Branding in the age of AI

A security-solutions company engaged Mercury to consolidate a brand scattered across presentation decks and a public website into a governed record and two developed sales assets.

The work established a source hierarchy, extracted supportable positioning, built claim traceability for material statements, and kept high-risk proof behind approval gates while producing a two-sided one-pager and an eight-page corporate brochure ready for design handoff.

The engagement had a concrete deliverable list: consolidate the brand and produce two working sales assets, a two-sided one-pager and a corporate brochure, that could help a sales team explain the company to buyers. Behind those deliverables sat a harder requirement. The company's story had to be assembled from what its own sources could support, at a moment when generative tools make it trivially easy to produce polished collateral full of statements no reviewer can trace.

Mercury's objective was a governed brand record the company could treat as the single account of how it describes itself, a traceable path from every claim in the collateral back to the evidence behind it, and a clean separation between language ready for review and language that had to wait for explicit approval. If those controls existed and the collateral drew only from them, the engagement would have done its job even before any page shipped.

The company's proof was genuine and widely scattered. Two presentation decks carried much of the positioning. The public website carried another layer, written at different times for different purposes. Read together, the sources supported a distinct profile: a customer-service operating theme, the F.O.C.U.S. framework, employee ownership, long-term private ownership, a Fusion Center, and a bench of technical and service capabilities. No single source held the whole account, and the materials did not give a future writer or reviewer a reliable way to identify the current version of a claim.

The strongest material was also the riskiest. Threaded through the decks and site were guarantee statements, service metrics, certification and audit references, coverage and response commitments, retention figures, and exclusivity language. Those are the claims that can persuade a buyer, and they are also the claims that damage a company when they surface unverified, outdated, or unapproved.

AI made the exposure sharper. A generative model could produce a credible one-pager from this material in minutes, and the polish would hide the sourcing problems. Models given loose inputs also drift toward the average of their training: the interchangeable language of every technical services firm. The company's actual distinctiveness lived in specific, checkable facts, and specific checkable facts are what fast generation tends to flatten first.

The work opened with intake and classification rather than writing. Mercury reviewed the two supplied presentation decks and walked the public website section by section, treating each source as evidence with its own weight. Establishing a source hierarchy served a practical need: when two documents described the same capability differently, a writer needed a rule for which one governed, and a reviewer needed to see which rule was applied.

With the hierarchy set, Mercury extracted what the sources genuinely supported. The service-centric security-solutions positioning held up across the material, and so did the customer-service operating theme, F.O.C.U.S., employee ownership, long-term private ownership, and the Fusion Center. The extraction drew a line between language the evidence supported as written and language that read persuasively while resting on material no one had recently verified.

Everything on the supported side went into a governed brand record with its public-use flag set to false. Internal drafting could build from it immediately, while nothing could reach public copy until a human turned that flag on. The record also narrowed the which-version-is-current problem: from this point there was one place the account lived, and it had an owner.

The riskiest material received its own control. Mercury built a claim-traceability record that captures, for each material claim, the exact wording, the source, the evidence excerpt, a risk rating, an approval owner, a review date, and usage notes. The high-risk proof categories identified in the source audit were held out of every draft until a named person decides to release them. A claim could no longer enter collateral anonymously, and the most persuasive claims could no longer enter at all without a decision.

Only then did AI-assisted production begin. Fable/Claude and GPT model lanes handled drafting and editorial QA under a standing rule that model output is input: every generated passage passed through human review before it could touch the record or a deliverable.

Drafting produced range before decision. Mercury developed three one-pager directions and three corporate brochure directions, each drawing only from the governed record. Reader-scannability research shaped how the one-pager candidates organized information. Brochure-architecture research shaped how the longer candidates carried a narrative. With generation cheap and claims controlled, the useful human decision moved up a level, from whether a sentence was right to which telling helped the company explain why buyers should trust it.

Comparison against those reader criteria settled the selection. The recommended one-pager direction became a two-sided v4 design handoff, and the recommended brochure direction became an eight-page narrative handoff. Both went to design with claims traced to the record and high-risk holds intact.

The company now has a source-backed brand record, review-ready with public use still gated, and a claim-traceability control in which material claims carry wording, source, evidence, risk rating, owner, and review date. Anyone producing a future asset can draft from the record, and anyone reviewing that asset can check its claims without re-deriving the company's truth from a pile of decks.

The company started with a brand that was true in pieces, supported by real operating history, scattered across sources of uneven reliability, and carrying its most persuasive claims in its least verified corners. It ended with a governed source system: one record of what it can say, one ledger of what it must hold, and two developed sales assets whose every statement can be traced and defended.

For technical companies whose collateral is moving faster than their ability to check it, Mercury offers brand and source consolidation, claim governance, and collateral production with AI speed under human authority.