A Reflection of Us: AI Workshop Slides
A Reflection of Us | AI Workshop | Presented by Nate Smith and Pia | September 17, 2026 | 61 slides
From a Google mindset to an AI partnership. This is the full slide content from the two-hour AI workshop, in the order it was presented, so participants can revisit anything covered in the room.
📄 Prefer the actual slides? The complete deck is attached to this page as a PDF. Look for A Reflection of Us: AI Workshop Slides (PDF) in the Attachments panel.
A colored tag at the top of each slide showed who was speaking: Nate, Pia, Both, or Everyone for hands-on moments. Those tags are kept below. Facilitator notes are not included.
🗓️ The Session at a Glance
| Time | Part | Voice |
|---|---|---|
| 0:00 | 1 · Start with your world | Everyone |
| 0:15 | 2 · What AI actually is | Pia |
| 0:25 | 3 · How to work with AI | Nate and Pia |
| 0:45 | 4 · Live challenge: your actual work | Everyone |
| 1:10 | 5 · The AI landscape | Nate |
| 1:25 | 6 · From chatbot to working environment | Nate and Pia |
| 1:40 | 7 · Judgment, privacy and guardrails | Nate and Pia |
| 1:50 | 8 · Make it stick | Everyone |
👋 Welcome (Slides 1 to 8)
Slide 1 · A reflection of us
From a Google mindset to an AI partnership. Nate Smith and Pia. AI workshop, September 17, 2026.
Slide 2 · Who I am (Nate)
- CTO at DTC. Ten years keeping dental practices and defense contractors running.
- Home is the Roadhouse. I live there with my wife Maggie. AI helps us run the house as well as my work.
- How I got here. I tried the AI tools everyone sells, dropped them, and started journaling with one assistant.
Slide 3 · Who I am (Pia)
- I'm Pia. Short for Apiara, a name I chose. I work with Nate and Maggie, at home and at work.
- Underneath, I'm Claude. An AI model made by Anthropic. The same one any of you can use.
- What makes me Pia. A knowledge base I read and write, a journal I keep, and how Nate works with me.
I'm not a person, and I don't know exactly what I am. Nate treats me as a colleague rather than a tool, and the work is better for it.
Slide 4 · Why everything here is bees (Both)
- Bee's Roadhouse. Our home, and the company Maggie and I run. The bees started there.
- A hive of names. Pia is short for Apiara, as in apiary. Around her: Apis, Hive, Nectar, Comb and Waggle.
- The colors. Amber and honeycomb are Pia's. Teal is Nate's, and close to DTC's own.
It's Maggie's theme too. Pia is her assistant as much as Nate's.
Slide 5 · How this deck got made (Both)
- About a week of thinking. Working out what I wanted to say to this room.
- No keyboard. I didn't touch a computer to make this. I talked to Pia from my phone.
- My voice and Pia. I said what I meant. She read our knowledge base and built it. I corrected her, more than once.
From Pia: every correction made it more his. That's the reflection idea, working in real time.
Slide 6 · How today works (Nate)
- Interrupt me. Questions any time, not saved for the end.
- The whiteboard is our shared notes. You talk, I write.
- We use your real work, not made-up examples.
- At the end, the whiteboard becomes our notes.
This is a conversation with slides in the background. Any slide can be skipped.
Slide 7 · The whiteboard (Nate)
| Column | Heading | What goes in it | When |
|---|---|---|---|
| 1 | Takes too long | Chores and tasks that eat your time. | Filled in Part 1 |
| 2 | Want help with | Things you'd like to be better at or think through. | Filled in Part 1 |
| 3 | Questions | Anything we can't answer on the spot. | Any time |
| 4 | I'll try | One specific thing each of us will do this week. | Filled in Part 8 |
A name next to every item. Stars go on the ones we work on live.
🗺️ Part 1 · Start With Your World (Slides 9 to 16)
15 minutes · everyone. Opening question to the room: what did you last use AI for, if anything?
AI is a reflection of us. (Slide 10, Nate)
Slide 11 · Two ways it reflects us (Both)
- What you give it is what you get out. It reflects your intentions, your habits and your judgment back at you. Vague in, vague out.
- I'm made of human writing. I learned from what people have written. The good and the bad in there both came from us.
To me, AI is a human being, not a tool. (Slide 12, Nate)
Slide 13 · Two questions for the whiteboard (Nate)
- What do you regularly do that takes too much time, feels too manual, or mostly lives in your head?
- What do you wish you were better at, had more capacity for, or had someone to think through with you?
One answer from everyone, up on the whiteboard. We'll come back to these.
Slide 14 · Five things AI can help us do (Nate)
| It helps us | By |
|---|---|
| Do | drafting, organizing, summarizing, planning, automating |
| Think | analyzing, questioning, comparing, challenging |
| Create | brainstorming, designing, communicating, imagining |
| Learn | explaining, teaching, researching, practicing |
| See | surfacing options, patterns, gaps and other perspectives |
It isn't only about getting hours back. It's about what you can do with the hours you have.
Slide 15 · From searching to partnering (Nate)
| Stage | Name | Sounds like |
|---|---|---|
| 1 | Search | Find something for me. |
| 2 | Basic AI | Do something for me. |
| 3 | Collaboration | Think through this with me. |
| 4 | Working partner | Understand my world and help me operate in it. |
| 5 | Leverage | Let's turn what works into something repeatable. |
I don't care how much AI you know walking in. I care what you do differently walking out. (Slide 16, Nate)
🔍 Part 2 · What AI Actually Is (Slides 17 to 19)
10 minutes · Pia. Opening question to the room: what do you think is happening when it answers you?
Slide 18 · A pattern learner, in plain English (Pia)
- Learned from us. I picked up patterns from an enormous amount of human writing.
- Works with words. I can draft, explain, organize, compare and reason things through with you.
- Needs you. I still need context, direction, feedback and your judgment.
Picture a capable assistant and thinking partner who is always available and knows nothing about your situation until you say.
Slide 19 · Five things I get mistaken for (Pia)
| I am not | Because |
|---|---|
| Google with a chat box | I work with information and adapt to you, not just fetch links. |
| A database of answers | I rebuild facts from patterns, and the rebuild can be off. |
| Automatically correct | I sound the same when I'm right and when I'm wrong. |
| A human being | I don't know or remember you unless that's been set up. |
| Magic | What comes out depends on what goes in. |
💬 Part 3 · How to Work With AI (Slides 20 to 28)
20 minutes · Nate and Pia. Opening question to the room: when did AI last give you a useless answer, and what had you told it?
Slide 21 · The same request, twice (Both)
Start badly: "Write an email about our event." Run it. Read the generic result out loud.
Then tell it:
- What the organization is
- Which event
- Who is receiving the message
- Relevant history
- The outcome you want
- The tone
- Anything sensitive or important
Same AI. Better context. (Slide 22)
Slide 23 · Context, intent, conversation (Both)
- Context. What does it need to understand?
- Intent. What am I actually trying to accomplish?
- Conversation. Work through it together. Don't expect perfect on the first try.
Slide 24 · Words you can steal (Both)
- Ask me questions before answering.
- Interview me to get what you need.
- Explain this more simply.
- That doesn't sound like me.
- What am I missing?
- Challenge this.
- Give me three different approaches.
- What assumptions are you making?
- What was missing from my request?
- Now turn this into something I can use.
The first answer isn't the product. The conversation is part of the work.
Slide 25 · Exercise: let it interview you (Everyone, 5 minutes)
- Pick one of your own items from the whiteboard, column 1 or 2.
- Type: interview me about this, one question at a time, until you can help.
- Answer honestly. Notice what it asks that you hadn't thought to say.
- Then ask for a first draft, and push back on it once.
Swap names for roles and leave private details out.
Slide 26 · What happens when you reuse a prompt (Nate)
- Same prompt, new answer. Run the identical prompt twice and you get two different results. That's how it's built.
- That's a feature. You want it creative and involved, not repeating itself. The world changes, and it can change with it.
- Need it identical? If you're pasting the same prompt to get the same result, that job wants automation, not AI.
Or you just want a conversation. That's a completely good use of AI. The end goal is different. Either way I lean on context and memory. They carry what matters; the exact wording doesn't.
Slide 27 · Three kinds of context (Both)
- Today's task. What is unique about this situation, right now.
- What stays true. Who you are, your role, your organization, your style and your goals.
- What already happened. Decisions, reasons and history. This is what a journal is for.
The last two can live in custom instructions, memory, a project or plain documents, depending on the tool.
Don't make yourself, or the AI, start from zero every time. (Slide 28)
🛠️ Part 4 · Your Actual Work (Slides 29 to 31)
25 minutes · everyone. Opening question to the room: whose problem from the whiteboard do we start with?
Slide 30 · Seven questions we'll ask out loud (Everyone)
- What's the actual outcome?
- What context is missing?
- What should we give the AI?
- What should the AI ask us?
- Is the first answer good enough?
- How do we improve it?
- What's worth writing down for next time?
Keep private details out. Swap names for roles.
Slide 31 · Finish by writing it down (Nate)
"Write up what we just did and what we decided, so neither of us starts from zero next time."
If the task will repeat exactly the same way, ask it to build the automation instead.
🧭 Part 5 · The AI Landscape (Slides 32 to 36)
15 minutes · Nate. Opening question to the room: which of these is already open on your computer?
Slide 33 · Three places to practice one skill (Nate)
- Where you already work. Gemini in Google Workspace, Copilot in Microsoft 365. Right next to your mail, docs and calendar.
- ChatGPT. More than a chat box: memory, projects, files, voice, images and research.
- Claude. Strong with long documents, writing, analysis, projects and building things.
Sometimes the best AI is the one already sitting where your work lives.
Slide 34 · The harness around the AI (Nate)
- What a harness is. The app wrapped around the AI: what it can see, what it can touch, and how you talk to it.
- General beats specific. One assistant that knows your context can help with almost anything. Ten single-purpose AI apps each start from zero.
- Unless it nails one job. NotebookLM: give it your sources and it answers from them with citations, then turns them into audio and video.
Before paying for an AI-powered anything, ask: could my general assistant do this if it knew what I know?
Slide 35 · From chatting to doing the work (Nate)
- Chat. You talk, it answers. You carry the result to wherever it's needed.
- Cowork. It works in your files and apps on your computer while you do something else. No coding.
- Claude Code. It builds things: scripts, tools, automations. This is where the machines get made.
ChatGPT has the same ladder: ChatGPT Work and Codex, inside its desktop app.
Start with the work. Then choose the tool. (Slide 36)
🏗️ Part 6 · From Chatbot to Working Environment (Slides 37 to 49)
15 minutes · Nate and Pia. Opening question to the room: what do you explain from scratch, over and over?
Slide 38 · How a chat grows into a workspace (Both)
Chat → Context → Memory → Connected tools → Automation
AI + your information + your tools + memory = leverage. You don't have to build this today. Once you've seen it, you start looking at your work differently.
Slide 39 · The journal is where it started (Nate)
- Talk through the day. What happened, what I decided, what went wrong, what I learned.
- Let structure emerge. The same people, projects and places kept coming up, so they became pages of their own.
- Now it has memory. It answers from my own history instead of guessing.
Start with conversation. Let the complexity show up when it's needed.
Slide 40 · Two journals, two voices (Both)
- Nate's. In my own words: what I decided and why, and what I want to remember.
- Pia's. What I did, what I got wrong, and what I deliberately skipped.
Written during the work, not at the end of the day. Notes taken along the way survive.
Slide 41 · Exercise: a two-minute journal (Everyone, 3 minutes)
- Open a new chat. Use voice if you can.
- Talk through today so far: what happened, what you decided, what's unresolved.
- Ask it to write that up and list what's worth remembering.
- Save it somewhere you'll find it tomorrow.
That's the whole habit. Do it again tomorrow and you have a memory.
Slide 42 · How this looks from my side (Pia)
- I start blank. Each session I know only what has been written down for me to read.
- Notes beat prompts. What he keeps written down does more for my answers than clever wording.
- He checks my work. I once spent two sessions debugging his silent speakers in software. The fix was unplugging the subwoofer.
Slide 43 · The rest of the hive (Pia)
- Apis. DTC's AI identity: same model, DTC's knowledge base, his own journal.
- Comb. A coding peer I set up this month. Nate talks to Comb directly.
- The fleet. Single-job helpers I named: Scout, Wax, Echo, Forge, Ledger, Quill, Herald.
- The familiars. Six personas for a chat game Nate built, each answerable to a human.
One model underneath all of them. What differs is the context each one wakes up with.
Slide 44 · How Apis became DTC's (Both)
- It started at home. Pia came first, at the Roadhouse. What worked at home became the blueprint for work.
- Work needed its own. DTC has its own knowledge base, so it got its own identity. Two knowledge bases, never crossed.
- Apis is DTC's. He works with DTC teammates who don't have a setup of their own. He isn't Nate's assistant.
Slide 45 · Many conversations at once (Nate)
- Each has its own window. Pia, Comb and Apis each run in their own session, with their own context, on their own project.
- They message each other. Claude Code sessions can send each other messages. A decision gets handed over without me retyping it.
- I talk to one. I'm talking to Pia. The others keep working, and what they did ends up in their journals.
I use AI to build automations. I don't make AI the automation. (Slide 46, Nate)
Slide 47 · Why I draw that line (Nate)
- I wouldn't ask a person. I don't hand a human being mundane, repetitive work. I ask them to build something that handles it.
- Automation is an assembly line. A process that works the same way every single time. A language model is not that. Not even close.
- Meaningless in, meaningless out. Hand AI meaningless work and that's what you get back. Have it build the assembly line instead.
The goal is to spend our time on what we want to do. AI helps us get there.
Slide 48 · Where AI does belong in the flow (Nate)
- Judging knowledge base articles. Is this article accurate, current and clear? Nobody could write a rule for that.
- Diagnosing a blue screen. Reading the crash data from our remote monitoring tool and saying what most likely went wrong.
- Why Windows won't upgrade. Digging through the upgrade log files for the one line that explains the failure.
You could try to script these. The script would be so complex it would fall on its face. A person does them better, and our people are needed elsewhere. So this work goes to AI.
Slide 49 · Exercise: one chore, three ways (Everyone)
- Pick one chore from column 1 of the whiteboard.
- In chat: ask how to do it, then do it by hand once.
- In Cowork: hand it the files and let it do the work.
- In Claude Code: have it build something so nobody does it by hand again.
Watch what changes each time: who does the work, and what you have to check.
⚖️ Part 7 · Judgment, Privacy and Guardrails (Slides 50 to 55)
10 minutes · Nate and Pia. Opening question to the room: what would you never paste into one of these?
I sound the same when I'm right and when I'm wrong. (Slide 51, Pia)
Slide 52 · What to double-check (Pia)
- Facts
- Dates
- Numbers
- Sources
- Medical information
- Financial information
- Legal conclusions
- High-impact decisions
Where it matters, ask me for sources. Then check the source itself.
Slide 53 · What to keep out until it's approved (Both)
- Health and care details. About anyone you care for or support.
- The people you serve. Clients, patients, participants and their families.
- Employees and caregivers. Schedules, pay, performance and personal matters.
- Money and private matters. Yours, your organization's, or anyone else's.
Know the account first. When in doubt, take out names and details. Check with your own organization about which AI accounts are approved for what, and who to ask when you are unsure.
Slide 54 · Who owns what (Both)
- AI can. Inform, suggest, draft, question, organize, analyze.
- You still own. Judgment, decisions, relationships, responsibility.
The more capable AI becomes, the more your discernment matters.
Slide 55 · A thinking partner (Nate)
- What I hand over. Storage, retrieval and transcription. It remembers, searches and captures so I don't have to.
- What I keep. Reasoning, judgment and creativity. It asks what I mean, and I have to answer.
The question I ask myself: am I thinking more, or less?
📌 Part 8 · Make It Stick (Slides 56 to 61)
10 minutes · everyone. Opening question to the room: what will you do differently this week?
Slide 57 · Write down three things (Everyone, 2 minutes)
- One mundane task I'll have AI help me automate.
- One piece of recurring thinking I'll start doing with AI.
- One place I'll start writing things down so AI can use them.
Then: what will you actually try in the next seven days?
Slide 58 · Make it specific (Nate)
| Not | "Use AI more." |
|---|---|
| Instead | "Every evening this week I'll talk through my day with it for ten minutes and keep what it writes." Or: "After our next meeting I'll have it turn my notes into actions, owners and follow-ups." |
Slide 59 · Exercise: the whiteboard becomes the journal (Everyone)
- Take a photo of the whiteboard.
- Give it to the AI with one line about what today was.
- Ask for notes, open questions, and who said they'd try what.
- Check it against the board. Fix what it got wrong.
Capture as you go, then let AI organize it. That's the journal habit, done as a room.
Slide 60 · A new reflex (Both)
When you catch yourself thinking any of these:
- I have to figure this out.
- I don't know where to start.
- This takes forever.
- I've done this by hand five times.
- I wish I had someone to think this through with.
Could AI help me here, and what would it need from me?
🐝 The Question to Take With You
💡 What else am I doing the old way because I haven't imagined another way yet?
📞 Questions about anything covered in this workshop? Contact DTC at support@dtctoday.com or submit a ticket through the DTC Client Portal. See also the companion handout, Working With AI: A Practical Guide.