Meet My Team (They're All AI)
I was on a Zoom call recently and someone asked me where my team was. Where was my marketing team. Then they asked about my website team. Then my legal team. Then my team that built my operations app.
I'm a solo operator building a prevention-first animal welfare nonprofit. Seven counties in Central Alabama, six core programs, a partner network, a 250-page book, a custom operations app, a 37-page website, a media ecosystem, and a phone system that triages calls.
One person.
Here's how. And how you can do the same.
The Team
I have five AI team members. Each one does a different job. Each one had to be set up and trained before it was useful. And each one produces work that looks like a human did it, because I spent the time making sure it would.
Claude
Claude is the backbone of the operation, and it comes in three modes. And utilizes plugins like marketing, legal, productivity, sales, brand voice skills and connectors.
Claude Chat is my thinking partner. Quick questions, brainstorming, talking through a problem. When I need a fast answer or a sounding board, that's where I go.
Claude Co-Work is my operations manager. This is where the heavy lifting happens. Writing grant applications, drafting partner agreements, building email sequences, creating leave-behinds, updating my business plan, managing my calendar, tracking 40+ active projects simultaneously. Co-Work has memory. It knows my organization, my people, my programs, my terminology, my brand voice, and my rules. When I say "draft the Remy Fund application," it already knows what AAF is, what our programs do, who our partners are, and how I talk. I don't re-explain anything. It just writes.
Claude Code is my developer. It built my website (37+ pages), my custom operations app, data visualizations, PDF generators, and automation scripts. When I need something technical built or fixed, Code handles it.
All three versions of Claude share the same memory system, so what one learns, the others know.
Stanley
Stanley is my social media strategist. It's an AI content tool that learns your writing style, tone, and patterns over time. It drafts LinkedIn posts and social content that sounds like me, not like a bot. It suggests content ideas based on what's trending and what's worked before, and it remembers my preferences across sessions. When I need to stay consistent across platforms without spending three hours writing captions, Stanley handles it.
Sona - Quo
Sona is my receptionist. She runs through Quo, an AI-powered phone system, and answers calls to our Pet Help Desk. She asks the caller's name, county, and reason for calling. She triages based on a custom knowledge document I built with our program rules, eligibility criteria, and referral pathways. She can schedule appointments, take messages, and route urgent calls. She answers questions about our programs using information I gave her, not whatever she scrapes from the internet. When someone calls after hours, Sona takes the call, gathers the information, and has it waiting for me in the morning.
Perplexity
Perplexity is my researcher. Real-time web search, factchecking, source verification. When I need current data or need to verify a claim before it goes into a grant application or a published document, Perplexity finds it and cites the source. It's the tool I trust for "is this number still accurate" and "where did this stat originally come from."
Recently, it was brought to my attention that I had used someone's statistics from their data and had failed to make sure that everything was cited correctly. And instead of getting upset or defensive I implemented an Accountability and Compliance Policy where every document, website page, graphic going out the door has to be checked for accountability and citations. Every statistic must be correct. Everything has to be cited to the proper source. I have skills set up for Claude and Perplexity to do this. This way I will never have someone call me, write me, or tell me in a Zoom meeting in the future that their data or statistics are showing up in any of Animal-Angels materials without being cited or being incorrect. I even created a Compliance Database (which is a part of the AWRN) to track to make sure that every document had met the accountability compliance check before it goes out the door.
Vislo
Vislo is my infographic designer. It's an AI tool that generates on-brand visuals from text prompts. I uploaded my brand kit (colors, fonts, style rules) and now I describe what I want in plain language, and it builds it. I created eight infographics in one sitting: surrender statistics, prevention vs. shelter costs, pet-inclusive housing data, door-to-door vs. events comparison, program metrics, spay/neuter impact, a landlord playbook, and a taxpayer cost breakdown. Vislo does 80% of the work in seconds.
OpenArt
OpenArt is my image generator. When I need custom graphics, social media images, or visual concepts that don't exist yet, OpenArt creates them from text prompts. I use it for campaign graphics, event visuals, character illustrations, and anything where I need an original image instead of a stock photo.
Wisper Flow
Wispr Flow allows me to talk at 100 to 150 words a minute instead of typing at five words a minute.
The Supporting Cast
The five AI tools above are the core team, but a few other platforms round out the operation.
Salesforce Nonprofit is my CRM. It tracks every contact, every partner, every outreach touchpoint, every grant, and every follow-up across the entire organization. Salesforce also has AI features built in (Einstein) that get smarter the more data you feed it. For a nonprofit managing relationships across seven counties with dozens of partner organizations, having everything in one system that actually talks to itself is worth its weight.
Adobe Illustrator and Photoshop handle graphics cleanup and creation. When Vislo or OpenArt produces something that's 90% right but the text is mangled or the layout needs adjusting, Illustrator cleans it up. Photoshop handles image editing, compositing, and anything pixel-level. Both have their own AI features now (generative fill, neural filters, content-aware tools) that speed up the design work even further.
Constant Contact handles email marketing and has AI-assisted content features for subject lines and copy suggestions.
None of these are the stars of the show, but they all have AI working under the hood, and together they fill the gaps between what the core five produce and what goes out the door.
The Part Nobody Talks About: Setup
Here's where most people go wrong with AI. They open ChatGPT, paste in a request, and get back something that sounds like a corporate brochure wrote itself. Then they submit that to a grant reviewer who reads 200 applications a month and can spot AI-generated content in the first paragraph.
The output is only as good as the setup.
With Claude, you can create something called a CLAUDE.md file. Think of it as the AI's brain. Mine is massive. It contains:
Memory. Every person I've talked to, every meeting outcome, every project status, every partnership, every phone call. When I mention "Rachel at Young-Williams," Claude knows exactly who she is, when we talked, what she said, and what the next steps are. I don't dig through notes or emails. It's all in the memory file.
Brand voice rules. My organization never uses em dashes. Never uses emojis. Never uses guilt-based language or words like "fur baby" or "forever home." Never uses AI-tell words like "delve" or "straightforward" or "leverage." Never dumps bullet-point lists where sentences would work better. We write like real people who care, not like a charity brochure. All of that is baked into the memory file, and Claude follows it every time.
Terminology. We say "The Bridge," not "emergency assistance." We say "prevention first," not "proactive animal welfare services." We say "keep the pet," not "reduce relinquishment rates." Every term is defined so Claude uses the right language automatically.
Project context. Claude knows my book is on V10, my website has 37 pages, Prevention Fest is July 18 at Black Creek Park in Fultondale, I'm waiting on Google Ad Grants verification, and which grant deadlines are coming up. It tracks everything so I don't have to hold it all in my head.
On top of the memory file, you can build **skills**. A skill is a set of instructions for a specific type of task. I have skills for writing grant applications, creating partner leave-behinds, building email sequences, writing social media posts, generating impact calculator numbers, scripting video content, and more. Each skill codifies the rules, formats, and examples that make the output consistent and on brand. Every. Single. Time.
The same principle applies to every AI tool on the team. Sona has a custom knowledge document with program rules and caller scripts. Stanley learns your patterns over time. Vislo has your brand kit uploaded. The setup is the work. After that, the tools just run.
Why This Matters for Animal Welfare
I was on the First Fridays AI call last month. About 25 people from shelters and rescues across the country. One organization mentioned they export animal medical records and paste them into ChatGPT to translate vet notes into plain language for fosters and adopters. They do it record by record, manually, every time.
With a Claude skill built for that specific task, you could standardize the output format, set the reading level, include the right disclaimers, pull in the relevant follow-up care instructions, and produce a clean, consistent summary every single time. No re-explaining. No fixing the tone. No stripping out AI-sounding language afterward. Build the skill once, use it forever.
And here's the thing most people miss: if you're translating medical notes into "plain language" but you haven't set a target reading level for your audience (The average reading level in America is the seventh to eighth grade level - approximately 54% of U.S. adults read below a sixth-grade level), you haven't solved the problem. You've just made it slightly less confusing. A skill lets you lock that in so the output hits the right level every time, without anyone having to think about it.
That's the difference between using AI as a novelty and using it as infrastructure.
The No-AI-Tells Rule
Grant reviewers, funders, and board members can tell when something was written by AI. The phrasing is too smooth. The structure is too perfect. The vocabulary is weirdly formal. There's a sameness to it that feels manufactured.
My rule is simple: if it sounds like AI wrote it, rewrite it.
But the better approach is to set up your tools, so the output never sounds like AI in the first place. Here's how I did it.
Put your actual voice into the system. Not a description of your voice. Your actual words, phrases, and patterns. The way you talk in real life is the way your documents should read.
Ban the A.I.-Tells! No EM dashes. No emojis. No icons. No dividers. No bold text within paragraphs. There's a whole set of AI tells, and you have to ban those in your settings. You don't have to worry about rewriting the document because of all the AI scattered throughout it.
Ban the AI-tell words. Every AI has favorites. "Delve." "Comprehensive." "Robust." "Utilize." "Pivotal." "Leverage." Kill them. Put them in your banned list and they will never show up in your output again.
Ban the AI-tell structures. "Not only... but also..." "It's important to note that..." "In today's rapidly evolving landscape..." Real people don't talk like that. Neither should your documents.
Write rules, not suggestions. Don't say "prefer short sentences." Say "use short sentences." Don't say "try to avoid jargon." Say "never use these words." The more specific your rules, the cleaner the output.
Test it by reading the output out loud. If you wouldn't say it to a colleague standing in front of you, it shouldn't be in the document.
What This Looks Like in Practice
I built a 250-page book. A custom operations app with bidirectional matching, crisis radar, dispatch tracking, foster management, event coordinator, a shared community event calendar, complete shelter and rescue operations management and landlord outreach dashboards. A full website with lead magnets, impact calculators, and program pages. Partner guides, sponsor proposals, grant applications, press releases, email sequences, video scripts, phone system knowledge documents, commissioner outreach packages, and eight branded infographics in one sitting.
I'm one person. A disabled veteran building something from scratch in Central Alabama.
These tools are not replacing the work that I have to do. They're making the work possible.
If you're a solo operator or a small team drowning in the gap between what needs to get done and who's available to do it, this is how you close that gap. Not by hiring ten people. By building a team that never sleeps, never forgets, and writes in your voice.
I have the background and experience as an electronics technician in the U.S. Navy. Then I went to work for the University of Hawaii for ocean floor mapping, where I took care of the sonar system and all the data acquisition computers and programs and I have my Bachelor's in Videography and Graphic Design.
What I didn't have before putting my team together was a way to make it all happen.
Set them up right, and nobody will ever know they're there.
Why I'm Sharing This
I'm not saying AI replaces staff. It doesn't. But for a small nonprofit trying to do prevention work across seven counties with limited funding, these tools have been a multiplier. They help me move faster, stay organized, and produce work I couldn't do alone.
If anyone's curious about any of these tools or how I've set them up, happy to answer questions.
#DataandTechnology
#MarketingandSocialMedia
------------------------------
BJ Adkins
Founder/Director
Animal-Angels Foundation
Pinson, AL
bjadkins@animal-angels.org
animal-angelsfoundation.org
------------------------------