0:00

On this episode, I'm going to

0:02

show you how to create two

0:04

AI assistants to make you the

0:06

most productive person in your company.

0:08

I'm going to show you how

0:10

to build a executive assistant and

0:12

a project assistant, and better yet,

0:14

I'm going to show you how

0:16

to do that across Gemini gems,

0:18

chat TVT, and claw projects, and

0:20

I'm going to compare and contrast

0:22

which are better, which one should

0:24

you actually use, for what? All

0:26

of that I'm born today's show

0:28

I'm Karen Flanagan the co-host of

0:30

marketing Against the Green. Let's go

0:32

build some AI assistance. Here's a

0:34

quick message from Hubspot. This

0:37

isn't your typical marketing

0:39

software at because Hubspot

0:41

isn't typical. It's marketing made

0:43

easy. Turn one piece of content

0:45

into assets for every channel. Convert

0:48

leads in no time and get

0:50

a crystal clear view of your

0:52

campaign performance. Hubspot can do all

0:54

that and get your results fast,

0:57

like double your leads in 12 months

0:59

fast. See, I told you

1:01

this wasn't a typical software

1:04

ad. Visit hubspot.com/marketers to get

1:06

started for free. So

1:20

this is a really great video. If you

1:22

just want to get started with the

1:25

AI, like I think this is one

1:27

of the easiest things you're going to

1:29

be able to replicate for your day-to-day

1:32

work because it's not my day-to-day work

1:34

as a marketer, it's not my

1:36

day-to-day work as a special whatever I

1:38

do, it's really just how we do

1:41

work. And I'm going to show you how

1:43

I actually started to use AI to be

1:45

way more productive. So people who follow

1:47

along my videos may have seen me

1:49

build an executive assistant for

1:52

Claude. Well, I'm always building upon and

1:54

iterating upon what I've done in the

1:56

past, and I've made that really much

1:58

better. I've also replicated it for the

2:01

three core platforms, Google Gemini, Open AI,

2:03

Chat, GBT, and Claude. And I'm going

2:05

to show you a comparison, because I

2:07

think it's pretty interesting in comparison, like

2:10

which ones are better for your day-to-day

2:12

work as a knowledge worker. The first

2:14

thing I actually want to start with

2:16

is something you probably haven't even thought

2:19

about for AI. But you should. Here's

2:21

a little sneak peek at something I

2:23

do that I think you should start doing,

2:25

and no one else would tell you this tip.

2:27

If you were using Google Drive or whatever

2:30

folder you were using, I'm going to show

2:32

you how to use your folder structure to

2:34

build these AI assistants. I am someone who

2:36

is pretty disorganized in terms of where I

2:38

keep things. And because of AI, that actually

2:40

no longer matters. It doesn't really matter where

2:43

I keep things. So there's two things you

2:45

have to do when you're building full structures

2:47

and the way I would do this. So

2:49

for me, in HubSPI have three large pods

2:51

that I oversee. I have three large pods

2:53

that I oversee. in the lower end of

2:56

our business, we have our overall demand, just

2:58

how hotspots go to market works, and then

3:00

we have all of our AI work. And so

3:02

what I do is I have a folder

3:04

structure that basically just has two different folders

3:07

of stuff. I have one where I put

3:09

every single meeting transcript, record everything, put

3:11

the meeting transcripts in there, and then

3:14

all of my strategic docs. And what

3:16

I'm going to show you is the

3:18

reason I do that, as I build

3:20

an AI executive assistant. for the meeting

3:23

transcripts and I build an AI project

3:25

assistant for all of the strategic docs.

3:27

That's kind of tip number one. I

3:29

always start to think about how your

3:31

folder structure allows you to build these

3:33

kind of AI assistance. And I think

3:35

to follow along my video, this is

3:37

my recommended structure. One, whatever your project,

3:39

break it into meeting transcripts, which is

3:41

all of the meetings you're having on

3:43

Zoom or whatever else and you're keeping

3:45

the transcripts, and then the street teacher

3:47

docs. Now at some point, I will

3:49

do a follow on to this video

3:51

where this is not just zoom meeting

3:53

transcripts, it's going to include slacks. But for

3:56

the time being, I think that is where I

3:58

would start. The other cool thing is... a lot

4:00

of people what you'll do is go to

4:02

zoom I'm going to use zoom because it's

4:04

the most common but whatever you're using you

4:06

go you record the meeting you get the

4:08

transcript and you have to put it in

4:10

the folder or what I'm doing is I

4:13

have set up zaps so every time the

4:15

meaning ends and the transcript is available it

4:17

just zaps it into the right folder so

4:19

that's number one okay let's go into build

4:21

this in Gemini because we haven't actually built

4:23

gems I focused a lot on Claude I'm

4:26

still going to do claw in this But

4:28

we haven't actually talked much about Gemini

4:30

gems. Gyms are basically Google's equivalent to

4:32

custom GPTs. So when you go into

4:35

Gemini, you will see the gem manager

4:37

here. So basically you can see all

4:39

your gems. So they have pre-made for

4:41

Google. Very similar to a chat TPT

4:43

store. I built two. I built our

4:46

project assistant and I built a scaled

4:48

assailant. So these are for one of

4:50

my pods. And so I go into

4:52

the project assistant. And so one of

4:55

the things I can do now is

4:57

showing me. the last summary of our

4:59

meeting, including key follow-up points.

5:01

Taking all the time to load, analyzing,

5:04

response, and it will go through and

5:06

it will show me my last meeting and

5:08

tells me everything I need to know.

5:10

And so let me go through how

5:12

we've actually built this. We're going to

5:14

jam manager, we're going to edit

5:16

executive assistant. And so what you

5:19

do is you build an executive

5:21

assistant, and this is the prompt here. You

5:23

basically tell it to... Do meat analysis. So

5:25

anytime you give it a meat and transcript,

5:27

it analyze the meat and transcript, it identifies

5:29

key decisions and action items, it tracks deadlines

5:31

and ownership, it flags high priority items. And

5:34

then I give it the exact output format.

5:36

Now if you want these prompts, you sign

5:38

up to my newsletter, I'm going to give

5:40

you this prompt and the project assistant prompt.

5:42

You can find my newsletter, sign up in

5:44

my LinkedIn. And so we'll go back to

5:46

jams. And so I put the prompt, and

5:48

then the next thing you can do is

5:50

you can basically upload everything from your Google

5:52

Drive. And so if you have a folder structure,

5:55

so again, what did I say? All of my

5:57

folder structure looks probably the same, and then you

5:59

can go. mean transcripts and grab all

6:01

the mean transcripts. Now word to all

6:03

of the people, building these AI assistants,

6:06

Google, chat, Chibati, and Claude, all do

6:08

a bad job of this in that you have

6:10

to kind of go in and upload the files

6:12

or you connect your Google drives, you go add

6:14

them, and add to go in and upload the

6:16

files or you connect your Google drives, you go

6:18

add them. I have to go in here and

6:20

add the new meaning transcript in here to get

6:23

all of the follow-up actions. What I want to

6:25

do in the future is being able to just

6:27

zap it straight, Anything that's happened in any of

6:29

your meetings is access to all of your meetings,

6:31

just remember to go in here, add them from

6:33

your Google Drive, I've already done that. and go

6:35

in at any meeting transcript, add the latest one.

6:37

So that's the only piece of work

6:40

you have to do is I go

6:42

remember to do meeting transcript. Now the

6:44

next one is super cool, the project

6:46

assistant. So this is a strategic project

6:48

assistant and it does some pretty cool

6:50

things. So it basically says you are

6:52

an AI project assistant with expertise in

6:54

project management risk analysis and organizational dynamics.

6:56

Now remember when you get this prompt.

6:58

If you follow along with my newsletter

7:00

and you really want to kind of

7:03

replicate this for yourself, you can actually

7:05

play around with this prompt, right? You

7:07

should just not copy and paste for

7:09

your needs. You should use it as

7:11

a foundation and then build upon that

7:13

foundation and tweet to your needs. Again,

7:16

the meeting exact one has access to

7:18

all of your meeting transcripts. So

7:20

remember the folder structure. Now your

7:22

project assistant has access to all

7:25

of your strategic. And what it will

7:27

do is it will analyze all the documentation

7:29

and basically give you a summary of their

7:31

project, including some really cool things, like team

7:33

structure responsibilities, any timelines, all of the budget,

7:35

risk assessment, technical specifications, stakeholder communication logs, previous

7:37

status reports, previous status reports, previous status reports,

7:40

and minute meetings. And this one here I'm

7:42

trying to think through. Someone may argue why

7:44

don't you just combine these things in two.

7:46

It's actually better to keep them separate. They

7:48

work much better if you have one very

7:50

focused on project on project management. And so

7:52

this is doing a little bit of like executive

7:54

assistant where it is grabbing previous minute meetings, but

7:57

I'm probably going to take this out and keep

7:59

it very central. or project management work. Now,

8:01

and then it provides actually a menu from

8:03

you to pick from. So you can basically

8:05

ask it to do a further project status

8:07

analysis, not sure what that looks like. You

8:09

can do a risk assessment, which is really

8:12

good at like pinpoints any potential risks. It

8:14

does a really good job. Well, some better

8:16

than other, I'll actually show you the three

8:18

perform very differently when I give it

8:20

this prompt. tactic generation. So we'll actually

8:22

pull out the core components of your

8:24

project and recommend different tactics to make

8:27

that component a success. Strategic recommendations, if

8:29

you're having to put something together for

8:31

your executive team, you can choose us.

8:33

It does a really good job. And

8:35

then any kind of document generation. So

8:37

I can basically create a monthly summary,

8:39

a quarterly summary, a quick summary for

8:41

the executive team, for our CEO. really

8:43

easily because it has all

8:45

the documentation. The minute meet-ins

8:47

compilation, again I'll take this

8:50

one because I've actually since

8:52

progressed to split these things out,

8:54

I had it all in one, and

8:56

then action, I'm tracking decision log maintenance,

8:58

it gives you the kind of history

9:01

of how it, it gives you the

9:03

kind of history of how it decision

9:05

log maintenance, it gives you the kind

9:07

of history of how it decision was

9:10

made, which I think is really important,

9:12

like let's create a one-page summary. for

9:14

my executive team detail and key accomplishments

9:16

and challenges. All right, gives key accomplishments,

9:19

give some key challenges, looking back over

9:21

all of the presentations and docs that

9:23

we've done, overall status, like pretty good

9:25

length. So it actually does a really

9:28

good job of being your project assistant

9:30

assistant. Now for any project, I have

9:32

a meeting assistant for that folder, and

9:34

then I have a project assistant for

9:36

the strategic docs folder. And again, just

9:39

going back to that project assistant assistant.

9:41

Really, what we want to do is

9:43

have it summarize all of those strategic

9:45

docs anytime you want. Now, when you're

9:47

summarizing the strategic docs, you can ask it

9:50

for specific parts of the docs. Like, show me

9:52

what is different from this month to last month

9:54

in terms of the progress we've made. Show me

9:56

that summary. Then you can get into the actual,

9:58

what are the options? you want. This is

10:01

the one that actually is really good because

10:03

I was trying this where it actually shows

10:05

you key components of a project and will

10:07

show you tactics. But then I will do

10:10

anything for you. Like we'll do all the

10:12

document generations, we'll do strategic recommendations, it will

10:14

do all of that stuff. Now that's the

10:16

way I'm sending it up. The reason I've

10:18

kind of glossed over some of the output

10:21

because this is actually my real assistant, right?

10:23

So it actually has access to a lot

10:25

of real documentation. Again, the documentation, you have

10:27

to go in to Gemini. What I really

10:30

like by Gemini is the context window is

10:32

really large. So you just have to go

10:34

in and remember to continue to like add

10:36

the latest stocks. There's no way yet to

10:38

like automatically add the docs into that context

10:41

window. Now, if you're watching the long LinkedIn,

10:43

this is where I will leave you because

10:45

the LinkedIn video can't be too long. If

10:47

you are subscribed to my newsletter. If you

10:50

are a marketing against a great podcast, listener,

10:52

you're going to get the full episode where

10:54

now I'm going to go on and show

10:56

you how to do this in Claude, how

10:58

to do it in open AI.

11:00

One of those is actually

11:03

even better than Gemini. So

11:05

Gemini is really good. Very,

11:07

very good actually. But

11:10

one of the ones I'm

11:12

going to show you is

11:14

surprisingly amazing. brought to you

11:16

by the Hubspot Podcast Network.

11:18

Creators are brands, explores how

11:20

storytellers are building brands online,

11:23

from the mindsets, to the

11:25

tactics, to the business side. They break

11:27

down what's working, so you can apply

11:29

that to your own goals. Tom just did

11:31

a great episode about social media

11:33

growth called 3K to 45K on

11:36

Instagram in one year, selling digital

11:38

products and quitting his job to

11:40

go full-time creator with Gannon Mayer.

11:42

Listen to creators or brands

11:45

wherever you get your podcast.

11:47

All right, so how does this look

11:49

in chatGBT and Claude? And so if

11:51

we go into Open AI and

11:53

we create some custom GPTs, I

11:55

would say that again, the get in

11:58

the documentation is not five. So what

12:00

I've done is created custom GBTs. I've given

12:02

it some prompts. I can just click and

12:04

say analyze the transcript for key decisions. I

12:07

can do what are the high priority items.

12:09

Can you provide a strongly follow-up, identify the

12:11

action items with ownership, which is really

12:13

great. Exact same prompt. There is no difference in

12:15

the prompt here. The only thing I will say is,

12:17

again, the best way to use this is after a

12:19

median. Again, you can add it to your G drive.

12:22

And then I can go in and I can go

12:24

in and I can go in and I can grab

12:26

my median transcripts. Let's go on, you just have to

12:28

go in and you add your latest meeting transcript. And

12:30

then you add that doc. So again, it will be

12:32

much better at some point if it just has access

12:34

to a folder where those folders are keeping your documentation

12:37

so you don't have to go in and just add

12:39

it. And anytime you want a meeting summary, you can

12:41

just go in and ask for any kind of meeting

12:43

summary. Okay, so this is the scale selling

12:45

executive system. So it works very same the

12:48

way. So it's a custom GBT. So I

12:50

go into explored GBTs. I go to my

12:52

GBTs. I have the exact same two assistants,

12:54

the same prompts. So if I go into

12:56

the executive assistant, I can

12:58

upload a meeting transcript here. Again,

13:01

I've got it connected to my G

13:03

drive. And then I can upload this. And I

13:05

can say analyze the meeting transcript for

13:07

key decisions and action items. So anytime I

13:09

need anything around a meeting around a meeting.

13:11

I've got it recorded as in the folder.

13:14

I can go in, I can upload it

13:16

here. So it does an incredible

13:18

job of getting you the meeting summary. I

13:20

think better the most meeting tools actually. So

13:22

I really never need to go and

13:24

figure out what is happening because I record

13:27

all meetings, everything is recorded from like one-to-one

13:29

to team units. So I have this like

13:31

database of unstructured data that I can go

13:33

and get insights from any time I want.

13:36

The one that's much better here in open

13:38

AI chat TVT than I think Google Gemini

13:40

is actually the project assistant. Now again, what

13:42

I really get frustrated on is just the

13:45

way you have to add the docs. So

13:47

let's say I wanted to use my assistant

13:49

here to get some worked on my project

13:51

and I'm going to add all of these

13:54

documents. So add all the latest strategic docs

13:56

again, not to continue to repeat myself.

13:58

Be so much better. if it could

14:00

just auto-axis a folder, so I didn't

14:03

have to add the docs. But then

14:05

I'm going to say provide a gap

14:07

analysis of the project plan. So

14:09

based upon our current course, where do

14:12

you think are gaps in challenges?

14:14

Takes a little bit of time. And

14:16

so it will do a pretty great

14:18

job. And so if you actually

14:20

do that, it will give you

14:22

like an overview strategy and goals.

14:25

It'll give you lots of the

14:27

strategic gaps, operational gaps, operational gaps.

14:29

It's pretty unbelievable to be honest with you

14:32

so like you really do have a project

14:34

assistant that can be strategic with

14:36

you and actually go back and forth and

14:38

so the other one that works really well

14:40

is when you ask it for like show

14:42

me tactics around components that you think are

14:44

the most important to make this thing a success.

14:47

So I would say chat cheapity is not

14:49

much better but is better in terms of

14:51

the project assistant if you build that versus

14:53

what I've seen on Gemini gems. both very

14:55

useful and you can prod and poke

14:57

Gemini to be better if I ask

14:59

it for more comprehensive details it will

15:01

actually do that. But that's how you

15:03

use the project assistant you build custom

15:06

GPTs and again when I finished I'll

15:08

have three sets of those a set

15:10

for each folder a meeting executive and

15:12

a project assistant. And then the last

15:14

one I'll just quickly cover is Claude,

15:16

because I want to talk a little

15:18

bit about some of the actual drawbacks

15:20

of Claude for some knowledge work like

15:23

this. So I would say at the

15:25

moment, like chat TVT number one, mostly

15:27

just because of its output in terms

15:29

of this project assistant. It just does

15:31

better strategic thinking around that for some

15:33

reason. The two I have here, I'll go

15:35

into project assistant. Claude's core drawback here is

15:38

I have them all connected to my G

15:40

drive because I use them all at work,

15:42

but it doesn't have the ability to pick

15:44

up power points, but like Google slides

15:46

just has text in this. I'm doing

15:48

something wrong, but it would not find

15:50

any of the power points. And then

15:52

also the context window is just too small.

15:55

So when you're building them in Claude

15:57

projects, you go in and you actually

15:59

add in your... project instructions,

16:01

right? And it's the exact same prompt,

16:03

and then you add your content. And again,

16:05

I don't know why they've done it this

16:07

way. If you see the other two, I

16:10

can just open up Google Drive and go

16:12

pick the files. Here you can't even do

16:14

that, right? It doesn't allow me to just

16:16

open Google Drive to go and actually pick

16:18

the files. But let's say I wanted to

16:20

give it some strategic docs. Like I can

16:22

add them in here. Yeah, you have to add them

16:24

one by one by one by one. So I

16:27

think the context window is smaller. And

16:29

being able to add the documents is

16:31

not as intuitive, and I cannot actually

16:33

add in power points, but it actually

16:35

gives you pretty great recommendations. I would say

16:38

the recommendations it gives you for these two

16:40

prompts, specifically for the project assistant, is actually

16:42

a little bit better than Gemini, or maybe

16:44

on a par with Gemini, which has been

16:46

a little bit better. So that is the

16:48

video. The main takeaways here is actually not

16:50

the output because the output is really good,

16:52

it's how I did it, right? I structured

16:54

the folders, meeting transcript, strategic docs, and then

16:56

I built an assistant for EEA, built an

16:59

assistant for project assistant, and then I built

17:01

that menu of options that you can actually

17:03

use it to get a summary, you can

17:05

use it to do further things on the

17:07

project for you. And that's a really

17:09

great way to start to actually integrate

17:11

these into your day-to-day work and actually

17:13

makes a big big impact. I use

17:15

them every single day. I kind of

17:18

like gravitate between chat GT and Gemini

17:20

because Gemini is so deeply hooked into

17:22

the Google suite of products. It's really

17:24

useful. Claude, I use it every day.

17:26

I use it actually to take some

17:28

of the documentation that the assistance billed

17:30

for me around my projects and I

17:32

ask Claude to rewrite to rewrite them

17:34

because it's a better writer. on the

17:37

one-pager it built for the executive team, and then

17:39

I'll put it into Claude and I'll say, hey,

17:41

can you rewrite this for my executive team? This

17:43

is the audience. This is what I'm trying to

17:45

do. So I used all three in some sort

17:47

of combination, but if I was going to end

17:49

this in terms of a summary, all of them

17:51

are pretty good to create a first pass

17:53

at a meeting executive. They all do a

17:55

really good job, because all of the meeting

17:57

transcripts are text or text files. Project

18:00

depending upon what documentation you use and

18:02

what you want to put in there,

18:04

I would say Google Gemini gems is

18:07

really good. It has access to all

18:09

types of documentation, really large context

18:11

window, easily hooks into G-Drive, can just

18:13

add them in there, chat chibeti,

18:15

really good. The output does a slightly

18:18

better job of like strategic analysis, strategic

18:20

recommendations, and then Claude really struggling because

18:22

you build projects, you have the

18:24

context window, context window a little too

18:27

small, and actually harder to like

18:29

upload the doc. documentation you want. So

18:31

that is today's video. I hope this is

18:33

a good way for you to think about

18:35

how to start to integrate AI into your

18:37

own work and until next time, enjoy this.