Skip to main content

Select an AI Product

This playbook helps choose the AI product or delivery track for an initiative. The choice matters because it determines the team, delivery stages, checks, architecture, and impact expectations.

When to use

  • the initiative has passed initial intake;
  • the problem is clear but the solution route is not;
  • several technical routes are possible;
  • a one-off experiment must be separated from a reusable product route;
  • similar initiatives exist in the portfolio.

Inputs

  • business problem;
  • process and users;
  • expected impact;
  • data type;
  • AI action type: hint, draft, recommendation, automation;
  • security constraints;
  • similar initiatives;
  • existing AI products and platforms.

Selection logic

Task typeLikely product
Answers over documents, policies, knowledge baseKnowledge assistant
Text generation, editing, summarizationCorporate LLM assistant
Classification, scoring, prediction over dataML platform
Data extraction from documentsDocument AI / OCR
Developer assistance and code analysisCode Agent
Automation across systemsAutomation AI / agentic workflow
Portfolio analytics, KPIs, impactAnalytics layer / BI

Decision checks

  1. Check whether the required data exists and can be used.
  2. Classify the decision impact: hint, draft, recommendation, automatic action, critical impact.
  3. Check reuse: can an existing AI product be extended?
  4. Record alternatives and the reason for choosing the route.
  5. Assign the product owner and next delivery stage.

Possible outcomes

  • existing AI product selected;
  • initiative merged with an existing product;
  • new AI product or catalog extension required;
  • initiative stays in assessment until data is checked;
  • initiative rejected because there is no data, owner, or acceptable risk route.

Common mistakes

  • choosing an LLM for every task;
  • starting a new product instead of extending an existing one;
  • ignoring automatic-action risk;
  • selecting a product without a product owner;
  • deciding before checking data;
  • mixing the business initiative funnel with the product delivery track.