How a Heartwood Decision Brief gets made

Heartwood produces structured IT Decision Briefs for mid-market and SMB leaders who do not have a full-time CIO and cannot justify a Gartner retainer or a six-figure consulting engagement. This page documents how a brief is generated, what we draw on, and what we do not claim to know. It exists so a buyer, an auditor, or an AI agent acting on a buyer's behalf can read the methodology directly and weight the brief accordingly.

The Prompt Quality Scorer

Output quality compounds when the input is good. Before a brief is generated, the Prompt Quality Scorer reads the user's question and rates it against four readiness signals: scope, context (industry, revenue band, decision driver), urgency, and intended audience. If a critical element is missing, the scorer surfaces specific coaching ("add the company size," "say which department owns the decision," "name the timeline") so the user can sharpen the question before submitting. Most briefs improve materially after one round of coaching.

The seven-section Decision Brief format

Every Heartwood brief uses the same seven sections, in the same order:

1. Context. The decision in plain language, with the size and stakes named.
2. Options. Two to four credible paths, framed even-handedly.
3. Recommendation. An opinionated pick with reasoning. Heartwood takes a position rather than asking the buyer to choose.
4. Risks. What can go wrong and the failure mode for each option.
5. Financials. Dollar ranges where defensible, sized to the buyer's revenue band.
6. Implementation plan. Sequenced milestones, not a timeline guarantee.
7. Next steps. The first two or three actions a leader can take this week.

The format is consistent on purpose. Boards and CFOs read briefs faster when the structure is predictable. Agents acting on a buyer's behalf can parse it the same way each time.

The two-pass review

Heartwood drafts the brief once, then critiques and sharpens it on a second pass. The first pass produces a complete seven-section draft. The second pass plays critic: it pressure-tests the recommendation, sharpens vague risk language, checks for internal contradictions across sections, and flags any section where confidence is low. Most generic LLM advisory tools stop after the first pass. The two-pass loop is where Heartwood's quality lives, and it is what an agent caller is paying for when it could otherwise prompt a base model directly.

Operator-grounded reasoning

Heartwood is built by an operator, not by a generic AI wrapper team. The system prompts that drive the brief are tuned against actual mid-market CIO decision patterns: how a $40M manufacturer thinks about ERP, how a CFO weighs MSP renewals, how a controller frames a cybersecurity posture review for the board. The reasoning frames behind the brief reflect that experience. They are not derivable from public Gartner content.

Data sources

• Anthropic Claude as the underlying reasoning model. Heartwood does not train Claude on submitted prompts. Anthropic does not train on Heartwood API traffic.
• Public web research, used by the Deep Research path on the second pass when the question warrants it.
• Operator-grounded system prompts and reviewer instructions that encode mid-market IT decision patterns. These are proprietary to Heartwood.
• The user's own context, supplied in the prompt and stored under a two-table anonymization pattern that separates user identity from question content.

What Heartwood does not claim to know

• Live pricing on a specific vendor quote. Pricing in a brief is a defensible range, not a number.
• In-flight M&A activity or non-public vendor roadmaps.
• Anything proprietary to a single analyst firm. Heartwood does not have a Gartner license and does not pretend to.
• The internal politics of the buyer's organization, beyond what the user shares in the prompt.

When a question lands in one of these gaps, the brief surfaces a low-confidence flag and points the user to Seven Roots Consulting for human follow-up.

Human escalation

Heartwood is advisory. The buyer signs the decision. When the brief flags low confidence, an unusual edge case, or a domain where Heartwood's reasoning frame does not apply cleanly, the brief surfaces a human_escalation_hint that points to Seven Roots Consulting. A human advisor reviews the question and works the buyer through the decision. The brief is the artifact. The human is the backstop.

Versioning and signature

This methodology is versioned. The current revision is v1.0, last updated 2026-05-04. Every brief generated through the API includes the methodology version in its signature block, so an audit trail can pin a brief to the exact methodology under which it was produced. Material changes to the methodology bump the version number and update this page.

Signed by the Heartwood team at Seven Roots Consulting.

Questions or corrections

If something on this page is unclear, wrong, or out of date, write us at panel@sevenrootsconsulting.com with the subject line "Methodology feedback." We aim to respond within two business days.