A useful AI strategy is short, specific and sequenced. It states where the company expects AI to change economics, what it will deliberately not attempt, and what has to be true operationally for any of it to work.
Merjora builds AI strategy from evidence rather than ambition: a quantified opportunity portfolio, an honest readiness view, an operating model that fits the company's size, and a sequenced roadmap with owners and review points. The strategy names what is out of scope, because a strategy that excludes nothing sequences nothing.
Why most AI strategies do not survive contact with delivery
They are written as ambition documents. They list capabilities rather than decisions, assume capacity that does not exist, and set no criterion for stopping anything.
The corrective is uncomfortable but simple: fewer initiatives, each with a named owner, a baseline, a review date and a stop rule.
What the strategy contains
The economic thesis: where AI plausibly changes cost, speed, capacity or revenue in this business
The quantified portfolio, with value ranges and confidence levels
Readiness: data, systems, skills, and the organisation's capacity to absorb change
Operating model: who decides, who builds, who owns the outcome
Governance: risk classification, review design, EU AI Act considerations, data protection
Sequence: what happens in which quarter, and what is explicitly deferred
Readiness assessed against what you will actually attempt
Generic readiness scores are of little use. Readiness is only meaningful relative to specific opportunities: this process needs this data at this quality, and this team needs to change this working practice. Assessed that way, readiness gaps become tasks in the roadmap rather than a maturity label.
Sequencing when capacity is the binding constraint
Mid-market companies rarely lack ideas; they lack the attention to run more than one or two meaningful changes at a time. Sequencing therefore weights organisational disruption alongside value, and deliberately places one fast-finishing improvement first so the measurement discipline is established before something hard is attempted.
What you receive
A short written strategy stating scope, exclusions and sequence
Quantified opportunity portfolio
Readiness gaps expressed as roadmap tasks
Operating model and decision rights
Governance and risk framework
Business impact
Planning cycles stop relitigating the same candidate projects
Capacity is allocated to a realistic number of initiatives
Governance is designed once rather than negotiated per project
Progress is visible because each initiative has a metric and a date
Where this is the wrong fit
Organisations wanting a maturity score rather than a set of decisions
Programmes where no exclusions are politically acceptable
Companies that have not yet quantified a single opportunity
Frequently asked questions
How is this different from an AI readiness assessment?
Readiness is one input. The strategy adds the economic thesis, the exclusions, the operating model and the sequence — the parts that determine what actually happens.
How long should an AI strategy document be?
Short enough that the executive team can hold it in mind. If it cannot be summarised on a page, it will not guide a decision under pressure.
Does the strategy cover EU AI Act obligations?
It covers classification and governance design at the level a mid-market company needs. Formal legal advice remains with your counsel.