Learn More about the Life First Business Lab : https://www.deeplyrootedbusiness.com/lfbl?podcast
If you've ever asked ChatGPT or Claude to write something for you and ended up with robotic, generic content full of emojis, buzzwords, and phrases you'd never actually say, this episode is for you.
Jessica and Rachel break down the exact framework they use to get AI outputs that require minimal editing while still sounding authentic and on-brand. They introduce the ROOTS Framework, a five-part prompting method that transforms AI from a generic content machine into a well-trained team member who understands your voice, goals, and business.
You'll learn why most AI failures are actually prompting failures, how to provide better context, and how to create reusable prompts that save hours every week.
Let's unpack:
Check out the Life First Business Lab: https://www.deeplyrootedbusiness.com/lfbl?podcast
A plug-and-play AI employee membership designed for non-techy business owners who need real support without building everything from scratch.
Meet Your Hosts
Jessica Walther is the founder and CEO of The Launch Collaborative and Sustainable Success Systems. As a launch strategist and systems consultant, Jess is dedicated to helping solo business owners and small-but-mighty teams build businesses that deliver both peace and profit. She specializes in creating sustainable growth strategies that align with her clients' values and lifestyles.
Rachel Lopez is the founder and CEO of Gal Marketing Agency, a boutique email marketing and strategy firm. With over a decade of experience, Rachel helps heart-driven entrepreneurs craft intentional marketing strategies that attract, nurture, and convert leads sustainably. Her human-first approach ensures that marketing efforts feel authentic and effective .
Together, Jess and Rachel blend systems, storytelling, and soulful strategy to help you grow a business that's deeply aligned with your life—not just your revenue goals.
Connect With Us:
Hang Out & Say Hi!
All right, so as an AI power user, sometimes I get personally
2
:offended where people complain about
AI and it not giving the output that
3
:they need because they type something
like, "Write me an Instagram caption
4
:about my new offer," and they get mad
when AI gives them, "Rocket ship emoji.
5
:Exciting news.
6
:I'm thrilled to announce my brand new
offer that will transfer your business!
7
:Exclamation Point.
8
:Exclamation Point.
9
:DM me to learn more."
10
:It's generic, and they get mad, and
then they blame my friend Claude,
11
:and it's really not his fault.
12
:Rachel: Yeah, I mean, the extra emojis, I
use a lot of exclamation points in my copy
13
:as it is just 'cause that's how I talk.
14
:But that is not everybody and making sure
the things that you would never say, maybe
15
:you would never say the word thrilled.
16
:Ultimately, those are the realities
of people that have tried and then
17
:come down and say , "Oh, I sound
everybody else on the internet."
18
:Which pretty much means you sound
like no one 'cause there's not
19
:truly an identity rooted in there.
20
:so people that have experienced
that rollercoaster of an emotion
21
:just ultimately end up deciding
that AI isn't good or it's not for
22
:them, it can't capture their voice.
23
:But we're here to tell you that it wasn't
the AI that failed you, it was the prompt.
24
:It was your request to the AI.
25
:handed it a vague sentence, and
expected it to ultimately read your
26
:mind, which I don't know about you,
I love AI and I love supporting it,
27
:but I'm not at that point where I
want my robots to be reading my mind.
28
:Jessica: Yeah, yeah.
29
:So the difference between getting
generic garbage or spending all
30
:of my usage tokens really, really
quickly, is to give it a better brief.
31
:Basically, AI is not a software, it's
not a tool, it's not a vending machine.
32
:It is basically an employee.
33
:So the more brief that you give it,
the better your output is gonna be.
34
:So today we're gonna be sharing
our exact framework we use to get
35
:outputs that we barely have to edit.
36
:It's called the ROOTS framework.
37
:It has five parts and it's gonna make
it super easy for you to remember so
38
:that whenever you're typing in your
AI, you're getting what you need
39
:faster and not wasting any time.
40
:So if you've used ChatGPT, Claude, and
you've gotten an output that's fine,
41
:technically correct, but totally soulless
and you felt like you needed to rewrite
42
:the whole thing, this is gonna be for you.
43
:You're gonna walk away with a
framework to follow so that AI feels
44
:more like you and less like a robot
45
:Rachel: Yeah.
46
:So quick energy check before
we dive into all of this.
47
:Really wanna make sure we're, you
know, holding ourselves true here.
48
:And if we're at that point where maybe
we've decided that AI doesn't work,
49
:obviously we talk about other things
on this podcast, but if you're over
50
:here thinking, " I'm gonna skip this
episode because AI has never given me
51
:anything that sounded like me, or it's
never given me any useful", we want you
52
:to hold that thought a little loosely
and ask you to stick with us today.
53
:Because like we had said,
it is not always the AI.
54
:It is most likely always the prompt.
55
:And that's the good news, is because the
prompt is the part that we can control.
56
:And the episode that we're walking through
today is ultimately the fix for it.
57
:We're gonna walk through roots one
letter at a time with examples, And
58
:we want you to stick around until the
end 'cause we're gonna show you how to
59
:save your roots prompts once, and then
you never have to type it again, and
60
:that's where the real time saving comes.
61
:Jessica: All right, so let's get into it.
62
:The first R is role.
63
:So before you tell AI what to do,
you have to tell it who to be, and
64
:this one move is gonna change the
output more than anything else.
65
:So think about the role.
66
:You need to tell it what
hat that it is wearing.
67
:You are a conversion copywriter.
68
:You are a registered dietician
who writes plain, like they're
69
:talking to their bestie, language.
70
:You are my operations project manager.
71
:Because without a role, AI is just going
to default to your average internet voice.
72
:With a role, it's going to start pulling
its knowledge from a specific lane.
73
:Rachel: If you ask a lawyer and
a kindergarten teacher the same
74
:question, you're ultimately gonna get
very different energetic response.
75
:You're gonna get a very different
context level of response.
76
:And that whole role of establishing
who it is that you're initially
77
:talking to with your AI conversation
is gonna set so much of the foundation
78
:before a single word gets written
in your prompt or in your output.
79
:So that little bit of context
is so incredibly powerful.
80
:And we say get as specified as possible.
81
:So you're not just a copywriter,
but you're a copywriter who writes
82
:for wellness practitioners or you're
not writing for corporate brands.
83
:Jessica: Yeah, basically you
get to describe what your
84
:dream employee gets to be.
85
:So right now I'm working on a new
employee for the lab called Bruce,
86
:and I want him to be able to write
SEO and AIO optimized blog posts.
87
:But I'm like, you're just not
an SEO optimized expert, Bruce.
88
:You not only write blogs that are
good searchable, but they spark
89
:aha and insights and they make the
readers wanna come back for more.
90
:The more you describe who you want the
AI to be, the more he will fall into that
91
:lane and get less generic, and the more
it's gonna sound like you because you're
92
:describing what the ideal person would
be who you would hire to do this for you.
93
:Rachel: Right.
94
:So then once you know who it
is that we're prompting, now it
95
:needs to know actually what, we're
trying to get it to accomplish.
96
:And this is where, again,
people get super weirdly vague.
97
:So this brings us to our step
two, which is O, objective.
98
:What is the goal of the prompt?
99
:Jessica: Yeah, so what are you
actually trying to accomplish?
100
:Not just the deliverable.
101
:Write a caption, that's a task.
102
:There's a difference
between a task and a goal.
103
:Write a caption that gets people
to comment so it boosts this
104
:post's reach, that's an objective.
105
:Write a subject line that encourages
the user to click, that's an objective.
106
:So getting that specific is gonna
produce two different kind of
107
:outputs and really help AI start to
be able to support your business.
108
:Rachel: Yeah, and I think this is so
important for the business owner too, to
109
:really understand, it's a good practice
to have when you're saying, "What exactly
110
:am I asking for, and what is the end
result that I'm hoping to achieve here?"
111
:I think so many times people are wasting
time when it comes to AI when they
112
:don't even know what the output is.
113
:They don't know, and sometimes
it's like, "I want it to look
114
:like this, I want it to sound like
this, I want it to feel like this."
115
:All of those various components.
116
:And when you haven't put that thought on
the human side of this prompt, it really
117
:does ultimately give you a less than
favorable output because you've kind of
118
:skipped a lot of that step hoping that
it's gonna fill in the gaps for you, and
119
:I don't believe we're at that point with
AI yet where it can read your minds.
120
:. Jessica: Yeah, yeah.
121
:So for example, this morning I was
writing an email to Ashley, the
122
:DOERS community that we're both in.
123
:And at first I was like, "Hey,
help me respond to this email.
124
:I wanna pitch this," blah, blah, blah.
125
:But I hadn't taken the time to like,
what are my objectives, giving it more
126
:context that she actually knows me.
127
:So it wrote in this way which
if I would've sent it to Ashley,
128
:she would've been like, "What the
hell kind of robot crap is this?"
129
:And I was like, no, write it
like she's a business peer.
130
:My goal is to get her very curious
about this thing that I'm telling her
131
:about, and then ultimately get, you
know, be able to pitch this to her.
132
:So by giving it that much context,
sometimes AI does that creepy
133
:thing where it references the
email, but it references too much.
134
:Don't be a creep and be like,
"Oh, that was so funny about
135
:your nephew," blah, blah, blah.
136
:I would never say that in real life.
137
:So I'm just like, don't be a creep.
138
:Do this.
139
:Make it concise.
140
:A lot of times sometimes when I'm
giving CEOs things, Claude likes
141
:to be extra and give you 14 pages.
142
:Well, that's the next step.
143
:Never mind.
144
:We'll, I'll save that for the next one.
145
:Rachel: So now that we've kind of gone
through the first two, so it knows who
146
:it is and it knows what it's trying to
achieve through the objective, now we have
147
:to make sure that it gives you something
that is actually of value and something
148
:you can use, which brings us to the
second O, which is step three, output.
149
:Jessica: Yeah, so Claude likes
to be a little bit extra.
150
:So if you do not tell it what you
want to get, you're gonna get a
151
:giant 14-page essay back every time
when all you wanted was a list.
152
:So you need to make sure you're specifying
the output, the format you want it in,
153
:because it likes to default to documents
and maybe you just want it typed back in.
154
:Do you want it Notion?
155
:The constraints, how many
words do you want it to be?
156
:Do you want it to be long?
157
:How do you want it structured?
158
:How many options and what to leave out?
159
:So if you say, "Give me three subject
lines under eight words each," or, "Write
160
:four short paragraphs, no bullet points,"
161
:Rachel: yeah
162
:Jessica: Specifics here are really
gonna save you a lot of time from going
163
:back and editing, and it's also gonna
save you on usage because you're not
164
:getting more than you actually need back.
165
:Rachel: Yeah.
166
:And this is something that,
everybody, regardless of your
167
:industry, has best practices, right?
168
:For an email marketing strategist, I
have, you know, how many characters I
169
:wanna hit in my subject line, I have,
you know, the subject lines that I
170
:know work really well with my clients,
and all of these various things.
171
:But Claude is not a subject or email
marketing subject line expert, and
172
:sometimes they can go two 200+ characters,
and you're like, "Man, nobody would ever
173
:read that," all of these various things.
174
:So when you tell it to keep it under
75 characters, start it with how to,
175
:a list number, anything like that.
176
:So many people now tell AI,
"Get rid of the em dashes."
177
:Em dashes are telling it.
178
:Jessica: No emojis, no hashtags.
179
:Rachel: Yeah, all of those things,
and it's really important to spell
180
:it out exactly how you want it.
181
:Tell it exactly the length of things.
182
:So Jess and I, we were just, prior to this
call, auditing and optimizing our blog
183
:writer, Bruce, and we were saying, "Hey,
it needs to have these components to it.
184
:It needs to never do this, and make
sure that these are the proper rules
185
:that it follows every single time."
186
:But what's gonna be great about this
is that once we've set this, next time
187
:we go to write a blog, we don't have
to reformat it and restructure it in a
188
:way that we know is gonna be optimized
for our desired output, because we have
189
:locked in essentially these end results,
the output of it all that we wanted.
190
:So it gets us to a point of less
editing time and time again.
191
:Jessica: And like I
said, AI is an employee.
192
:You gotta give it feedback.
193
:He's gonna get better every time.
194
:The key is that you're saving
your prompts and reusing them
195
:so you can optimize them more.
196
:For example, every time we go to write
this podcast, he kept opening the
197
:entire podcast with real talk, and
that is one of the newly banned words.
198
:Do you have any specific bans on your AI?
199
:Rachel: The 'quietly', God,
quietly is so obnoxious.
200
:Once you see it pop up time and time again
in your outputs, and you start to see it
201
:consuming the content of other people and
you're like, "I would never say that."
202
:I was writing an email the other day, and
I actually did use quietly, but not in the
203
:way that AI uses it, and I was like, "I
just am so jaded now, I can't say that."
204
:So I had to find a different word to use.
205
:But it's important to know these things
or else you're gonna have to keep auditing
206
:and refining every output you get.
207
:But it's gonna learn if you tell
it, "Don't do this, do this instead,
208
:and this is the output that I want."
209
:Jessica: All right, so
moving on to our next thing.
210
:This is gonna be the one thing that is
really gonna help turn generic AI slop
211
:into something that actually sounds human,
actually sounds like you specifically,
212
:and that is your tone, the T.
213
:Rachel: Yeah.
214
:So want to hand it your voice, right?
215
:And this is something that we put at
the forefront of our foundation in
216
:the Life First Business Lab, where
you're building your business brain,
217
:where you get your copy Bible, you
have your voice of the customer, all
218
:of these things that are constantly
being referenced in your prompts.
219
:You want it to know how should it sound.
220
:Are you a warm, direct, a
little bit funny type of person?
221
:Or are you more straight to the
point, big sister type energy?
222
:All of those various things that make you
more you in your voice and all of that.
223
:I could probably pick out an Ashley
email from a mile away because I know
224
:how she talks, and the things that
are important for your audience, is
225
:that they want you to be able to say,
yeah, that's the person that I follow.
226
:That's their language.
227
:That's how they speak.
228
:That's how they communicate."
229
:And that's a reason why there's so many
times where people go, "AI doesn't work
230
:for because it doesn't sound like me.
231
:There's so much refinement
that I have to do on my end."
232
:It's this T step, the tone step,
is the most important part.
233
:Jessica: Yeah.
234
:Actually when we were training the
Claude that writes our podcast and
235
:really helps us with all the things in
the lab, I just took transcripts from
236
:our actual Speaking Voice podcasts
and kept feeding it into Claude.
237
:Actually, now he goes and automatically
pulls our transcripts and puts it in
238
:there so that he can constantly be
analyzing our voice, pulling out our
239
:frameworks that we like to use all
the time so that it's learning me.
240
:So yeah.
241
:Pro tip is paste in.
242
:If you don't wanna sit there and explain
it or you don't know what you, how to
243
:describe what your voice sounds like,
is go find some things that you really
244
:love that you've either written or
said or do and feed it to it and say,
245
:" Analyze this and match this voice."
246
:Rachel: Yeah, exactly.
247
:So then that brings us to the last
letter, which is S, and I think this is
248
:the one that has really helped me utilize
prompts and our AI tools and stuff.
249
:And it helped me kind of go
from yeah, this is a good output
250
:to holy cow, did it do that?
251
:How did it get me to the point
where now, I can take this, give
252
:it my human touch, refine it just
barely, and then take it and go?
253
:And that is step five, the specific.
254
:So this is the context about your
business and really takes you from
255
:where it knows the inner deep workings
of your business, why you have specific
256
:offers, and it makes things way less
generic than it could possibly be.
257
:So I think this is the one that I
would say is the most important.
258
:Jessica: I say this all the
time, the more context you can
259
:give it, the better it can be.
260
:For example we just launched
the Life First Business Lab.
261
:And we were able to launch
this in record time.
262
:I think from the moment we had the idea
to launch was maybe a week and a half.
263
:And this was setting up the product,
getting the onboarding emails, writing
264
:our sales page, designing our sales
page, and it's because I had already
265
:done so much pre-work with training our
Claude on everything about our business.
266
:So it knew everything about our offer.
267
:We had had it do some voice-of-the-client
research, so it knew everything about
268
:the audience and their exact pain points.
269
:It knew everything about the results that
the lab was gonna get them, and everything
270
:that also made Rachel and I different,
what is our specific history, why we would
271
:build this or why it's important for us.
272
:And we were able to kinda
churn that out so quickly.
273
:It sounds just like us.
274
:We only had to go back and do a couple
of tweaks to get it there, and that's
275
:really what the difference is about having
it just going, opening Chat randomly
276
:and say, "Write about, you know, gut
health issues," if you're a dietician.
277
:Or somebody who has taken the time to
create a copy bible in the business brain,
278
:and going in there and being like, "Okay,
let's write this email to my client.
279
:This is the offer I'm trying to sell."
280
:It knows everything about it and kinda
spit out everything that it needs.
281
:Rachel: Yeah.
282
:So, I always say this when it comes to
your email list, but the same is true
283
:when it comes to your AI prompts and all
of that, where it's trash in, trash out.
284
:So if you are putting generic information
in and you are getting generic information
285
:out, the specifics of it all is really
where the game changer is, right?
286
:Where the pivotal moment of saying,
"Okay, this is something that maybe
287
:I can be inspired from," and then
rewrite everything from scratch, to then
288
:actually giving it the specifics and
saying, "This is exactly what I needed.
289
:I just need to refine it 10%
15%," whatever the case may be.
290
:So that's all five.
291
:I'm gonna do a quick recap, and then
we're gonna shift into one more area.
292
:So in the roots format, we have
the role, so who we want it to be.
293
:Before we give it any information, we
tell it exactly who we want it to be.
294
:Then we go to the objective.
295
:Now that we know who it is, we
wanna tell it what it's actually
296
:trying to then accomplish.
297
:What's the goal?
298
:Then we go into the output.
299
:We want to tell it exactly what we are
looking for when we get that prompt,
300
:whether it is saying, "Please do not
ever give me another Google document.
301
:Give me a direct text output
and make it look like this."
302
:That's what we wanna
really communicate to it.
303
:Then we send it to the most important
information, or one of the most important
304
:pieces, which is the tone, the portion
that takes it from sounding every person
305
:on the internet that's ever used a random
prompt to actually sounding like you,
306
:how it should sound warm or a little big
sister-y, all of those various things.
307
:we root it into the actual specifics
in the context of your business, your
308
:offers, your pricing, your audience,
exact pain points, all of that.
309
:And that gives us the roots framework.
310
:Jessica: So the mindset shift we want
you to walk away with is that you're
311
:not bad at AI, AI doesn't suck, AI isn't
garbage, you're just under-training it.
312
:So think about it.
313
:You would never onboard a new hire and
expect them to create a task and output
314
:it perfectly, especially if you gave it
zero context or no background or no voice.
315
:So, the reframe here is AI is
not a software you operate, it
316
:is a team member that you brief.
317
:It's the same rules that would apply
here that would apply to a great manager.
318
:You wanna make sure that you're
giving it everything that it
319
:needs to set it up for success.
320
:And I know what you're thinking:
"Jess, that sounds like it's gonna
321
:take a lot of effing time, and I
don't have effing time to write a
322
:paragraph prompt every single time."
323
:And that's why you wanna make sure
that you are saving your prompts.
324
:We also have a cool little tool inside
the lab called the Workflow Wizard,
325
:which helps you write the prompts like
this, following the ROOTS framework.
326
:So you can basically go in there and give
it a very vague of what you wanna do, your
327
:short little, fast prompt, and it will
help you fill in all the gaps for what
328
:you're missing so that you can then go
into your Claude and not have to waste,
329
:a lot of time and credits going back and
forth trying to get it where you need.
330
:Rachel: All right, so a couple of
things before we wrap up this episode.
331
:First, we want you to share this one.
332
:We all know the business bestie
that we all have chatted with,
333
:and they're like, "I can't use AI.
334
:It doesn't capture my voice."
335
:We want you to share this episode
with her or him, and kind of share
336
:the fix that maybe she's been missing.
337
:Maybe this is the one shift that
has been missing that can help that
338
:person actually use AI to a more
achievable and less frustrating manner.
339
:And then second, we want you to tell
us that you are using AI, right?
340
:DM us at Deeply Rooted
Business on Instagram.
341
:Tell us which letter of the
ROOTS framework that you skip the
342
:most, because I think it's really
likely gonna be tone or specifics.
343
:And we read all of our messages.
344
:You can send us an email too.
345
:We love emails as well.
346
:Jessica: And if you want a cheat code
for having AI do things and really become
347
:a whole assistant for you, we want to
invite you to the Life First Business Lab.
348
:This is where you're gonna get a
pre-trained, already on the roots AI
349
:assistant that you can install in Claude.
350
:Once you install it, it does a quick
little onboarding with you so it can
351
:get all the context and tone about your
specific business, so that you can go in
352
:and write your blog posts quicker, emails.
353
:We are creating and training
new employees every single week.
354
:And we know that this is such a good
product for busy small business owners
355
:like ourselves, 'cause you don't
have time to be training a bunch
356
:of different assistants because you
would've hired one already if you did.
357
:Rachel: These are all assistants
that we use in our own business,
358
:and that we've created because
they're not fluff assistants, right?
359
:Sales call closer, Sana, has
been super impactful and really
360
:beneficial in my business.
361
:We use Bruce every single week,
every other day kind of thing.
362
:So check it out.
363
:The details are in the show notes.
364
:It's an incredible product that
we're both very, very, very proud of.
365
:So,
366
:Jessica: and until next
week, we're rooting for you.