How We Use AI
Windmill Strategy uses AI as a tool, directed by experienced strategists, never as a replacement for expertise, judgment, or accountability. Because our clients are technical B2B companies where accuracy and buyer trust matter, we hold our AI use to specific standards. This page explains what those are, so you always know how your work is being done.
Our guiding principle
AI is best at structure, synthesis, and acceleration, organizing information, summarizing large datasets, drafting from source material, and reformatting. It is not the strategist. Final recommendations come from human judgment that accounts for your business model, sales process, competitors, constraints, and goals. A lot of marketing has gotten faster and worse at once: generic content assembled by a tool and published without anyone who understands the business. We built our approach around the opposite conviction: the strategy and the subject-matter expertise are the valuable part, and those stay human. AI helps us execute more efficiently, not skip the thinking.
What we use AI for
We use established AI tools to accelerate specific, bounded tasks: organizing and summarizing analytics exports (SEMrush, DataForSEO, GA4, Google Search Console, Google Keyword Planner), synthesizing meeting transcripts into action items, drafting from approved messaging and source material, and structuring research. A strategist provides the inputs, and reviews the output against known sources for accuracy.
What we won’t let AI do
- Invent facts about your company. Our internal prompts are built to protect against AI fabrication of specifications, certifications, capabilities, industries served, pricing, proof points, case study results, quotes, or dates. Where a claim would require a fact the tool wasn’t given, our process forces it to flag the gap for a human to fill, not guess.
- Replace your subject-matter experts. Content built in a vacuum is generic by definition, and neither AI search nor your buyers reward it. We interview your experts; AI helps us organize and draft what they tell us, not invent it.
- Overstate what the data shows. Our team tags every analytical claim by how well the data supports it, and we use measured language (“likely,” “appears to”) rather than false certainty. Where a number is directional or estimated, as AI-visibility measurement often is, we say so plainly.
- Ship unreviewed. Nothing written entirely by AI leaves Windmill without human development and review for accuracy, quality, and fit.
How we protect your data
We are deliberately cautious about what goes into any AI tool. We do not enter one client’s information into another client’s work, and we never knowingly enter credentials or login details into an AI tool. Where we use AI tools on client work, we use business or enterprise tiers configured so your inputs are not used to train the underlying models, and we keep each client’s work separated. Our full data-handling commitments are set out in our client agreements, and we’re glad to discuss any specific requirements you have.
Human accountability
Responsibility can’t be handed to a machine. Every deliverable that leaves Windmill has a person behind it who reviewed it for compliance with our standards. If you ask whether AI was involved in something, we’ll answer honestly, we won’t present AI output as work we personally produced when we haven’t.
What if you’d prefer AI not be used on your work?
Some organizations, particularly in regulated industries, or with specific brand or legal requirements, prefer that generative AI not be used in creating their deliverables. If that’s you, tell us and we’ll plan and scope your work accordingly. Our value has always been the strategy and expertise; the tools we use to execute are a conversation we’re happy to have.
Why we publish this
We keep this page public because the same honesty we bring to measurement, we bring to our own methods. If a vendor won’t tell you how they use AI, or how they arrive at the numbers they report, that’s worth asking about.