We get the question almost every week, and it comes in two flavors. The technical founder asks “how are you using LLMs in your engagements?” with hope. The traditional brand person asks the same question with suspicion. Both deserve a precise answer.
TISSA is AI-native, in the specific sense that we use models heavily in the production of every Master Book and on every Owner’s Rep retainer. We are also AI-skeptical, in the specific sense that we do not let models touch the upstream decisions a brand depends on. Both postures are deliberate. The line between them is the entire bet of the firm.
What models do well.
Three things, demonstrably, at production quality.
Voice enforcement at scale. Once the Verbal chapter of the Master Book is written — tone descriptors, lexicon, paired examples — a fine-tuned model can rewrite arbitrary copy into compliance in seconds. We use this to QC vendor-delivered copy at Gate 2. A 4,000-word RFP response gets scored against the lexicon in under a minute; the deviations come back as a structured diff, which a human reviewer accepts or rejects line by line.
Application drafting. First-draft renderings of sales sheets, one-pagers, and email templates against the Master Book’s specifications. The model does not invent the spec; it executes it. The output is always reviewed at Gate 2, but the first draft arrives in hours rather than days, and the review focuses on judgment rather than on transcription.
Audit photo classification. The Quarterly Field Audit captures dozens to hundreds of photographs of brand-in-the-wild. A multimodal model triages them against the 4C rubric, flags the off-standard ones, and produces a draft scorecard. The Steward then walks the flagged set on foot and finalizes. Without the model, the triage was the bottleneck. With it, the audit shifts from sampling to comprehensive coverage.
What models cannot do.
Two things, durably. The reasons are structural, not temporary.
Set the strategic posture. The Foundations chapter is a position, not a synthesis. It states what the brand believes about its market, where it sits versus competitors, and what it explicitly refuses to do. A model produces a competent-sounding average of every brand strategy in its training corpus. That is exactly the wrong output: brands win by departing from the average, not regressing to it.
A model gives you the average of every brand. The job of the principal is to know which direction to walk away from it.
Make a judgment call inside a Council. When two ratified positions in the Master Book come into apparent conflict — which happens every fifteen to twenty Decision Memos — the Council must decide which precedent governs. The model can lay out the conflict cleanly. It cannot say “we are going to live with the second one and amend the first.” That is a stake. Stakes are what brands are made of.
Where we use them inside TISSA.
Concretely, here is the model surface across a typical engagement.
In the Express Diagnostic, models help triage photo and screenshot evidence, score it against the 4C rubric, and surface the top three drift patterns. The principal writes the Decision Memo. The model writes nothing the client sees.
In the Master Book build, models draft applications, render variants for review, and run consistency QC across the six chapters as the book gets closer to v1.0. The voice of the prose — Foundations, Governance — is written by the principal in full.
In the Owner’s Rep retainer, models run voice QC on vendor deliverables before Gate 2, classify field-audit imagery, and propose Decision Log entries the Steward edits and signs. The Council itself is human; the Memo is signed by humans; the precedent is set by humans.
Why this changes the consulting model.
The interesting question is not whether AI displaces brand consulting. It is what kind of brand consulting AI lets a small firm sustain.
Before models, the production overhead of a Master Book engagement required junior labor to be commercially viable at any price below $80K. The pyramid was structural. With models, a single principal can hold the strategic spine and let the production layer be a tool rather than a team. The ratio of senior judgment to total output rises from roughly 15% in a pyramid model to over 70% in a principal-led, model-augmented model.
That is what “AI-native” actually means in practice — not that the deliverable was written by a chatbot, but that the firm shape has been redesigned around what a senior practitioner can hold when production is no longer the bottleneck. The result is more judgment per dollar, not less.
The model-augmented production layer is one reason our pricing sits where it does. The methodology it serves is The Brand Operating System.
