Your brand voice guide is your best AI training data

Her answer was a custom AI bot named Lenny, trained on ASU’s brand guidelines and built to capture individual administrators’ voices. 

“AI has enabled us to take the essence of each department and create capacity for anyone within those organizations to share consistent, powerful stories,” says Becker, founder of SAB Creative & Consulting. 

Here’s how she built it.

The skill set that matters isn’t technical

When Becker demonstrates Lenny in action, the difference between generic AI and a custom-trained bot becomes immediately clear. 

She prompts both standard ChatGPT and then asks Lenny to write a LinkedIn post about Accelerate ASU, the university’s dual enrollment program that allows high school students to earn college credit.

Standard ChatGPT produces a lengthy, article-style response riddled with bullet points and generic language.

Lenny generates a concise, platform-appropriate post that incorporates specific statistics, such as how many learners are enrolled at any given time, and uses ASU’s brand guide and internal positioning to better communicate in the organization’s preferred voice.

“People assume a lot of technical prowess is needed, but I’m here to demystify that,” Becker says. ​“One of the best groups of people to create bots is professional communicators.”

Why communications professionals? Building a bot requires the same skills you already use to create a communications strategy.

This reframing matters because it reveals where autonomous marketing differs from traditional marketing automation. Instead of just setting up triggers and workflows, you’re training an agent that understands context, maintains voice consistency, and adapts outputs based on comprehensive brand knowledge.

Four steps from brand guide to working bot

For ASU’s Learning Enterprise, Becker followed a systematic approach that any marketing and communications professional could replicate:

Step 1. Build your knowledge base

• First, Becker gathered style guides and brand positions for the university’s various institutions and brands, transcripts from leadership talks, past owned media articles, and social posts that exemplified the organization’s voice and detailed style guidelines.
• She then organized these materials into Word docs, PowerPoints, and PDFs for Claude to reference. 

(Read more about creating your own customer knowledge center.)

Step 2. One document, one source of truth

• Because AI platforms limit the number of files you can upload, Becker recommends putting all voice attributes and brand position data into a single document.
• A document like this is available to ASU staff. People can suggest changes to the document to update Lenny’s existing knowledge base, making the knowledge base collaborative and transparent for everyone. 

Step 3. Upload, instruct, and test

• Next, Becker uploaded her consolidated documents into ASU’s enterprise ChatGPT workspace to create Lenny. She says that she programs AI chatbots in a ​“platform-agnostic way” that could work across any tool.

Step 4. From org-wide voice to individual style profiles

• Here’s what makes Lenny particularly powerful: it knows how individual leaders at ASU communicate.

When Becker prompts Lenny to write a social post in a specific executive’s style, the bot generates content that matches the executive’s specific voice and tone, without additional context.

The result? An executive uncertain about what to post on LinkedIn can prompt Lenny for a draft. A new communications team member can use the bot to understand the organization’s writing style. ASU’s staff can access Lenny to become more confident storytellers.

“ASU was the first university to adopt ChatGPT at an enterprise level, meaning everyone in the community has access,” Becker explains. ​“It means anyone can feel empowered to be a communicator and a storyteller.”

Three problems that generic AI can’t fix

For organizations with numerous brands, multiple stakeholders, and limited resources, custom bots address specific pain points that generic AI can’t touch:

1. The brand consistency problem at scale: When a university has hundreds of programs, each with its own communicators, maintaining voice consistency becomes nearly impossible. Lenny ensures everyone pulls from the same brand knowledge, whether they’re in admissions, alumni relations, or academic affairs.
2. Scrappy teams need a force multiplier: Many nonprofits and smaller institutions describe themselves as ​“scrappy.” They lack the budget to build robust communications teams but need to tell compelling stories. ​“AI becomes a creative muse for humans and helps set a tone for the team,” Becker explains. ​“This is a repeatable and scalable strategy to start telling more stories and getting your voice out there.”
3. Turn everyone into a communicator: From program directors who need to write newsletters to faculty members who want to promote their research, custom bots can give reluctant communicators a starting point that matches organizational standards.

The pragmatic case for engaging with AI in higher ed

While helping organizations like ASU, the University of Massachusetts Amherst, and Washington University in St. Louis integrate AI into their communications, Becker has witnessed plenty of skepticism surrounding AI in education, particularly around student cheating.

Her response is pragmatic: ​“Never in the history of history has banning something been effective.”

She points to the irony of AI detection software that flags legitimate student work as AI-generated. A colleague submitted papers he’d written in high school and college—years before AI existed—and the software flagged them for cheating. Had he been a modern student who’d done authentic work, he could have faced academic probation.

“There are very real dangers and unknowns and complexities that surround the use of AI,” Becker acknowledges. ​“But ignoring that is only going to make the divide worse.”

Instead, she encourages educational institutions to accept that AI is already achieving mainstream adoption.