Opportunity Research Toolkit — Interviews, Process Maps, & Data Checks

Purpose

This toolkit helps small teams run a short domain discovery that surfaces high-impact, non-obvious AI opportunities. Combine targeted stakeholder interviews, a lightweight process map, and quick data probes to avoid superficial brainstorming and identify ideas worth prototyping.

When to use

• Launching a one- to three-day discovery sprint in a team, department, or small organization.
• Validating whether a proposed automation or assistant idea has data and process fit.
• Finding systemic opportunities that individual contributors or managers may not see day-to-day.

Timebox & Roles

Suggested timebox: 1–3 days for a focused scan. Core roles:

• Facilitator: runs interviews, synthesizes findings.
• Domain SME(s): clarifies process detail and exceptions.
• Data Analyst: runs quick queries and verifies data availability/quality.

Quick workflow

1. Prepare: pick a target process or domain scope and schedule 3–6 short interviews.
2. Interview: use the 10-question script (20–30 minutes each).
3. Process map: draft a simple swimlane-style flow from the interviews and observation.
4. Data probe: run a few lightweight queries or checks to confirm feasibility and scale.
5. Score & prioritize: use the scoring template to rank surfaced ideas and recommend next steps.

Interview script (10 short questions)

Use 20–30 minute interviews. Ask the question, capture a brief story or example, and probe the impact and frequency.

1. What are the tasks you spend the most time on each week?

   Intent: reveal time drains and repetitive work. Probe: ask for an estimate of hours/week and variations.

2. Which tasks feel manual, repetitive, or error-prone?

   Intent: find low-hanging automation opportunities. Probe: examples of mistakes and their consequences.

3. When you have a difficult decision, where do you look for information?

   Intent: understand information flows and knowledge gaps. Probe: sources, search pain, and delays.

4. What frequently causes rework, escalations, or customer complaints?

   Intent: identify quality and exception hotspots. Probe: volume and typical root causes.

5. Are there tasks you do only because the system or process doesn’t support the right information at the right time?

   Intent: surface knowledge or integration gaps that an AI assistant or document intelligence could fill.

6. Tell me about a recent unusual or high-effort case.

   Intent: capture edge cases that reveal hidden complexity or systemic issues.

7. If you could remove one annoyance or speed up one task, what would it be?

   Intent: prioritize user-perceived value and quick wins.

8. What data or reports do you wish were easier to access or more reliable?

   Intent: identify data-readiness and reporting gaps for analytics or ML.

9. Which tasks require the most judgment or back-and-forth with other teams?

   Intent: find collaboration or decision-support opportunities where AI could surface context or suggestions.

10. What would success look like if this work were 30–50% faster or more accurate?

    Intent: link potential solutions to measurable outcomes (time saved, fewer errors, increased throughput).

Process mapping checklist (lightweight)

Create a 1-page map (or whiteboard photo) that shows steps, actors, systems, inputs/outputs, and pain points.

• Define scope: start and end events; keep it narrow.
• List actors and systems (human roles, applications, databases).
• Capture triggers, decision points, and handoffs.
• Mark where delays, rework, or inspections occur.
• Annotate frequencies (daily/weekly/monthly) and average durations where known.
• Note data artifacts (spreadsheets, reports, scanned documents, logs) and their owners.
• Identify exceptions and how they are resolved today.

Quick data probes (practical checks)

Purpose: verify volume, signal quality, and basic availability without a deep engineering effort.

• Sampling: pull a 30–90 day sample of relevant records to understand variety and edge cases.
• Completeness checks: measure missing values for key fields (percent missing).
• Value distribution: compute top N values for categorical fields and detect unexpected values.
• Time-series sanity: chart daily/weekly counts to find gaps or spikes that affect modeling/automation.
• Linkability: verify whether records share stable keys or identifiers for joining across systems.
• Example probe queries (conceptual):
  • Count events by day and by source system.
  • List top 10 most frequent error/exception codes.
  • Sample 50 recent cases with full fields to inspect data quality.

Template for scoring surfaced ideas

Score ideas on a 1–5 scale (1 low, 5 high). Multiply by weight and compute a weighted total.

• Impact (weight 30%): potential time saved, cost reduction, revenue upside, or quality improvement.
• Effort (weight 25%): engineering effort, integration complexity, and change management (higher score = lower effort).
• Data readiness (weight 20%): availability and quality of required data.
• Strategic fit (weight 15%): alignment with org priorities and regulatory constraints.
• Risk & compliance (weight 10%): safety, privacy, and operational risk (higher score = lower risk).

Interpretation: use the weighted score to categorize candidates as Prototype (top tier), Explore (investigate feasibility), or Park (defer). Capture 1–2 evidence points from interviews or probes that drove each score.

Deliverables & recommended next steps

• A one-page summary for each candidate use case: description, pain addressed, estimated benefit, evidence, and weighted score.
• Top 1–2 candidates recommended for quick prototypes (1–4 week POC).
• Short list of required data access and engineering tasks to start a POC.

Common pitfalls to avoid

• Chasing shiny technical ideas instead of measurable pain and frequency.
• Assuming clean, integrated data exists without quick probes.
• Over-scoping: prefer a small, testable slice rather than an enterprise-wide project first.

Tips for facilitation

• Keep interviews conversational and fact-seeking — ask for recent examples.
• Record short audio (with permission) or take structured notes tied to the scoring template.
• Use whiteboard photos for process maps and attach them to each candidate summary.

Artifacts you can reuse

• Interview notes template (one row per interview, include evidence bullets).
• One-page process map template.
• Scoring spreadsheet or simple interactive form to collect scores and compute totals.

How this toolkit can evolve

Consider converting the interview notes and scoring template into interactive forms so teams can capture and store responses, compare candidates over time, and build an organizational memory of discovery outcomes.