AI Initiatives
What is an AI initiative
An AI initiative is a managed portfolio unit through which a business need moves from idea to implemented solution and confirmed impact.
An initiative describes:
- what problem needs to be solved;
- who it matters to;
- what impact is expected;
- which AI product can implement it;
- what data, documents, integrations, and approvals are needed;
- who is responsible for implementation;
- how the result will be validated.
If AI products answer "what do we solve with?", AI initiatives answer "what business task are we solving and why?".
Why separate AI initiatives as their own entity
Companies often mix three different levels:
- Idea — a raw request or hypothesis.
- AI initiative — a structured business need that can be evaluated, prioritized, and moved through the funnel.
- AI product — a tool, platform, or service through which the initiative is implemented.
If these levels are not separated, chaos appears: business brings an idea but ownership is unclear; the AI team starts a pilot but impact is not recorded; IT waits for requirements while business waits for a solution; duplicate requests are implemented by different teams; new AI products appear for isolated cases even when an existing platform could solve them; after a pilot, no one knows who owns adoption and support.
That is why the AI initiative is needed as a management container that brings together problem, value, implementation route, owners, artifacts, decisions, and impact.
How an AI initiative differs from an AI product
AI product and AI initiative are related, but they are different entities.
| Entity | Meaning | Example |
|---|---|---|
| AI product | Reusable platform, tool, or service | LLM, RAG, ML platform, code agent |
| AI initiative | Specific business task solved with AI products | Request processing automation, knowledge base search, metric forecasting |
| Delivery track | Implementation route depending on selected product | Prompt setup, knowledge base assembly, model training, service integration |
One AI product can be used in many initiatives. For example, RAG can support policy search, operator support, legal assistance, employee onboarding, and project documentation analysis.
Conversely, one initiative may require several AI products: LLM for text generation, RAG for fact retrieval, workflow automation for routing, BI/DWH for data, and an applied AI service as the user interface.
Logic of an AI initiative
An AI initiative should start not with "let's implement AI," but with a business situation.
Correct logic:
- We have a problem or opportunity.
- We understand who it matters to.
- We hypothesize the potential impact.
- We assess which AI products can help.
- We choose an implementation route.
- We validate the result and record impact.
Wrong logic:
- We have a new LLM.
- Let's find where to apply it.
- Let's launch a pilot.
- Later we will figure out whether it had impact.
Sometimes new AI products do create demand and ideas. But even then, the initiative should be framed through a business problem and expected result.
What goes into an AI initiative description
A minimal initiative card should include:
| Block | What is recorded |
|---|---|
| Problem | What currently works poorly, slowly, expensively, or riskily |
| Business sponsor | Who owns the problem and wants the result |
| Users | Who will use the solution |
| Expected impact | Expected benefit: money, time, quality, risk, speed |
| Solution hypothesis | How AI can help |
| Potential AI product | What can implement it: LLM, RAG, ML, agent, automation |
| Data and sources | What data, documents, or systems are needed |
| Constraints | Security, compliance, architecture, access, data quality |
| Initiative lead | Who moves the initiative through the funnel |
| Status | Where the initiative is in the business funnel |
| Next step | What should happen next |
| Impact validation method | How we know the initiative produced a result |
This should not become a heavy project passport at the start. Early on, a short card is enough; it gets enriched as the initiative moves through the funnel.
Lifecycle of an AI initiative
An AI initiative moves through the business funnel.
Simplified:
Idea → Evaluation → Delivery → Awaiting impact → Closed
Idea
The initial request is recorded. Main question: is there a potential business problem worth exploring?
Evaluation
Problem, impact, users, data, constraints, and possible AI products are clarified. Main question: should the initiative move into implementation and through which AI product?
Delivery
The initiative is implemented through the appropriate delivery track. Main question: can we get a working solution that can be tested with users?
Awaiting impact
The solution is implemented or piloted, but impact still needs confirmation. Main question: is the solution used and does it deliver the expected benefit?
Closed
The initiative is closed when impact is confirmed, the solution is handed to operations, scaled, or rejected for clear reasons. Main question: what outcome was recorded and what did the company learn?
When an idea becomes an AI initiative
Not every idea automatically becomes an initiative.
An idea becomes an AI initiative when it has:
- a clear business problem;
- an interested sponsor;
- potential users;
- an impact hypothesis;
- preliminary understanding of how AI can help;
- an owner of further movement;
- a status in the business funnel.
Before that, it is only a raw request, hypothesis, or suggestion.
Connection to the AI product portfolio
The AI initiative portfolio and AI product portfolio must be connected.
AI initiatives show which business needs exist in the company. AI products show which reusable capabilities the company uses to solve them.
This connection matters for three reasons.
1. Do not create a separate solution for every case
If every initiative becomes a separate product, the company quickly gets a zoo of pilots, services, and prototypes.
The right first question is: can this initiative be solved through an existing AI product?
If yes, the initiative goes through that product. If no, the company decides whether to build a new product, improve an existing one, or reject the initiative.
2. Develop AI products based on real demand
Initiatives show which capabilities business needs repeatedly. Many document-search initiatives strengthen the RAG direction. Many development tasks strengthen the code agent. Many forecasting tasks strengthen the ML platform. Many text and document tasks strengthen the LLM loop.
That way, AI product roadmaps are built from recurring business needs, not from the abstract desire to implement a trendy tool.
3. Choose the right delivery track
Different AI products require different implementation paths.
An LLM task may need prompt selection, quality check, and user training. A RAG task needs documents, knowledge base quality assessment, source upload, answer validation, access setup, and support ownership. An ML task needs data, metrics, training, validation, integration, and monitoring.
Therefore an initiative must be connected to the product not only by meaning, but also by delivery logic.
Key statuses of an AI initiative
Every initiative needs a clear status.
| Status | Meaning |
|---|---|
| New idea | Request received but not yet analyzed |
| In evaluation | Problem, value, feasibility, and product route are being clarified |
| In delivery | Implementation is in progress through the selected AI product |
| Awaiting impact | Solution is used, but impact is not yet confirmed |
| Completed | Impact confirmed or final result recorded |
| Rejected | Initiative is not viable, not possible, or not prioritized |
| In operations | Solution is handed over to stable use and support |
Statuses are not for reporting bureaucracy. They show where an initiative is stuck, who owns the next step, which initiatives need decisions, where data/resources/approvals are missing, and which initiatives already deliver impact.
Criteria of a good AI initiative
A good AI initiative answers several questions:
| Question | What should be clear |
|---|---|
| Why? | What business problem or opportunity exists |
| For whom? | Who the user and sponsor are |
| Through what? | Which AI product potentially fits |
| On what? | What data, documents, or systems are needed |
| How will we validate? | How quality and impact will be assessed |
| Who leads? | Who manages the initiative |
| What is next? | What next step and gate decision are needed |
If these questions cannot be answered even preliminarily, the initiative is not ready for delivery.
Core idea
An AI initiative is not "an idea to use AI" and not "a small AI project".
It is a managed business need that moves through a single loop from idea to impact.
It connects business problem, expected value, AI products, delivery track, accountable participants, artifacts, stage gates, and result confirmation.
That is how the company stops managing AI as random experiments and starts managing it as a portfolio of business change.