-contact resolution
Feasibility	Medium — knowledge is fragmented but exists
Risk	Medium-high — product advice and regulated language
Adoption	High if embedded in the service desk, low if a separate portal

Classification. Strategic bet with a foundation dependency on approved knowledge ingestion, access-controlled retrieval and evaluation.

Likely-case benefit framing. Capacity released for advisors (minutes saved × volume × adoption), not headcount cuts on day one.

Output

Qualified use-case scores with evidence, category labels (quick win / strategic bet / foundations / defer), business case with benefit categories, and best / likely / downside scenarios plus an explicit ask for the decision forum.

Common failure modes

• Ranking by executive enthusiasm instead of evidence
• Counting 100% of theoretical hours as cash savings
• Ignoring foundations that make later use cases cheaper
• Skipping the “rules vs AI” test
• Single-point estimates without downside scenarios

Solution Engineer checklist

Solution Engineer checklist

• Suitability test completed for each candidate
• Scores recorded with evidence notes
• Quick wins / strategic bets / foundations / defer labelled
• Benefits separated by category and certainty
• Best / likely / downside scenarios documented
• Explicit ask prepared for the decision forum

Practical exercise

Score three candidate use cases with the matrix. Force yourself to defer at least one. Write the one-paragraph rationale an executive would need to accept that deferral.

Key takeaways

• Prioritisation protects capacity; not every opportunity should proceed.
• Score value, feasibility, risk and adoption with transparent evidence.
• Separate benefit types and certainty; always show downside scenarios.
• Foundations often unlock more value than isolated demos.

What to do next

1. Continue with Enterprise AI architecture.
2. Frame the executive ask using Consulting and executive communication.
3. Apply VALUE quality gate before sending the case.