Building an AI Strategy

Core Concepts

What an AI strategy is

An AI strategy is a structured set of decisions that connects your business goals to the AI capabilities you build or adopt. It answers four questions: Where does AI create value for us? Why does that value matter to the business? What capabilities do we need? Who owns delivery and success?

Without these four answers, you have a list of tools, not a strategy.

The four components

Every AI strategy has four components. Each one depends on the one above it.

1. Business goals The outcomes your organization needs to achieve: revenue, retention, margin, speed, quality. These come from your existing strategy; AI does not invent new goals.

2. Friction points The specific problems, bottlenecks, or gaps that prevent you from reaching those goals. These are where AI has something to do. Without named friction, AI investments go looking for problems.

3. AI levers The specific ways AI can reduce or remove the friction. A lever is a concrete AI-powered action: automate a document, surface a recommendation, flag an anomaly. Not "use AI to improve onboarding" but "use AI to generate onboarding documentation from a project template in under two hours."

4. Capabilities The technical and organizational infrastructure required to operate the levers: models, data pipelines, integrations, workflows, and the people who maintain them.

The strategy hierarchy

These four components form a hierarchy. Every capability traces up to a lever. Every lever traces up to a friction point. Every friction point traces up to a business goal. If a capability cannot trace to a goal, it does not belong in the strategy.

Business Goal
  └── Friction Point
        └── AI Lever
              └── Capability Required

At X-company, one path through this hierarchy looks like:

Retain mid-market clients
  └── Client onboarding takes 5 days (too slow, creates early churn risk)
        └── Auto-generate onboarding documentation from project intake data
              └── GPT-4 integration with intake form + document template library

Business goal alignment method

The alignment method is how you move from a goal to a lever without jumping straight to tools.

1. State the business goal in measurable terms.
2. Map the friction: what is the specific obstacle between today's state and that goal?
3. Ask: what does AI do well that maps to this friction? (classify, generate, retrieve, summarize, predict)
4. Define the lever as a concrete AI action with a before and after state.
5. List the capabilities the lever requires.

This method prevents the most common strategy failure: picking a tool and working backward to a justification.

An AI strategy vs. an AI tool list

The difference is direction of reasoning.

A tool list starts with tools and asks "what can we use these for?" An AI strategy starts with goals and asks "what do we need AI to do?"

X-company's CTO had a tool list. Twelve tools, evaluated on capability, adopted on interest. No tool was tied to a retention target or a revenue goal. No one owned the outcome.

Key Points

• An AI strategy connects business goals to AI capabilities through friction points and levers
• Every capability in the strategy must trace to a business goal
• The alignment method moves top-down: goal first, tool last
• A tool list and a strategy look similar on a slide but produce entirely different outcomes
• Ownership and success metrics are not optional: they are what makes it a strategy

---

Actionable Takeaways

1. Pull up your organization's current AI initiatives. For each one, ask: what business goal does this trace to? If you cannot answer in one sentence, it is not part of a strategy yet.
2. Pick one business goal your team owns. Use the alignment method to map at least one friction point and one AI lever this week. Do not start with tools.
3. If you have a tool list masquerading as an AI strategy, name it honestly in your next planning conversation. The team cannot fix a problem that has not been named.
4. Write the strategy hierarchy for one initiative as a four-line chain: goal, friction, lever, capability. Share it with one other stakeholder and ask if the logic holds.
5. Assign an owner and a success metric to each AI lever before any capability work begins. If no one owns the outcome, the work will drift.

---

Practical Examples

Before: X-company's tool list

Six months ago, X-company's CTO shared this in the engineering Slack channel:

"AI tools we should evaluate: GitHub Copilot, Notion AI, ChatGPT Enterprise, Intercom Fin, Otter.ai, Glean, RunwayML, Midjourney, Clay, Grain, Tome, Perplexity."

Three tools were adopted. Copilot went to the engineering team. Notion AI was used by a few people in product. Intercom Fin was trialed for support. No tool had a named business goal. No tool had a success metric. By quarter end, no one could say whether any of them changed anything that mattered.

After: X-company's AI strategy (first draft)

The product and leadership team ran the alignment method against their three highest-priority business goals. They produced three initiatives, each with a named lever, owner, and success metric.

Goal 1: Reduce mid-market churn in the first 90 days

Goal 2: Increase engineering throughput without headcount growth

Goal 3: Shorten the sales cycle for enterprise prospects

The CTO's Copilot evaluation made it into the strategy as a component of Goal 2. The other eleven tools did not survive the alignment filter.

The strategy hierarchy in practice

A product manager at a 40-person fintech wanted to adopt AI for "customer support." Before running the alignment method, the plan was to deploy an AI chatbot.

After running the method:

• Business goal: reduce support cost per ticket by 30%
• Friction: 60% of tickets are password resets and account verification requests
• AI lever: self-serve identity verification workflow powered by an LLM-guided flow
• Capability: LLM with structured decision logic, integrated to the identity provider

The result was different from the original plan and more specific. The chatbot idea survived but was scoped to one ticket category with a clear success metric.

---

Implementation Workflow

Use this workflow to build the first draft of an AI strategy for your organization or product area.

1. List your top three business goals. Pull from your current OKRs, board priorities, or annual plan. Write each goal with a measurable outcome: not "improve retention" but "reduce mid-market churn in the first 90 days by 15%."

2. Map friction to each goal. For each goal, write one or two specific friction points that are preventing you from reaching it today. Be concrete: name the process, the delay, the gap, or the error rate. If you are unsure, talk to one person who works in that area before continuing.

3. Apply the AI alignment filter. For each friction point, ask: does this involve classifying, generating, retrieving, summarizing, or predicting something? If yes, write a candidate AI lever as a sentence: "Use AI to [action] so that [friction is reduced]."

4. List the capabilities each lever requires. For each lever, list the technical components needed: a model, a data source, an integration, a workflow change, a person to maintain it. Keep this list short; it is not an architecture diagram.

5. Build the hierarchy. Write the four-line chain for each initiative: goal, friction, lever, capability. If any link feels loose or forced, revise or discard the initiative. A chain that does not hold means the lever does not belong in the strategy.

6. Assign ownership and success metrics. For each initiative that survives step 5, name one owner and one measurable success metric. The metric must connect to the business goal, not the tool adoption rate.

7. Compare against your current tool list. Review any AI tools currently in use or under evaluation. For each tool, ask: which initiative does this serve? Tools that do not map to an initiative are either candidates for a future initiative or candidates for removal.

8. Write the one-page strategy summary. Three to five initiatives, each with a goal, friction, lever, capability, owner, and metric. This is the artifact you bring to leadership for alignment. It fits on one page because it is a set of decisions, not a roadmap.