AI Data Foundation Plan

AI Data Foundation Plan

Use When

• Use this skill when the requested outcome is specifically a implementation plan and the supplied brief falls within ai data foundation plan.

Do Not Use When

• Use ai-readiness-diagnostic when its narrower output is the real deliverable; do not use this skill as a generic substitute.
• Do not use it to publish, send, spend, alter a live account, or make unsupported legal, platform, performance, or certification claims.

Required Inputs

Capability and Permission Boundaries

Drafting is permitted within the supplied brief. Publishing, sending, spending, changing live accounts, or claiming certification requires separate explicit authority. Minimum capabilities are read access to supplied files and search across the authorised evidence set. Use only the files, tools, accounts and evidence made available for the engagement, expose every unassessed check, and obtain explicit authority before any mutation.

Degraded Mode

Fallback: if files, network access, platform data, language review or production tools are unavailable, return the narrowest useful qualified implementation plan; mark unavailable checks not assessed and never convert them into a pass.

Decision Rules

Workflow

1. Confirm the exact implementation plan, consumer, market, channel and approval boundary; route to ai-readiness-diagnostic if it is the closer match.
2. Inventory supplied facts, source provenance, constraints and missing inputs; stop if the objective, audience or authority is unknowable.
3. Select the domain method and record the material decision behind it before drafting.
4. Produce the smallest complete implementation plan; keep facts traceable and placeholders visibly unresolved.
5. Test the result against the decision table, domain quality criteria and anti-slop gate; recover by narrowing or qualifying unsupported portions.
6. Deliver the artefact with evidence, assumptions, unassessed checks and the next approval or verification step.

Outputs

Evidence Produced

Quality Standards

• Preserve the domain guidance and East African market context below; replace it only when the requester names another market.
• Use British English unless the target language or market requires otherwise, and verify names, figures, quotations and platform rules before use.
• Make the key choice visible, cover failure and edge cases, and keep the result ready for its named consumer.
• Run the repository's anti-ai-slop ship gate; a blocking factual, cultural, safety or permission defect stops release.

Anti-Patterns

• Writing before the objective and audience are known. Fix: stop and obtain the missing brief fields.
• Reusing a neighbouring skill's template because the headings look similar. Fix: route by the requested implementation plan, not vocabulary overlap.
• Adding a price, result, quotation, platform limit or cultural claim without a traceable source. Fix: verify it or qualify/remove it.
• Treating missing access, evidence or native-language review as approval. Fix: mark the check not assessed and narrow the result.
• Publishing, sending, spending or changing a live account from drafting authority alone. Fix: obtain explicit action-specific authority and retain the approval record.

References

• ai-readiness-diagnostic is the nearest routing comparison for this skill.
• Repository agent guide defines the engine-wide market, safety and anti-slop gates.

Purpose

Produce a structured data foundation action plan grounded in Canvas Step 1 of Venkatesan and Lecinski's The AI Marketing Canvas (2nd ed., 2026). AI is only as good as the data it is trained on. A client with poor data will get poor AI outputs regardless of which tools they purchase. The most common reason AI marketing fails at Step 2 is not tool quality — it is data quality.

The four customer moments that AI marketing serves — Acquisition, Retention, Growth, and Advocacy — each require specific, clean, structured data to function. This skill delivers the audit, schema, and plan to build that foundation.

Primarily relevant for larger East African clients (banks, NGOs, telecoms, universities) with existing data but no structured marketing use of it.

After completing this plan, refer the client to ai-readiness-diagnostic to confirm their Canvas step progression, and to ai-vendor-evaluation when selecting CRM or AI tools.

Required Inputs

Ask for the following before generating any output:

1. Client business name — trading name and legal entity if different
2. Industry — financial services, NGO, telecom, education, retail, etc.
3. Country and city — defaults to Uganda/Kampala if not specified
4. Organisation size — approximate staff count and customer/contact count
5. Current data sources — list every source, including spreadsheets, WhatsApp contact lists, paper records, and legacy software
6. Primary marketing goal — what AI capability the client wants to unlock (e.g., segmented email, churn prediction, personalised WhatsApp messaging)
7. Current Canvas step — from ai-readiness-diagnostic; if not run, ask the client to describe their AI marketing activity to date
8. Existing CRM or data tool — name and version, or "none"

Step 1 — Data Asset Inventory

Map all current data sources across five categories. For each source identified, record the four attributes listed below.

Five data categories to map:

• Contact data — customer name, phone number, email address, location
• Transaction data — purchase history, amounts (UGX), dates, payment method (Mobile Money / bank transfer / cash)
• Engagement data — social media interactions, email opens, website visits, WhatsApp message responses
• Behavioural data — content consumed, products viewed but not purchased, pages visited, app activity
• Feedback data — survey responses, Google/Facebook reviews, complaint records, WhatsApp conversation threads

For each data source, record:

Present the inventory as a table. Flag any source rated below 50% completeness or updated less than monthly as a priority gap.

Step 2 — Data Quality Assessment

Rate each data source on four dimensions. Target scores are noted; flag any source below target.

EA-specific data problems to check for explicitly:

• SIM swapping — customers in Uganda frequently change SIM cards; a contact may have 2–4 phone numbers on file, not all current
• Mobile Money numbers — MTN MoMo and Airtel Money numbers may differ from the customer's primary voice SIM; record both separately
• Name variations — Luganda and English name forms of the same person are often recorded inconsistently (e.g., "Nakato Sarah" vs "Sarah Nakato" vs "S. Nakato"); flag duplicates
• Unrecorded cash transactions — particularly common in retail and financial services; a customer's LTV is systematically underestimated if cash sales are not captured
• WhatsApp as de facto CRM — WhatsApp contact lists are valuable but unstructured; they must be exported and cleaned before any AI tool can use them

Produce a quality scorecard table with one row per data source and columns for each dimension. Assign RAG status (Red / Amber / Green) per cell.

Step 3 — Minimum Viable Customer Schema

Design the client's minimum viable customer record — the specific fields required for their stated marketing goal. Do not produce a generic schema; tailor fields to the client's industry and goal.

Core fields — required for any AI use:

Behavioural fields — required for Retention and Growth AI moments:

Segmentation fields — required for Acquisition AI moment:

For each field in the schema, note: (a) whether it currently exists in a data source, (b) the source location, and (c) what action is needed to populate it.

Step 4 — 90-Day Data Foundation Plan

Produce a task-level plan with named deliverables per 30-day block. Assign a responsible role (e.g., Marketing Manager, IT Officer, Data Analyst) to each task. Milestones must be concrete and measurable — not vague objectives.

Days 1–30: Audit and Clean

• Week 1 — Complete the data asset inventory (Step 1). Produce the inventory table and share with senior management for sign-off.
• Week 2 — Run the data quality assessment (Step 2) on the top three data sources by customer record count. Produce the RAG scorecard.
• Week 3–4 — Identify and close the single most damaging data gap. For most EA clients this will be either: (a) deduplicating phone numbers, or (b) exporting and structuring WhatsApp contacts.
• Week 4 — Select a CRM or data management tool (see recommendations below). Complete tool sign-up and configure the minimum viable schema as the record template.

Days 31–60: Consolidate

• Migrate all customer records from existing sources into the chosen CRM.
• Apply the minimum viable schema to all migrated records; flag incomplete records for follow-up.
• Write and circulate a data entry standard — one-page document specifying exactly how new customer records are created (format, required fields, who enters data).
• Set up a consent capture process (see Uganda DPA 2019 requirements below). Every new customer record created from Day 31 onward must have documented consent.

Days 61–90: Connect and Test

• Connect the clean data source to the first AI or automation tool identified in the client's marketing goal (e.g., Mailchimp for email segmentation; Africa's Talking for SMS broadcast; HubSpot workflows for lead nurturing).
• Run the first segmented campaign using the clean data.
• After the campaign, measure: Was the data clean enough to run without manual correction? What fields were missing or incorrect?
• Produce a data health score: calculate the % of records meeting the full minimum viable schema. Set a target of 70%+ by Day 90, 90%+ by Month 6.

Uganda Data Protection and Privacy Act 2019 — Compliance

Address the following requirements explicitly in the plan output.

Obligations relevant to marketing data:

• Obtain explicit, informed consent before storing any personal data.
• Inform customers of: what data is collected, the purpose, how it is used, how long it is retained, and who it may be shared with.
• Honour opt-out requests within 48 hours.
• WhatsApp broadcast lists require opt-in from each recipient — do not add contacts without consent.
• Appoint a Data Protection Officer if processing data at scale (>1,000 records is a reasonable internal threshold for larger clients).

Consent capture template — include this verbatim in the output, adapted to the client's name and service:

[CLIENT NAME] — Data Consent Notice

By sharing your contact details with us, you agree that [Client Name] may store and use your name, phone number, and purchase history to:

• Send you relevant updates, offers, and service information via WhatsApp, SMS, or email
• Improve our products and services based on your feedback
• Contact you about your account or orders

Your data will not be sold or shared with third parties without your separate consent. You may withdraw consent at any time by messaging "STOP" to [WhatsApp number] or emailing [email address].

Data is retained for [X] years or until you request deletion. This notice complies with the Uganda Data Protection and Privacy Act 2019.

Adapt the retention period and contact channels to the client's actual practice. Translate into Luganda or Swahili if the client's primary customer base is not English-speaking.

Recommended CRM Tools for the EA Market

Recommend one tool from this table based on client size and budget. Cross- reference with ai-vendor-evaluation for a full tool selection process.

Selection guidance:

• If the client has fewer than 200 contacts and no budget: Google Sheets + AppSheet.
• If the client has 200–2,000 contacts and some budget: HubSpot or Zoho free tier.
• If the client is an NGO or CSO: apply for Salesforce Nonprofit first.
• If the client has complex relational data needs: Airtable.

Quality Criteria

Assess the output against these criteria before delivering to the client:

• The data asset inventory is exhaustive — it includes WhatsApp contacts, paper records, and POS systems if present, not only digital CRM sources.
• All four data quality dimensions (completeness, accuracy, consistency, currency) are assessed and scored in a RAG table, not described generically.
• The minimum viable schema is tailored to the client's industry and stated marketing goal — it is not a copy-paste generic template.
• The 90-day plan contains named, measurable tasks for each 30-day block with an assigned responsible role; no block contains only vague objectives.
• Uganda DPA 2019 compliance is addressed with the consent capture template adapted to the client's name, channels, and retention period.
• The CRM recommendation is specific to the client's size and budget, with a brief rationale, not a list of all options.
• EA-specific data problems — SIM swapping, Mobile Money number duplication, Luganda/English name variations, unrecorded cash transactions — are explicitly identified where relevant to the client's data sources.
• The output closes with a clear statement of the data health score target and what Canvas step the client will be ready to attempt once that target is met.

Output Format

Deliver the plan in five clearly labelled sections matching the steps above:

1. Data Asset Inventory (table)
2. Data Quality Assessment (RAG scorecard table)
3. Minimum Viable Customer Schema (table with current-state notes)
4. 90-Day Data Foundation Plan (three 30-day blocks, task-level)
5. Compliance and Consent (DPA 2019 notice, adapted)

Close with a one-paragraph Next Step statement confirming: what Canvas step the client is currently at, what the 90-day plan will unlock, and which skill to invoke next (ai-readiness-diagnostic to re-score, or ai-marketing-canvas-assessment to begin full canvas development).

Output is a structured text document suitable for sharing with the client's senior leadership team as a standalone briefing paper.

For data-product ownership, lineage, freshness, audience boundaries, and AI marketing controls, load data product and AI foundation principles.