AI strategy is the set of decisions about what AI to build, for which workflows, in which sequence, measured against which outcomes.
AI implementation is the work of building it: the Foundation, the training, the improvement loop.
Strategy without implementation is a plan. Implementation without strategy is building in the wrong order. Most companies need both, in the right sequence, from the same firm.
AI strategy — the decisions that determine the outcome
AI strategy for a $5M to $25M company is not a document. It is the set of decisions that determine whether the AI investment produces compounding returns or a well-intentioned plateau.
Workflow selection: which operational tasks AI handles. The wrong selection (workflows that are low-frequency, high-judgment, or insufficiently AI-amenable) produces weak adoption. The right selection (workflows that are high-frequency, high-frustration, and structurally amenable) produces the first successful uses that anchor the adoption habit.
Sequence: which workflow goes first, which goes second, and which waits until the Foundation is calibrated. The sequence determines the quality of the Foundation that each subsequent workflow builds on, the pace of adoption, and the readiness for Phase 3 automations.
Foundation design: which context documents make AI produce company-specific outputs rather than generic ones. The Foundation design decisions (which document types, what content, at what depth) determine whether AI outputs require 15% editing or 45% editing.
Measurement: which metrics prove whether the decisions are right and make every subsequent decision correctable. The company that measures the right things can identify within 30 days whether the workflow selection was correct and adjust before the wrong choice produces three months of weak adoption.
Restraint: what the company deliberately does not build: the workflows that are not AI-appropriate, the automations that are premature, the second AI tool that fragments the context before the first is producing compound improvement.
These decisions require sector-specific operational knowledge, implementation pattern experience, and the discipline to prioritise adoptability over impressiveness. This is AI strategy.
For a deeper look at how strategy consulting differs from other types of AI work, see what is AI strategy consulting — and for the specific deliverables that distinguish strategy from implementation, see what your AI consulting firm should deliver in 30 days.
AI implementation — the work that makes the strategy produce results
AI implementation converts the strategy decisions into an operational system that the team uses and that compounds over time. It has four components:
AI Foundations: the structured extraction of sector-specific vocabulary, communication conventions, and quality standards from the company’s function leads, and their translation into context documents that make AI produce company-specific outputs.
The Foundation that a sector-experienced practitioner builds in two weeks takes a sector-naive team four to six months to calibrate to the same quality through iteration.
Team training: the individual anchor workflow sessions (25 to 35 minutes per team member on real current work), the day-seven follow-ups, the peer advocacy structure, and the targeted resistance engagement that produces 70% or higher adoption within 30 days. This is labour-intensive, individual, and not replicable through a group session or a training guide.
Improvement loop: the weekly review of AI-assisted outputs, the context document updates, and the custom instruction refinements that make the AI system compound over time. This requires discipline (it is the first thing to be deprioritised when operational demands return), quality judgment (knowing which output quality gap reflects which context document gap), and the ability to transfer that judgment to the internal AI system owner.
AI system owner development: the capability transfer from the practitioner to the internal team member who will maintain the system after the engagement ends. This happens through observed practice alongside the practitioner, not through documentation or a handoff meeting.
Which one does your company need?
”We have no AI in place and no plan — where do we start?”
Start with the strategy decisions. Not a strategy document: the specific decisions about which workflows, which sequence, what Foundation elements.
These decisions should be made with the implementation in mind, which is why they are better made by the firm that will also implement rather than by a separate strategy consultant.
For most $5M to $25M companies: engage a firm that scopes both strategy and implementation together. The strategic decisions are made during the first two to three weeks of the engagement and the implementation begins immediately after.
”We have an AI tool deployed but no Foundation and poor adoption.”
This is an implementation problem, not a strategy problem. The tool is in place. The Foundation build and the adoption programme are what is needed.
Engage for Phase 1+2 implementation: Foundation build and team training.
”We have a Foundation and some adoption but we’re plateaued at month three.”
This is an implementation quality problem. Diagnose which failure cause applies and engage for the specific remediation, not a new strategy exercise.
”We have a well-built Foundation, a trained team, and a running improvement loop, and we want to build Phase 3 automations.”
This is a Phase 3 implementation problem. The strategy decisions are already made (correctly, given the Phase 1+2 outcomes). Engage for Phase 3 automation architecture and build.
The general principle
Strategy without a plan to implement it produces a roadmap. Implementation without the strategic sequencing decisions produces the wrong things built well.
The best outcome is a firm that covers both — making the strategic decisions with the implementation in mind, and executing the implementation against those specific decisions.
For a broader look at what separates strategy-first from tool-first approaches, see strategy-first vs tool-first AI consulting — and for a concrete sequence that combines both, the four phases of a Phos AI Labs engagement shows how strategy informs each implementation phase.
Common questions on AI strategy vs implementation
”Can the same firm provide both AI strategy and AI implementation?”
Yes, and this is the recommended structure for a $5M to $25M company. The strategic decisions are shaped by implementation experience (knowing what fails in practice), and the implementation quality is shaped by whether the strategic decisions were correct.
A firm that only delivers strategy hands off to an implementation team that does not know why the decisions were made.
”What comes first — strategy or implementation?”
Strategy first, then implementation. But the two phases should be compressed together in the first engagement weeks rather than treated as separate projects.
The strategic decisions (workflow selection, sequence, Foundation design) should be made in weeks one and two. The Foundation build (the first implementation deliverable) should begin in week two. The gap between strategy and implementation should be days, not months.
”What deliverables should I expect from an AI strategy engagement?”
Not a document. The deliverables of an AI strategy engagement are decisions: the specific workflow selection (named, prioritised, sequenced), the Foundation design specification (which documents, what content, built to what quality standard), the measurement framework (four operational metrics, tracked weekly).
Also the restraint decisions (what is not being built and why).
A firm that delivers a deck and a roadmap at the end of the strategy engagement has delivered a plan. A firm that delivers decisions and begins the Foundation build in the same engagement has delivered AI strategy.
Phos covers both — the strategy decisions and the implementation work — in a single embedded engagement
AI strategy is the decisions: what to build, in what sequence, for which workflows, measured against which outcomes. AI implementation is the work: Foundation build, team training, improvement loop, AI system owner development.
The two are sequential — strategy first, implementation second — but they are most effectively executed by the same firm, because the strategy decisions are shaped by implementation experience, and the implementation quality is shaped by whether the strategy decisions were correct.
Phos AI Labs makes the strategic decisions and executes the implementation in a single embedded engagement. Thirty minutes, no deck. Start here.
Related articles