Voice Governance for
Consequential Communications
When your company speaks,
someone acts on it. Or throws it away.
Treger Voice finds and fixes language failures in the decisions, notices, instructions, screens, and AI-generated answers people depend on.
For organizations in health care, education, and financial services, and the technology companies that support them.
See the method at work
The same notice, before and after.
A Voice Audit starts with a document someone actually received. This one is fictional: its names, figures, timelines, and phone numbers are invented, and the governed version is a proposal the institution would confirm and approve before using it.
Before
Dear Member, Thank you for your interest in the Complex Care Program. Following clinical review, we are writing to inform you that your referral has been conditionally queued pending panel capacity reassessment1. Program enrollment is contingent upon care-team availability and panel turnover, evaluated on a rolling basis. Enrollment timelines cannot be guaranteed and are subject to change without notice2. Failure to respond to outreach attempts may result in removal from the queue3.
Findings
- 1Critical
The patient cannot tell what happened.
- 2Critical
The wait has no end date.
- 3Critical
The removal threat has no instructions attached.
After
You are on the waitlist: number 14 of 31. Most people in your position are enrolled within three to five months. To keep your place, respond within 10 days when we reach out. Our calls come from 410-555-0123, and our letters look like this one. Keep seeing your current providers; this program works alongside your regular care. If your health changes while you wait, call us.
Note: The governed version tells the patient their place in line and how to keep it.
Voice Governance, defined
Someone has to answer for every sentence your company sends.
Different teams have long produced different versions of your company's voice, and mass email multiplied them. AI adds something new: it can compose a distinct message for every customer, on the fly, and send it before anyone has read it. The rules for what those messages may say, who decides, and who checks them have to come from the company.
At the moment of decision, the message is part of your product, so a language failure is a product failure.
How it breaks down
Your mission reaches customers through people who didn't write it.
Leadership defines the promise.
What leadership authorizes, inside the company and outside it, establishes what the company means. Where it fails: The two versions differ, or the tone shifts once the news is bad, and everything downstream inherits the difference.
Teams put it into their own words.
Sales, support, policy, and product each translate the promise for their own work. Where it fails: Without shared standards and decision rights, sales promises one thing, support explains another, and the policy announcement reads like it came from another company entirely.
People are left to interpret it.
The patient, the student, the borrower, the employee receive the result. Where it fails: They can't tell what to do next. This is also where the consequences can be measured, which is why an audit starts here.
Services
Start with an audit.
Where is our language costing us?
Voice Audit
Send me three to five high-stakes communications: the customer letter, the policy announcement, the benefits notice, the form, the onboarding screen, the chatbot transcript. I audit them against published plain-language and accessibility standards, against one another, and against what your leaders have promised.
Two to three weeks after I receive them, you get a prioritized list of problems, the consequences each one can cause, and rewrites of the passages that need attention first.
Fixed scope and fee. No platform, subscription, or retainer.
If the audit finds the same failure across your messages, I design the standards, decision rights, and AI permission rules to address it.
When the problem goes beyond individual messages.
Voice Governance System
One standard for how you speak, tested against what you send. I build it with the people who will own it, so it can run without me.
What are we authorized to say, and who decides?
Language Assurance for AI
I help you define what your AI systems may say, and I test whether they do. Your team or vendors build and run the systems.
Can we defend what our AI said?
What clear language
can buy you
A notice that answers its readers' questions can keep them off the phone.
You keep the rules your AI must follow and the test results that show whether it does.
The Business Case: the research on what unclear language costs →
Case Examples
Six case examples.
Fictional cases built on real document types, with severity-ranked findings, the cost of leaving each problem unresolved, and a governed version.
- 01Read the audit →
Hospital discharge instructions
The packet that leaves critical instructions buried.
- 02Read the audit →
A care-navigation waitlist notice
It never uses the word “waitlist.”
- 03Read the audit →
A financial aid email
A cancellation warning with no deadline.
- 04Read the audit →
A community lender's decline email
The "no" that should have been a "not yet."
- 05Read the audit →
An onboarding flow
The error screen won't say which detail didn't match.
- 06Read the audit →
An AI chatbot's answer
An eligibility determination the available facts do not support.
Selected communications work from before Treger Voice
- 20+national affiliates, a common voice developed for them
- 1,300+nursing homes and K–12 systems in a COVID-19 outreach strategy
- 600+communications a year under one content strategy
- 24jurisdictions on a statewide platform
William Treger spent twenty-five years inside newsrooms and organizations in the health and human services, education, finance, and technology sectors. He developed this framework and refers to the practice as Voice Governance. Treger Voice is the new practice he founded to offer it.
Show me the communications you're worried about.
You don't have to know exactly what you need. Start with the email that keeps generating calls, or the chatbot nobody's watching. Send one over and I'll return a short read at no charge: the three things I'd fix first, on one page. Take out names, account numbers, and other personal details before you send it. If an audit would help after that, we agree on the scope and fee before work begins.