Treger Voice · Voice Governance for Public-Serving Institutions

Why This Matters

Voice — the way you talk about your organization and what you do in the world — is infrastructure, not decoration.

Your voice, as carried by your words, matters. And it lands in a crowded room: your customers experience most institutional mail as clutter, and they triage accordingly.

For example: a confusing eligibility notice is not simply a communications defect. If your customer, the recipient, does not understand you, it is an access failure — and they will either call you or ignore you. Both cost you.

A person who cannot understand a letter does not file the appeal. An applicant who misreads a deadline loses the benefit. A patient who cannot follow discharge instructions returns through the emergency room. Each of these failures has a cost, and that cost lands on both the person and you.

Most organizations manage their facilities, their data, and their finances as systems, with owners, standards, and controls. Language is usually the exception: it is managed through style guides, habit, and individual judgment.

That approach worked when producing language was slow and expensive. Add AI language generation into the mix, and that old problem arrives at a brand-new speed.

What Changed

Your organization can now generate a thousand letters an hour. It cannot review a thousand letters an hour.

Generative AI did not create this problem, but it did expose one that has been around for a long time. We have had brand guides, legal review, plain-language programs, and accessibility standards for decades, but these mechanisms were never connected in one system. AI has widened the gap between what organizations intend to say and what they actually say, and it has widened that gap faster than anyone can monitor.

This problem shows up in three ways.

No one owns the voice

Different teams speak for your organization in different registers. Legal writes one way, program staff another, the chatbot a third. When no one has the authority to resolve those differences, the customer resolves them instead, by deciding which version of the institution to believe, or not believe.

The AI has no supervisor

Which communications does AI draft? Which can be sent without human review? Which topics require legal or subject-matter approval? And who is accountable when it explains a denial incorrectly? Many organizations deploying AI in customer communications cannot answer these questions yet.

No one measures whether it works

Reading level, comprehension, task completion, avoidable call volume, appeal rates, and complaint themes are all measurable. Very few organizations measure them together as one system.

Services

There are three ways we can start, depending on where you are.

Some organizations need the problem documented. Some need a system built. Some have a system and need it to survive AI deployment. Every engagement can be scoped to the question that is right in front of you.

Voice Audit

For organizations that suspect something is wrong and need it documented.

A structured assessment of how your institution currently communicates across channels. We evaluate notices, letters, forms, portal copy, call scripts, and chatbot output for comprehension, accuracy, accessibility, consistency, and whether the tone fits the stakes of the situation. Audits can focus on a single high-risk customer journey, one program, or a representative cross-channel sample.

You get

A findings report, a severity-ranked inventory of communication risks and breakdowns, and a baseline for measuring improvement.

It answers

Where is our language costing us?

See a sample audit finding →

Voice Governance System

For organizations ready to build the operating system.

The complete architecture. A voice constitution grounded in your mission and your customers’ needs. Standards for clarity, terminology, empathy, and evidence. Specific guidance for high-stakes situations: denials, delays, errors, enforcement, and sensitive life events. A decision-rights framework that names who may speak, who approves, and who resolves disputes. An assurance process that keeps the system honest over time.

You get

Governing documents your organization can operate, as opposed to a style guide that sits on a shared drive.

It answers

How should you speak, who can speak for you, and who is accountable?

Language Assurance for AI Deployment

For organizations putting AI in front of the people they serve.

This is the implementation layer. System prompts, retrieval sources, model instructions, permission boundaries, escalation triggers, pre-release testing, output sampling, and incident review. Designed to fit inside your existing AI governance framework instead of next to it.

You get

Controls a CIO can point to, and evidence they are working.

It answers

Can we defend what our AI said to that person?

Also available.

Fractional voice leadership. Ongoing counsel, governance oversight, and cross-functional coordination for organizations that need the function but not the headcount.

Rapid response. Steady, accurate language when the stakes are high and the clock is running.

How It Works

We read everything. Name the owner. Build the system. Test whether it holds.

Every engagement adapts to your organizational needs. The sequence rarely changes.

Immersion

We read the actual artifacts first: the denial letter itself, not just the policy governing denial letters. We interview the people who write these communications and, where possible, the people who receive them. The output is a clear map of the distance between what you as an institution believe you say and what you actually say.

Authority

We establish who owns the voice. This is a governance question before it is a language question, and it is the phase most organizations skip, either by preference or without noticing. Skipping it produces documents no one enforces.

System

We build the constitution, the standards, the situational guidance, the terminology assets, and the decision rights. We train the people who will use them. The goal is internal fluency: a voice that belongs to the organization.

Assurance

Pre-release testing. Sampling. Audits. Customer research. Incident review. Correction. A system that cannot detect its own failures is not a system.

Sectors

In commerce, an inconsistent voice weakens a brand. In government, it determines whether someone can eat.

Public benefits. Health systems. Child welfare. Housing. Education. Any institution whose language stands between a person and something they need.

These organizations communicate with people under stress, people with limited English proficiency, people with disabilities, and people navigating administrative systems they did not have a hand in designing. For these institutions, communication is not adjacent to service delivery; it is service delivery.

The same institutions are now deploying AI into exactly those communications.

Making sure the language holds through that deployment: that is our work.

Selected Case Work

Twenty-five years, mostly in rooms where the intricacies of language, style, and audience mattered.

At a glance · Statewide eligibility platform serving SNAP, TANF, and Medicaid applicants · 1,300+ nursing homes and K–12 systems reached during COVID-19 response · 20+ national affiliates aligned under one voice · 600+ annual communications governed under one system

Government and Human Services

Single Benefits Platform. UX and plain-language content architecture for SNAP, TANF, Medicaid, and adjacent programs on a statewide eligibility platform.

Child Support Banking Transition. Messaging for a statewide change in payments infrastructure, reaching families who could not afford to misunderstand it.

COVID-19 Public Health Response. Outreach strategy across 1,300+ nursing homes and K–12 systems during a period when the guidance changed weekly.

Energy Assistance Programs. Communications for households applying for utility and weatherization support.

Health, Nonprofit, and Advocacy

Volunteers of America. Narrative reset and voice alignment across 20+ national affiliates operating under one name without shared language.

Maryland Heart Gallery. Campaign content and event strategy for foster-to-adopt awareness.

Jewish Community Services. Marketing and communications leadership across a multi-program human services agency.

Finance, Education, and Culture

2008 Liquidity Crisis. Financial data strategy and public relations for a national treasury association during a period of acute institutional distrust.

Graduate Curriculum Design. Content and architecture for public policy, leadership, and entrepreneurship programs.

Gordon Center Messaging System. Content strategy governing 600+ annual communications across arts, education, and wellness programming.

William E. Treger, founder of Treger Voice — William E. Treger · Founder

About

“Mission is a noun. Voice is a verb.”

I started as a reporter. I have spent twenty-five years helping institutions sound like themselves.

Early stops at The Village Voice, The Nation, and Vanity Fair gave me a reporter's instinct for how language moves through the world.

Since then, I have worked with public agencies, health systems, nonprofits, universities, and financial institutions, writing and rewriting and refining what they say and what they do.

My mother taught high school English in Baltimore, Maryland for thirty years, so I was raised by an editor with a loving yet sharp red pen who taught me the difference between a sentence that lands persuasively and one that does not.

I hold a BA in English Literature from the University of Michigan and an MBA from the University of Maryland's Smith School of Business. I am currently based in Washington, DC.

When your organization speaks, someone acts on it. Or throws it away.

Voice — the way you talk about your organization and what you do in the world — is infrastructure, not decoration.

Your organization can now generate a thousand letters an hour. It cannot review a thousand letters an hour.

There are three ways we can start, depending on where you are.

Voice Audit

Voice Governance System

Language Assurance for AI Deployment

We read everything. Name the owner. Build the system. Test whether it holds.

Immersion

Authority

System

Assurance

In commerce, an inconsistent voice weakens a brand. In government, it determines whether someone can eat.

Twenty-five years, mostly in rooms where the intricacies of language, style, and audience mattered.

Government and Human Services

Health, Nonprofit, and Advocacy

Finance, Education, and Culture

I started as a reporter. I have spent twenty-five years helping institutions sound like themselves.

The discipline this practice is built on.

If your language is off and you want a second opinion, start there.