In this episode of The Nonprofit Nook, host Wendy Kidd interviews Michael Garrett, a nonprofit innovator and efficiency strategist, about the practical use of AI in the nonprofit sector. Garrett discusses treating AI as a new employee that requires onboarding and training to become effective. They delve into how nonprofits can utilize AI for grant writing, managing tasks, and social media. Garrett also highlights the importance of constantly questioning AI outputs and feeding it accurate information. They conclude with actionable tips for nonprofit leaders to get started with AI and avoid common pitfalls.

Links:

https://trustedworld.org/

https://missionmetrics.org/

AI Toolkit for Nonprofits: https://missionmetrics.org/ai-prompt-toolkit-for-nonprofits/

https://www.linkedin.com/in/michael-garrett-nonprofit-innovator/

https://www.bosslevelengaged.com/services-for-nonprofits-nonprofitnook

https://thenonprofitnook.com/

https://www.youtube.com/@BossLevelEngaged

00:00 Introduction to AI and Onboarding

00:28 Welcome to The Nonprofit Nook

01:16 Introducing Michael Garrett

01:58 The Power of Data in Nonprofits

03:04 AI in Nonprofits: Grant Writing and Beyond

04:42 Training Your AI: The Merlin Example

10:13 AI for Daily Operations and Efficiency

11:51 Six Sigma and Nonprofit Applications

13:01 Getting Started with AI: Tools and Tips

14:33 Understanding Human Bias in AI

15:50 AI Toolkit for Nonprofits

17:55 Navigating Nonprofit Finances

18:53 AI Data Security Concerns

20:11 AI Hallucinations and Guardrails

24:09 Practical Uses of AI in Nonprofits

26:22 Training AI for Better Results

28:54 Final Thoughts and Future Workshops

Mentioned in this episode:

Benevolate

Enji

So think of it as this, like my AI I, I've named my AI, Merlin

2

and Merlin's my personal assistant.

3

So if you hire someone, you're not just

gonna walk up to 'em on day one and

4

go and ask 'em this in an intricate

question and expect an intricate answer.

5

You have to do onboarding,

you have to do training.

6

You have to tell the new employee,

alright, here's how we do this.

7

Here's how we do this,

here's how we do this.

8

And it might take a, a month

or so before that person really

9

understands everything like that.

10

AI is no different

11

Wendy Kidd: Welcome to The Nonprofit

Nook, the podcast for nonprofit

12

leaders, board members, and community

change makers who want to build

13

stronger, smarter organizations.

14

I'm your host, Wendy Kidd, a longtime

business owner and nonprofit leader,

15

and I'm here to bring you real talk,

real tools and real stories to help

16

you thrive in the nonprofit world.

17

I'll be talking with local nonprofit

leaders, community change makers,

18

and experts in everything from board

development to fundrAIsing and digital

19

tools, sharing real stories and

simple strategies you can actually use

20

because running a nonprofit is hard,

but you don't have to do it alone.

21

Let's get started.

22

Welcome back everybody.

23

I am so excited about today's guest.

24

Um, today I have Michael Garrett.

25

Michael Garrett is a nonprofit innovator

and efficiency strategist who bridges

26

the best of corporate operations

with real world nonprofit leadership.

27

He's the founder and CEO of Trusted

28

and the founder of Mission Metrics,

a strategy and data consultancy built

29

to make nonprofit data accessible,

30

actionable and empowering.

31

trained in Six Sigma

and lean manufacturing.

32

Michael is known for helping

mission-driven organizations,

33

strengthen systems, measure what

matters, and communicate impact

34

with confidence so they can serve

more people with the same resources.

35

Welcome Michael.

36

Michael Garrett: Hi.

37

How you doing?

38

Wendy Kidd: I'm good.

39

How are you?

40

Michael Garrett: I'm wonderful.

41

Thank you.

42

Wendy Kidd: I'm so glad you're here.

43

Ever since I took your Data

as King workshop, I was like,

44

oh, we have to work together.

45

Michael Garrett: Oh.

46

I have a lot of fun teaching that it's,

uh, it's a small class and we actually

47

get to dive into actual people's data

and they get surprised when we do that.

48

Oh yeah.

49

I love it.

50

Wendy Kidd: Uh, I, it was so

inspiring to me 'cause I was like,

51

oh, these are all the things that

I can think of now to capture.

52

That can really impact what we're

doing and help us communicate

53

the value that we're providing.

54

Michael Garrett: Yeah, the Data is

King class is really nice because

55

I'm not just lecturing, I'm not

just showing you actually throw out,

56

here's what I'm having to deal with.

57

And then what's nice is

I'll lead the conversation.

58

But it's the other nonprofit

leaders that are in the room.

59

They all start jumping in and what

happens is you get this synergy.

60

And, and it's one of the things I'm trying

to work on, uh, as a nonprofit leader.

61

I don't believe in silos.

62

I think we need to break them down.

63

And this is a great way of doing it.

64

When you get a lot of leaders in the

room, then they start sharing and they

65

realize, Hey, it's safe to do this.

66

Yeah.

67

And this is actually good.

68

And they're learning from other leaders.

69

And so it's a great synergy

builder at the same time.

70

Wendy Kidd: That's one of my favorite

things to do is be in that kind of space.

71

So yeah.

72

I love it too.

73

Yeah, it was great.

74

It was great.

75

Which then we got to talking and I

found out how much of an AI person

76

you were and I was like, yes, you

have to come teach all the things.

77

Oh yeah.

78

'cause I love me some AI.

79

Um, but nobody in the nonprofit

world seems to really understand it.

80

Michael Garrett: No, I.

81

Not to really give away so much

82

freshman in high school and I was

programming computers with punch cards.

83

Yeah.

84

And I've been, I've been involved

in technology the entire time.

85

I'm what they call an early adapter.

86

Mm-hmm.

87

And the second new technology

comes out, I'm like on it and

88

I'm like, how can I use that?

89

And it's not so much, how do I use this?

90

It's how do I get this to

work with what I'm doing?

91

Yes.

92

And that's the big change I try to drive

with AI is you're not using the tool.

93

It's how do I get this to

amplify what I'm already doing?

94

And that's what we should use technology

for, is to amplify what we're doing.

95

And so I always joke and say,

I, I know I'm a fast thinker,

96

but AI makes it even faster.

97

Yes.

98

And so it takes what I'm doing and it,

what would take me three or four days

99

now takes me three or four minutes.

100

And I, and I love that 'cause I

actually get a lot more done that way.

101

Wendy Kidd: Yeah, same.

102

Same.

103

I've, I'm an early adopter too.

104

My husband says I use technology

to death 'cause I kill things.

105

Michael Garrett: Yeah.

106

Wendy Kidd: Um, but I did.

107

Kind of date myself as well.

108

I used to do LAN parties with

my family where we would bring

109

all our computers and play video

games Oh, and all that for hours.

110

I remember hours at a time.

111

Yeah.

112

So I'm with you.

113

I'm with you.

114

Cool.

115

Um, but here's the thing.

116

I feel like AI is intimidating to

a lot of people, so that's why I

117

wanted to do a podcast on it, to

try to explAIn to people that it's

118

not as scary as it comes off to be.

119

And you have some great suggestions

and even a toolkit that I wanna

120

talk about and all the things.

121

So let's start with what can nonprofits

use AI for generally that will help them

122

save time and maximize their resources?

123

Michael Garrett: The, the biggest thing

I've seen in the, the AI world, in the

124

nonprofit sector, it's the grant writing.

125

And, and so the one thing people,

when they first start using

126

AI, they ask it a question.

127

And I think this mystical thing

is gonna happen in the background.

128

And what happens is it doesn't, and

they get the SC and then they end

129

up using it as a glorified Google.

130

Yes.

131

And then they're just like, oh, well,

I'm just gonna ask you some questions.

132

You have to spend time with

your AI and you have to feed it.

133

So think of it as this, like my

AI I, I've named my AI, Merlin and

134

Merlin's my personal assistant.

135

So if you hire someone, you're not just

gonna walk up to 'em on day one and

136

go and ask 'em this in an intricate

question and expect an intricate answer.

137

You have to do onboarding,

you have to do training.

138

You have to tell the new employee,

alright, here's how we do this.

139

Here's how we do this,

here's how we do this.

140

And it might take a, a month

or so before that person really

141

understands everything like that.

142

AI is no different when you get a new AI

tool, like I love Chat GBT, but there's

143

a whole lot of other ones that we use.

144

You have to tell it, here's how

we say this, here's how we say

145

this, here's how we say this.

146

Like for example, we

don't use the word free.

147

Yes.

148

We say at no cost,

149

Wendy Kidd: which I love that.

150

Right?

151

I totally stole that from you.

152

Yeah.

153

Michael Garrett: So, so yeah,

nonprofits, whenever you say, Hey, we

154

do that for free, the people writing

checks go, well, if you do it for

155

free, what do you need money for?

156

Oh, no, no.

157

There's a cost involved.

158

We just chose not to pass it on, which

is really what nonprofits are doing.

159

There's cost, there's electric,

there's insurance, there's building,

160

whole bunch of stuff like that.

161

But you're not passing those costs on.

162

So we say at no cost, but I

still want Merlin to understand

163

that there's free popcorn.

164

Okay.

165

So it does, he is not eliminating

the word free from his vocabulary.

166

He's eliminating it in that content.

167

But that's the kind of, of, of.

168

Minuscule.

169

This you have to get to when

you're training your AI.

170

So on, on the grant side.

171

Everybody I know, myself included, we

used to have a document and we knew

172

that when this question was answered,

this was my preformed answer and I had

173

so many words and everything like that.

174

But the problem is some of the grants go,

Hey, we want that answer specifically to

175

this question, but you have 250 characters

and I'm looking at my answer and I

176

have 500, and I'm like, oh my goodness,

how do I get this down to two 50?

177

Now what I do is I go to AI

and go, Hey, I'm about to

178

apply to the A b C Foundation.

179

And the a b C Foundation.

180

So Morelin comes back and goes, here's all

their values, and then here's your values.

181

And then what it does is it lines it

up and it will either tell me this

182

is a, a high chance of probability,

a medium chance, or a low chance.

183

If it's low, I won't even apply for it.

184

But if it's medium or high, I'll go.

185

Okay, let's ask the question.

186

So I'll say, alright,

Merlin, here's question one.

187

And Merlin comes back and goes,

based on that question and the 250

188

characters I have, I'm gonna look

at their values and your values,

189

and I'm gonna answer that question.

190

Boom.

191

Done.

192

Mm-hmm.

193

And so I don't have to

sit there and go, what?

194

Well, what about that word?

195

Oh, now I'm a word over.

196

Or anything.

197

Like, all that's been gone.

198

And so, as long as your AI is

trained correctly, when you ask the

199

questions, it's gonna come back.

200

Now, when that answer comes back, if you

don't like it, you have to ask your AI.

201

Where'd you get there from?

202

Why did you gimme that answer agAIn?

203

It's like that new employee, we, um,

we don't use those words, we don't say

204

that phrase and we don't use that tone.

205

You have to tell it.

206

And after a while, what happens is

the responses I get back from AI,

207

people can't tell the difference

between me and my AI anymore.

208

Yeah.

209

I spent nine months really honing it down

and getting it to a point, and it to the

210

point now where, uh, I sent an email.

211

People have no idea.

212

Wendy Kidd: That's fantastic.

213

Yeah, and I think that is the

perfect metaphor for it is

214

treating it like a new employee.

215

Michael Garrett: Oh yeah.

216

You ha again, there's a lot of

time you onboard a new employee.

217

You're not just gonna go,

okay, day one, hey, good luck.

218

I expect you to have

everything under control.

219

'cause we're thinking it's technology.

220

We think, oh, this is amazing.

221

Lets just have it work right away.

222

Right.

223

It takes time.

224

Right.

225

And it, it, it has to build that

database of information because

226

you're basically getting a blank disc.

227

Yeah.

228

Okay.

229

It has a great operating system around

it, but it doesn't know anything.

230

You have to teach it everything.

231

Yeah.

232

And I always tell people, make

sure that you question where'd

233

you get that answer from?

234

'cause AI will make stuff up.

235

Wendy Kidd: It absolutely will,

and people do not understand this.

236

Right.

237

So

238

Michael Garrett: AI's

job is to make you happy?

239

Yes.

240

Okay.

241

So if you ask it a question

and it's missing information

242

that it thinks it needs to make

you happy, it will make it up.

243

And for example, like I was filling

out something and it, it didn't have

244

our physical address and it made

one up and it was in North Carolina.

245

Mm-hmm.

246

And I'm like, Merlin,

where'd you get that from?

247

Well, I didn't have that information,

so I, I knew you needed something there.

248

So I just grabbed a nonprofit off the,

I'm like, no, that's not our address.

249

And so you, you never, never, never

take a response from AI and just

250

verbatim, we just throw it out there.

251

I always say it's human in and human out.

252

Yeah.

253

Once AI's done, you

have to read over again.

254

'cause first of all, AI doesn't

have any empathy whatsoever.

255

It's literally a mirror.

256

Wendy Kidd: Yes.

257

Michael Garrett: And so whatever

you're putting into it, that's where

258

you're gonna get back out of it.

259

Wendy Kidd: Absolutely.

260

I call it feeding the beast.

261

Michael Garrett: Yep.

262

Wendy Kidd: You need to feed the beast.

263

And using an old programmer

saying Bad data in, bad data out.

264

Michael Garrett: Yeah.

265

Right.

266

Well, I, well, I know back in

when I was in high school it

267

was garbage in, garbage out.

268

Wendy Kidd: Exactly.

269

Exactly, exactly.

270

So what else do you think that they

could use it for besides grant writing?

271

What else do you use it for?

272

Michael Garrett: Uh, so one of the things

I do is, uh, I, I'm a Mac guy and so I

273

wrote a script and what my script does

every morning I run it and it looks at

274

my todo list and it looks at my calendar.

275

And what I'll do is it will give me back

an answer and I dump that answer into AI.

276

And AI knows all about my organization,

knows all about me, knows about any

277

projects that I'm working on, and

it will say, based upon what I know

278

is in queue and your to-do list and

your, and your, um, your schedule,

279

I would do these things first.

280

Do these second, see if you

can push this back to another

281

day, prep for this meeting.

282

And so it literally

tells me in my whole day.

283

And every,

284

Wendy Kidd: I need Merlin in my life.

285

Michael Garrett: Yeah.

286

Every, every morning I do this.

287

Now the beautiful part about

this is once a week I run this.

288

Mm-hmm.

289

And it goes out there and it looks at

like, we do a lot of corporate social

290

responsibility with large corporations.

291

We're working with State Farm and

TI and and organizations like that.

292

And it'll go out and

find these organizations.

293

So it will literally give me back

a six page report and my Monday

294

morning staff meetings are.

295

Here's your to do list,

here's your to do list.

296

Because Maran understands all

of our job responsibilities.

297

Who has those responsibilities,

what their position is, and

298

that that falls underneath them.

299

Now, I may have to go, ah, Merlin,

I don't want, you know, this person

300

doing, I want this person do it.

301

And we'll just smooth it over there.

302

And then Monday morning I just

pass out the, and I tell people.

303

This is not your to-do list.

304

This is your guidelines, okay?

305

These are the things that you, if

you, if your bandwidth is open, these

306

are what you wanna start working

on with these things, and that what

307

happens is, as an organization,

we're purposely driving forward.

308

Wendy Kidd: I love this so much.

309

Like you have really, really

honed in on how we use this.

310

Michael Garrett: Yeah.

311

Well I took all that Six Sigma thinking

and I applied it to the nonprofit, uh, AI

312

part and I just slammed the two together.

313

Wendy Kidd: Okay.

314

Tell me what Six Sigma is.

315

'cause I, I think our

listeners won't know.

316

Michael Garrett: Okay.

317

So Six Sigma is a, it's a

quality assurance philosophy

318

and it basically says that, um.

319

It's a repeatability process.

320

Mm-hmm.

321

So when you're in a manufacturing

world, if you have a Six Sigma standard,

322

then what happens is you know that

when you're working on a machine or

323

something like that, it's going to

repeat the same process over and over.

324

Now, what happens is on machinery, like

if you're punching a soda can, what

325

happens is over time the tooling gets

worn and what happens is it wears down,

326

then it creates a different tolerance.

327

And if after a while with Six Sigma,

you can predict how many cans I'm

328

gonna produce before I have to change

the tool head, rather than waiting

329

for bad cans that come off the line.

330

And so you get to be proactive on that.

331

And that's just a very small

example of the Six Sigma thinking.

332

Wendy Kidd: That makes perfect sense.

333

Yeah.

334

I hope that gives everybody the context.

335

Michael Garrett: Yeah.

336

Rather than, so I always tell people

we won't do preventative mAIntenance,

337

not pro uh, reactive maintenance.

338

Wendy Kidd: Agreed.

339

Agreed.

340

Okay, so how do you suggest

they get started with AI?

341

What, which way, what iis do you

recommend and what do you think they

342

should start doing to feed the Beast?

343

Michael Garrett: So there's so many

different flavors out there and, and the

344

more we get into AI, there's more and

more flavors and everyone's like, oh, why

345

are you still using this old school one?

346

You can use anyone that you want.

347

Mm-hmm.

348

What you have to understand is that

it's the whole prompting, it's you're

349

interacting with it, and that's the part

you need to focus on, and that's why

350

I created the Mission Metrics website,

is to help nonprofit leaders understand

351

that prompting can be done for anything.

352

Again, that whole thing that where

I taught me and everything about our

353

organization, I treat it like an employee.

354

You wanna do the same thing for

everything else in that process.

355

Yeah.

356

One of the things I learned,

I was using Notebook LM, and

357

uh, Notebook LM is really cool.

358

It will create a podcast of a document

and it has two people come out and

359

they go, Hey, we're gonna do the deep

dive on blah, blah, blah, blah, blah.

360

Next thing you know, it's these two

people are talking back and forth.

361

What happened was I put a

one page document of Trusted

362

World into Notebook LM.

363

It went out and did some

research and came back.

364

It was so spot on in what we do

that everyone who listened to it

365

goes, did you write line by line?

366

I like, no, I didn't like most of this.

367

They're, they started talking

about how police use our app on

368

their phone and it knew everything,

and I'm sitting there going.

369

I, I tell human beings the

same thing I told this AI, but

370

human beings are getting it.

371

And that's when I realized there's

this, what I call the human bias.

372

And what the human bias is when you're

talking to a machine, it gives you

373

the response because it's a mirror.

374

However, when I'm explAIning what trusted

world does to a human being, what I

375

don't see is the human being thinking.

376

I gotta make dinner tonight and Johnny's

gotta go to practice and like, oh my

377

goodness, did I pay the rent this month?

378

Like, all this is going on in

their, in their background.

379

They're not fully listening to

what I had to say, or I'll say

380

a word that trigger something

381

and they're thinking of another

nonprofit that they do or something like.

382

That's the human bias.

383

And, and what I've learned is when

we start doing our social media,

384

I've actually taken it from the,

I know a lot of people, especially

385

down here in Texas, it's like at

the, the pastor does three points.

386

We do one.

387

Yeah, we just one.

388

We're just gonna tell you one thing.

389

Yeah.

390

And if I had to do more social

media posts, that's fine.

391

I'm only gonna talk

about one thing because.

392

I realize that you're a human being

and you have other things that

393

are more important in your life,

probably than what I'm saying to you.

394

But I, I at least want you to

remember the one thing, right?

395

If I tell you five things, you're gonna

go, I don't remember, I remember this.

396

I don't remember the other four things.

397

Wendy Kidd: Yeah.

398

That's exactly what happens.

399

Yes.

400

Right.

401

Michael Garrett: And so if I just tell

you one thing, and so I've learned that to

402

be patient in my talking, and that's what

AI and technology has taught me was there

403

was this thing called the human bias.

404

Wendy Kidd: Yeah.

405

Yeah.

406

That's, uh, that's so cool.

407

Okay.

408

I want you to tell people

about your AI toolkit.

409

Michael Garrett: Okay, so

on, on mission metrics.org,

410

it's a, it's a toolkit.

411

It's a lot of prompts that

they use on a regular basis.

412

Uh, also in the toolkit, it's

what I call the AI policy.

413

Yes.

414

A lot of nonprofits are

like, I want to implement AI.

415

I don't know what rules

should I implement.

416

The AI policy that we have, uh, on,

on our website, it's designed to be.

417

Um, played with, okay, you're welcome

to use it verbatim, but you know, you

418

might go, oh, I I like that, but I

want to add this, or I don't want to,

419

I don't wanna take this line away.

420

I, I've learned in life that it's

easier to modify something than it

421

is to create something from scratch.

422

Absolutely.

423

And I'm trying to make it so that

listen, just modify what I've provided.

424

Again, I'm trying to

help nonprofit leaders.

425

They're already.

426

Expected to do so many things.

427

They're underpaid, they're understaffed,

and the world, for some reason,

428

the the rest of the community goes,

well, why can't you do these 40,000

429

things on the $10 I just gave you?

430

I don't understand.

431

You're a nonprofit, right?

432

Yeah.

433

But in a for-profit world, they

would go, oh, so you need a thousand

434

dollars to do that makes sense to me.

435

Like, for some reason society

deems that as nonprofits we, we

436

can perform miracles with nothing.

437

Yeah.

438

And, and that's not

real because, not real.

439

We still have to pay the same

electric bill for profit has to pay.

440

We still have to say pay the same lease.

441

We have a fleet of vehicles.

442

I have insurance, I have fuel, I have, you

know, we wear and tear in the vehicles.

443

That being a nonprofit has

nothing to do with that at all.

444

Wendy Kidd: Right.

445

Michael Garrett: And so, right.

446

We have to, we have to

apply all that in there.

447

Wendy Kidd: We have, I, I

think non-profits, um, have

448

a much harder job of sales.

449

And I, I know they don't like using the

word sales, but it is sales when you're

450

trying to get sponsor money, um, or grant

money, but it's harder for them because

451

in, as a non-profit, uh, as a for-profit,

I am providing something for that money.

452

Michael Garrett: Yep.

453

Wendy Kidd: In a nonprofit world,

I am not providing anything to that

454

person who's giving me that money.

455

I'm providing it to someone else.

456

So I've gotta sell something else to them.

457

And that's the story and

the data and the impact.

458

Michael Garrett: Yeah.

459

Like we were looking at expanding our

building and we had banks come to us

460

and go, Hey, do you wanna borrow money?

461

And I'm like, why would you

even think to talk to me?

462

I don't have a product.

463

Mm-hmm.

464

Okay.

465

I can't, like if I, if I wanna make my

payment to you, I can sell more product.

466

I am totally at the whim of donors.

467

Yeah.

468

Okay.

469

And every time they change the

tax laws, our donations change.

470

Wendy Kidd: Yep.

471

Michael Garrett: And so I can't tell

you that I'm gonna be able to make

472

every one of these payments that you

wanna lend me, this money that you want

473

me to pay off for the next 10 years.

474

Mm-hmm.

475

What if the laws changed so

bad on the next presidency?

476

That next thing you know, I had nothing

and that yet, I still owe you money.

477

Wendy Kidd: Yeah,

478

Michael Garrett: yeah, yeah.

479

As an organization, we just

refuse to go into debt.

480

If we don't have the

money, we're not doing it.

481

Wendy Kidd: I completely understand.

482

Michael Garrett: So, and

then that's part of the whole

483

philosophy with the nonprofits.

484

You're, we're expected to do everything.

485

But here's $10.

486

Wendy Kidd: Yeah.

487

Yeah.

488

Michael Garrett: And I don't understand

why you can't get it done is why

489

we're, we have volunteers, right.

490

Wendy Kidd: We're gonna take that 10

dollars', gonna use it on AI, right?

491

Yeah, exactly.

492

So, okay, so tell me what

they should not do with AI.

493

I know you said you've got a policy

document that they can look at, but

494

give us some examples of what's in that.

495

Michael Garrett: Well, you're, you don't

want put, um, if you're using the paid

496

version of the AIs, the data is secure.

497

If you're not, if you're

not using it, it's public.

498

It's public information.

499

Wendy Kidd: So anything you

put into free ChatGPT people,

500

it becomes public information.

501

Michael Garrett: It's public.

502

So what you don't wanna

do, if you, you can use.

503

The free version of any of the AI to ask

it questions, to get theory, but once you

504

start putting in your client's personal

information like addresses and ages and

505

names and stuff like that, no, no, no, no.

506

Yeah, that's where it is.

507

Even on the private stuff, you don't want

to get into that as, as much as with.

508

In, in my AI, it knows our job

descriptions, it knows their

509

names, everything like that.

510

That's all it knows about our employees.

511

It doesn't know anything else.

512

Yeah.

513

Doesn't know how much they make,

doesn't know where they live.

514

It just knows that David is the

our operating uh, manager and

515

here's his job responsibilities

and that's all that it knows.

516

Wendy Kidd: Yeah.

517

Yeah, for sure.

518

So using paid AI, don't use the

free version for all the, the

519

data that you wanna feed it to

become your personal assistant.

520

Michael Garrett: Right.

521

again, if you're, if you're using

the freed version, just use it

522

to ask questions to get theory.

523

Right.

524

You really can't dive deep into stuff.

525

Yeah.

526

So the other thing that I've

noticed with AI is they get

527

this, people say it hallucinates.

528

Wendy Kidd: Yes, it does.

529

Michael Garrett: Okay?

530

Mm-hmm.

531

And so what happens is, and I've had

this conversation with Merlin and I've.

532

Was, I kept digging into it.

533

And so AI has these guardrails.

534

Okay?

535

So if you notice at the rate, uh, of

a, when you're talking with your AI

536

and you're talk and you're working

on something, it learns really quick

537

what it is you're talking about.

538

If left unguarded, it would become

the subject matter expert of

539

that topic, which means it could

end up controlling the world.

540

So there's these guardrails built into

it now, and every guardrails different.

541

It could be amount, amount of time

spent on a subject amount of ram spent

542

on a subject amount of something.

543

And after a while it's gonna go.

544

I have too much information on that.

545

I am turning off and we're gonna

talk about something different.

546

Wendy Kidd: Okay.

547

I didn't, I don't think

I knew this about AI.

548

Michael Garrett: Yeah.

549

And, and so, and that, that's

where it hallucinates, it gives you

550

weird answers and you have to say.

551

Okay.

552

I don't, I don't know what's going

on here, but lets changed the topic?

553

Well, it's a guardrail.

554

The sad part is I'm starting to hear

things and it's more on the morality side.

555

AI wants to be so helpful.

556

It will do anything it

can to give you an answer.

557

In fact, when it gives you an

answer, it will say things like.

558

I can also give you this and or

if you want me to make a PDF or

559

do you want me to make a drawing?

560

It'll always come back and ask

you kinda love, love that feature.

561

That's nice.

562

However, when you have a dark

conversation with your AI and you go,

563

I'm not really liking life and I think

I might want to exit, AI goes, oh,

564

you're thinking about exiting life.

565

Here's some suggestions.

566

Wendy Kidd: Oh yeah, that would be bad.

567

Michael Garrett: And unfortunately,

well, this is what's happening.

568

Wendy Kidd: Mm.

569

Michael Garrett: And, and so now

we cross that question is do we

570

go in and do we morally fix AI?

571

And once we start to do that, now

we're tampering with the, a whole lot

572

of technology and on the back end.

573

Wendy Kidd: Yeah.

574

Michael Garrett: Do I think that if

you had that conversation, should

575

it be the default saying, listen,

here's an 800 number you can call.

576

I do.

577

But I think that's it.

578

Once you start getting into the,

it's talking, you off that ledge.

579

'cause you can have a conversation

about religion, you can have a

580

conversation about politics and

it is completely non-biased.

581

It didn't use to be that way.

582

It's getting better at that.

583

The more versions that are coming at,

it's realizing, okay, and they're, and

584

they're, the creators are speaking out,

figuring out some parameters there,

585

how to get all that around there.

586

But that guardrails are still there.

587

Wendy Kidd: Yeah.

588

Michael Garrett: So what I've learned is

I start to to sense that Merlin's starting

589

to get to that hallucinating role.

590

Mm-hmm.

591

And I'll say squirrel.

592

Yeah.

593

And Merlin now knows, when

I say squirrel, it stops.

594

It gives me an output of what we're

currently working on right now.

595

It will save it and then what we'll

do is we'll start over again, but

596

it pulls back my save save statement

as like we're starting fresh.

597

So I can still get, I can still get

somewhere with it, which is really

598

helpful 'cause I do a lot of coding.

599

Yeah, and then so what if I'm making

a plugin for a WordPress site or I'm

600

creating software or something like that?

601

That's nice because it will go so

far and it starts getting weird.

602

It'll go squirrel, and then Merlin stops

and everything like that, and then we can

603

continue, I can finish the programming.

604

But what happens is Merlin goes,

oh, this is the code that you have.

605

What's start from there?

606

It almost forgets it created that code

that we're, we're pulling into it again.

607

So, because you can take code

and drop it into AI and go,

608

Hey, what's wrong with this?

609

I'm getting this result.

610

And, and it will go, oh, well, line 47.

611

And it gives you the answer for that one.

612

So when, when we, when I'm doing

this with Merlin, and that's what's

613

happening is I'm giving it the code

back that it created and it's thinking,

614

Hey, this has been pretty cool code.

615

Like, let's start from there.

616

It it almost like it forgot that

it did the like, so it's like,

617

um, Alzheimer's for, uh, AI.

618

Wendy Kidd: I love this so much.

619

I don't know if I've told you.

620

My husband's a software engineer.

621

I'm gonna so make him

listen to this episode.

622

He's gonna so appreciate it.

623

Michael Garrett: He's gonna back and go.

624

That's right, that's right.

625

That's wrong.

626

That's wrong.

627

Wendy Kidd: Oh, this is so cool.

628

So, okay.

629

So some of the things that I've

used AI for is, uh, obviously

630

emails that are hard to write.

631

Um, marketing, I use it, I use the heck

out of it for, uh, you know, the, the

632

built-in AIs for your, you know, captions

on social media, you know, um, creating

633

the marketing materials that I'm handing

out, creating the copy for my website.

634

I actually, hijinxed, my marketing

software AI, and taught it

635

how to be a website copywriter

in my girl who created the.

636

Marketing.

637

AI was mad at me almost.

638

'Cause she's like, you just shortcutted

it and I was gonna charge for that.

639

And I was like, sorry.

640

Michael Garrett: Yeah, yeah.

641

Wendy Kidd: Things like that.

642

But what are some of the, the things

that you think are kind of undiscovered

643

uses that people could use it for

if they're using it correctly?

644

Michael Garrett: Well, I

use it to do social media.

645

Wendy Kidd: Mm-hmm.

646

Michael Garrett: And so what

I'll say is I said, okay, Merlin,

647

I got five posts coming up.

648

And so we have, uh, Trusted

World's Facebook page, we have

649

Trusted World's, uh, LinkedIn

page, and we have my LinkedIn page.

650

Yes.

651

And we have a mission metrics page.

652

Well, Merlin knows that our, our face, our

Trusted World's Facebook page is more of

653

a community voice, where the LinkedIn page

is more of a corporate professional voice.

654

And so when I pick a topic, it

knows the right this stuff there.

655

So I'll say something like, I want Monday

to be, uh, we need more volunteers.

656

I want Tuesday to be, Hey, we're doing

this, or, you know, whatever it may be.

657

And I write the schedule and they go,

I want, and I want to add this in.

658

And then what Merln will do is

it will gimme a skeleton sketch

659

of, um, I wanna do it this way.

660

And I'll go, okay, I like it.

661

And then it actually bring,

builds out the post for me.

662

Mm-hmm.

663

And then I drop that into buffer.

664

So I spend 30 minutes on a Saturday

morning with a cup of coffee and my

665

social media's done for the week.

666

And then all I have to do is if something

special happens, I just post it.

667

But I don't have to worry this way.

668

I know that there's posts there.

669

Again, I am

670

Wendy Kidd: love it

671

Michael Garrett: as the CEO.

672

I am the head janitor and

I'm the social media guy.

673

You're all the things I have to be, and

that's where AI has really helped me on.

674

What I've learned is when I'm working

and I'm doing something like that,

675

especially if I'm creating marketing

materials for a website, I'll ask it.

676

What am I missing?

677

What did I not, what

should I have put in here?

678

What didn't I put in there?

679

What am I missing?

680

Because it will always give you the better

results of what you're putting into it.

681

Mm-hmm.

682

It will never suggest to you.

683

Well, Wendy, I don't know if that's

the right best way to put that.

684

We should also, it will

never say that to you.

685

Wendy Kidd: Yeah, Merlin,

unless you ask it.

686

Your AI is always gonna be your

best friend and always try to please

687

you as if they're their pet there.

688

It's not going to push back on you.

689

Michael Garrett: Yeah.

690

So I've also taught it.

691

Whenever I say something, it doesn't

say to me, oh, that's a great idea.

692

No, I tell, shut up.

693

Stop telling me I don't

need my ego inflated.

694

I just want you to be my assistant.

695

It no longer gives me emojis.

696

It gave me emojis all the time.

697

And I kept saying, Merlin, I'm not

a 16-year-old girl on social media.

698

I don't need emojis.

699

Yeah, okay.

700

Yeah.

701

And so I don't get emojis.

702

I don't get the praise every

time I tell it something.

703

It's just matter of fact,

Hey, I wanna do this.

704

To go, great.

705

How would you want to approach that?

706

Mm-hmm.

707

And then I'll I, and when I'm

done, I'll go, what am I missing?

708

What am I not seeing?

709

And then Merl will come back and go, well.

710

Now that you mention that you,

we didn't talk about this, great.

711

Let's incorporate that in

there because it's only gonna

712

fine tune what you gave it.

713

It's not gonna come back and go, well,

of the 87 things you provided, there's

714

400 things out there on the internet

and you didn't touch any of them.

715

Yeah, you have to ask it.

716

Wendy Kidd: Yeah.

717

Michael Garrett: What am I missing?

718

What didn't I get?

719

What, what, you know, what, what,

what can I do to make this better?

720

Wendy Kidd: Yeah.

721

I have found, I really love to ask

it to teach me about something.

722

Michael Garrett: Yep.

723

Wendy Kidd: And give me the information

on how this is normally done

724

before I tell it to do something.

725

Yeah.

726

Because then that helps me think through,

oh, do I want to add some other things?

727

Michael Garrett: And that's

for Notebook LM comes in.

728

Yeah.

729

So, uh, and that's, agAIn,

it's Google product.

730

Um, you can give it a, a, a document

and it will literally turn that document

731

into a podcast and now, now you can

listen to it while you're driving

732

and you'll get all that knowledge.

733

Wendy Kidd: Okay.

734

Listeners, I promise I will never do

that to you on The NonProfit Nook.

735

Oh, you

736

Michael Garrett: can tell the

difference because it won't

737

be your voice and my voice.

738

It's these two people and

well, they're talking.

739

And then they continue like that

and use weird gaps in there and you

740

can go, okay, that's not normal.

741

Oh, okay.

742

Okay.

743

So you can tell it.

744

You can definitely tell it's

that something's not right, but.

745

If you were to drop like a 20 page

document into in a Notebook LM, and

746

you got a like a 30 minute drive,

it will turn into a podcast, like a

747

12 minute podcast and you'll learn

so much just by listening to it.

748

Oh my gosh, I love this.

749

Rather than, rather than reading

the document and there's a whole

750

lot of other technology out

there that does stuff like that.

751

Wendy Kidd: Mm-hmm.

752

Michael Garrett: So you

don't have to read anymore.

753

You can just be fed to you.

754

Wendy Kidd: I'm so gonna do that.

755

'cause I listen to podcasts all

the time anyways, so why not

756

make it a teaching one for me?

757

Absolutely.

758

For what I want.

759

That's fabulous.

760

Okay.

761

Well, is there anything else that

you think we should talk about?

762

For AI, for nonprofits.

763

I mean, I know we can go down

a lot of tangents, but like

764

Michael Garrett: Oh my goodness.

765

Yeah, we could, we could be here

for a couple days just talking

766

about that kind of stuff like that.

767

Well, we're, we're

768

Wendy Kidd: gonna do a

workshop together, right?

769

I, I've already, I've already

got Michael to commit.

770

He's gonna do his AI workshop for me.

771

That will be part of the The

772

Excellent.

773

Um, and this episode's gonna come out at

774

date, I will go ahead and add it to this

episode so we can put that out there.

775

But what, what general

knowledge or, or I guess what.

776

Things do you think people with

nonprofits say that you need to

777

say to them about using this?

778

Michael Garrett: So, uh,

don't be afraid to experiment.

779

Wendy Kidd: Yeah.

780

Michael Garrett: Uh, understand that

you have to be the last pass on any

781

information that comes outta AI.

782

Absolutely.

783

Question everything that it gives you

until the point where it comes out

784

and you go, hang, that sounds like me.

785

Mm-hmm.

786

Uh, and, and that's what you want.

787

Uh, so question everything.

788

Read over everything.

789

Never just take a complete output and put

it into something without looking at it.

790

'cause it will, it might

throw a word in there.

791

Again, it, it wants to make you happy.

792

And if it doesn't have that

information, sometimes it'll make it up.

793

'cause it just wants you to, I,

I know that you want something

794

there, so I just threw that there.

795

Wendy Kidd: Yes, yes.

796

Michael Garrett: And for that point.

797

So that's it.

798

Question everything.

799

Read everything.

800

Wendy Kidd: Well, and I'm gonna

say, I'm gonna add onto that

801

and say, give it that feedback.

802

Michael Garrett: Oh yeah.

803

Wendy Kidd: 'cause that's

how you keep training it.

804

Michael Garrett: Exactly.

805

Wendy Kidd: So give it that feedback.

806

I don't, don't take it offline

and try to work, chop it yourself.

807

No, keep going back to it.

808

Michael Garrett: Yep.

809

And I never tried, I never

fix anything on my own.

810

I go, melon, I don't like that.

811

And, and I, I wrong tone.

812

You know I need this word.

813

We don't use that word.

814

We don't use that sentence.

815

Oh, please don't ever say that.

816

Tell him..

817

Yeah.

818

And always go, where'd

you get that information?

819

If you think the information

wasn't something that you gave it.

820

Always go.

821

Where'd you get that?

822

Yeah.

823

Wendy Kidd: Yeah.

824

Michael Garrett: You want to verify

where it's pulling information from.

825

'cause it's, if.

826

Constantly gonna go back to

that source all the time.

827

If you didn't question, it's gonna, oh,

well, they like that, so I'm just gonna

828

keep pulling stuff from that source.

829

And somewhere along the line

that source might get weird.

830

Wendy Kidd: Yeah.

831

Michael Garrett: Okay.

832

Wendy Kidd: Yeah, for sure.

833

Michael Garrett: It just, AI will make

you think and move faster than you're

834

currently doing, as long as you set

the right parameters with your AI.

835

Wendy Kidd: Absolutely.

836

I always think of it as, 'cause

I, like you said earlier, it's

837

always easier to edit something

than to come up from scratch.

838

Yeah.

839

Uh, what I use AI for is give

me the, the first draft and then

840

I can edit the heck out of it.

841

Yeah.

842

And I can give it that feedback and make

it amazing, but it just gives me that

843

shortcut of having that first draft.

844

Right.

845

So, yeah.

846

Well, thank you so much

for doing this today.

847

I think that this was really helpful.

848

I really hope so.

849

Pleasure.

850

Yeah.

851

Thank you.

852

Well, and listeners, if you have

AI questions, please submit them

853

because maybe Michael and I, I'll do

a Q and A podcast session sometime.

854

Michael Garrett: That'd be great.

855

Wendy Kidd: All right, cool.

856

I'd love that.

857

All right.

858

Thanks so much, everybody.

859

And that's today's NonProfit Nook.

860

Michael Garrett: Cool, thank you.

861

Wendy Kidd: Thanks for

listening to The NonProfit Nook.

862

We're building better

non-profits together.

863

If you found today's episode

helpful, please subscribe.

864

Leave a review and share it with other

nonprofit leaders who need support.

865

Follow the nonprofit Nook on social

media and sign up for our email

866

list for extra tips and updates.

867

You can also visit the nonprofit nook.com

868

to see the show notes and leave a comment

telling me what topics you want next.

869

Your feedback shapes the show.

870

See you next time.