01 /AI use case evaluation

Know which AI use cases deserve your next quarter.

Evaluate AI use cases on three dimensions and turn them into a sequenced implementation plan. Everything runs in your browser, so nothing leaves your machine.

02 /Definition

What is a use case?

A use case is one specific business task or process step where AI is applied to produce a measurable outcome. It has a trigger, a bounded task, and a number that changes when it works. If you cannot name that number, you do not have a use case yet.

A use case is not a technology, not a department, and not an ungrounded wish. "We want to use AI in finance" is a direction; "extract line items from supplier invoices so the team stops re-typing them" is a use case.

TRIGGERAn invoice arrivesTASKExtract and checkMEASURABLE OUTCOME20 hours saved monthly

03 /Scope

Horizontal or vertical

Horizontal

Enterprise-wide copilots and chatbots.

Easy to deploy and fast to scale, because everyone gets the same tool. The benefit is spread thinly across employees, which makes it hard to trace in business results.

Vertical

Embedded in one specific function or process.

Higher direct economic impact, because the work itself changes. It is also harder to land: fewer than 10% of deployed use cases make it past the pilot stage.

HORIZONTAL — WIDE, SHALLOWMany people, thin benefit eachVERTICAL — NARROW, DEEPFew people, deep benefitBoth blocks cover the same area: the same total value,distributed very differently.

Source: McKinsey & QuantumBlack, Seizing the Agentic AI Advantage, June 2025

04 /Solution class

Which solution class fits

Three questions come before any scoring: how far the logic can be written as rules, how many dependent steps it takes, and what the task has to produce. They do not reject use cases, they route them.

Most cases that fail an agent check are still perfectly valid AI use cases of a different class — a prediction model, a grounded retrieval setup, or a drafting assistant. Only fixed rules end the AI conversation, because conventional automation does that job better.

USE CASEFIT CHECKFixed rules, same input same outputConventional automationNO AI REQUIREDOutput is a prediction or a categoryClassical MLGrounded answer, single stepRetrieval (RAG)New content, one or a few stepsGenerative assistantMulti-step across systems, or an executed actionAgentic workflow

Adapted from the agent decision tree in the Microsoft Cloud Adoption Framework for AI agents. The five solution classes are Pilotic's own extension.

05 /Method

How the evaluation works

Every use case that passes the gates is scored on three dimensions from 0 to 100. Each dimension has its own sub-criteria, so a score can always be traced back to the answers that produced it.

Business impact

  • Strategic alignment
  • Business value
  • Change timeframe

Technical feasibility

  • Implementation risk
  • Safeguards
  • Technology fit

User desirability

  • User pain
  • Value proposition
  • Change resistance

The three dimensions and their sub-criteria follow the Microsoft Cloud Adoption Framework for AI agents. Weighting, thresholds and the quadrant model are Pilotic's own.

Impact and feasibility together place the use case in one of four quadrants; desirability tells you how hard the adoption will be once it is built.

Strategic Initiative — High long-term value, high complexity. Requires dedicated roadmap, sponsorship, and platform foundations.STRATEGICINITIATIVEQuick Win — High value, low barrier. Prioritize first for immediate momentum and rapid ROI.QUICK WINDeprioritize — Low value, high friction. Archive or avoid; return on effort is insufficient.DEPRIORITIZEFill-in — Low complexity, modest value. Good for training, secondary backlog, or quick wins when capacity permits.FILL-INHIGH IMPACTLOW IMPACTHIGH FEASIBILITYFEASIBILITYIMPACT

Four quadrants

Hover or focus a quadrant to read what it means for sequencing.

06 /Sequencing

From score to plan

A score is only the entry point. Assess establishes whether the use case is ready to enter a four-stage cycle: Envision, Align, Launch, and Scale.

The plan considers six organisational perspectives across that cycle rather than treating implementation as a technology task alone.

ENTRYASSESSScore the use caseDELIVERY CYCLEENVISIONShape the target flowALIGNSecure sponsor and ownerLAUNCHPilot with real usersSCALEStandardise and expand
  • Business
  • People
  • Governance
  • Platform
  • Security
  • Operations

Stages and perspectives: AWS Cloud Adoption Framework for AI (CAF-AI)

07 /Start

Take one use case through the model.

The assessment takes about three minutes. You will leave with three scores, a quadrant, and the gates that apply.

Start assessment