AI Assistant Does Your Job! Build AI Bots with Zapier Central & OpenAI's Assistant API

Summary

今回ご紹介するのは、ZapierのCentralとOpenAIを利用して、効果的な自動化とビジネスの労働を支援する新しいアシスタントを作成する方法です。GoogleドキュメントやMicrosoft Wordをデータソースとして使用し、ビジネスプロファイルを作成。このプロファイルを使って、メール自動返信などのタスクをスムーズに行うことができます。また、この技術を応用して、ビジネスの自動化プロセスを簡素化し、時間と労力を節約する方法も学ぶことができます。

Takeaways

• 🤖 利用ZapierとOpenAIのAPIを作成した新しいアシスタントを学ぶことができます。
• 📄 データセットとしてビジネスプロファイルを作成し、より効果的な回答を得ることができます。
• 📑 GoogleドキュメントやMicrosoftドキュメントを使ってアシスタントのコンテキストを提供します。
• 📈 ビジネスのルールやFAQを文書化し、自動返信の際に参照できるようにします。
• 🔗 ウェブサイトや連絡先情報を提供することで、アシスタントがより正確な対応ができるようにします。
• 📌 重要なビジネス情報はPDF形式でエクスポートして、APIで使用できるようにします。
• 🛠️ OpenAIのAPIを使用して、ビジネスプロファイルのPDFを参照してアシスタントを設定します。
• 📩 メール返信の際に、特定のトリガーワードに基づいてGmailの検索を実行します。
• 💬 ChatGBTブロックを使ってアシスタントと会話形式でメッセージを生成します。
• 📄 メールのドラフトを作成し、自動返信システムをZapierのアクションとして設定します。
• 🔄 自動化されたプロセスにより、大量のメールに対応する効率的な方法が得られます。

Q & A

今日のビデオで作成する新しいアシスタントはどのようなものか？

新しいアシスタントは、W2や従業員保険を必要としないZapierのCentralを利用して作成されます。このアシスタントはビジネスとその関連する労働力を理解するためのデータセットを作成し、効果的な応答を提供するためのものです。

AIプロバイダーとして使用されるOpenAIの何の特徴を利用する予定ですか？

OpenAIのアシスタント機能を利用して、ビジネスプロファイルのためのデータセットを作成し、より効果的で正確な回答を提供することができるようにします。

GoogleドキュメントやMicrosoft Wordを使用してアシスタントをトレーニングするために提供すべき情報は何ですか？

会社の名前、場所、設立者、CEO、業界、使命宣言、提供されるサービスなどの関連情報を提供します。これにより、AIはより効果的な出力を提供することができます。

Eメールの応答を自動化するために設定するZapierの行動はどのようなものか？

Zapierで行動を設定し、Gmailの受信Eメールに対する応答を自動化します。このプロセスには、Eメールの検索、Chat GPTの会話を生成、そしてドラフトの作成や送信が含まれます。

AIが生成したEメールドラフトを確認するために、どのようなステップを実行しましたか？

ビデオのデモンストレーションでは、AIが生成したEメールドラフトが作成され、その内容が確認されました。これには、太阳能パネルの見積もりに関する初期の問い合わせに対する回答が含まれています。

このビデオで学べることの1つとして、どういったAIモデルの出力をより効果的にするためのヒントは何ですか？

AIモデルの出力をより効果的にするためのヒントの1つは、カスタムインストラクションを使用することです。これにより、特定の出力を求める際に必要な文脈を理解し、より効果的な出力を提供することができるようになります。

ZapierとOpenAIを組み合わせて何を可能にすることができるか？

ZapierとOpenAIを組み合わせることで、自動化されたタスクやビジネスプロセスをスムーズに実行することができます。これにより、Eメールの応答やデータの入力/出力、そしてその他のビジネス関連の自動化が可能になります。

このビデオで作成した「新しいアシスタント」は無料で使用できる吗？

はい、ビデオで説明されているように、現在は無料でこの「新しいアシスタント」を使用することができますが、将来的には料金が発生する可能性があると述べられています。

AIが生成したEメールドラフトを改善するために追加で必要な情報は何ですか？

AIが生成したEメールドラフトを改善するためには、おそらくアシスタントAPIにさらに具体的な情報（例えば、カスタマーサービスの電話番号やウェブサイト）を追加する必要があります。

このビデオの目的は何ですか？

このビデオの目的は、ZapierのCentralとOpenAIを利用して、ビジネスプロセスの自動化と効率化を図ることです。具体的には、Eメールの自動応答やデータの処理を通じて、ビジネスの効率を高める方法を学ぶことができます。

このビデオで紹介された技術を実際に実践するためにはどのようなスキルが必要ですか？

このビデオで紹介された技術を実践するためには、AIモデルとの対話やZapierの使い方に関する基本的な知識が必要です。また、ビジネスプロセスを理解し、それに基づいて自動化を適用する能力も重要です。

Outlines

00:00

🤖 利用ZapierとOpenAIで新しいアシスタントを作成

このセクションでは、動画の視聴者にZapierとOpenAIを利用して新しいアシスタントを作成する方法を紹介しています。このアシスタントは、従業員として扱う必要がなく、W2や従業員保険を必要としません。データ源の追加や行動の設定を通じて、ビジネスにおけるこの技術の活用方法を理解し、効果的な回答を得るために、OpenAIのアシスタント機能を使用します。また、Eメールへの自動返信を通じてこの技術の力を示すために、Googleドキュメントを作成し、ビジネスプロファイルを設定する方法も学びます。

05:02

📄 データ源の作成とZapierへのアップロード

この段落では、GoogleドキュメントまたはMicrosoft Wordドキュメントを使用してデータ源を作成し、Zapierにアップロードする方法について説明しています。データ源は、アシスタントが効果的な出力を提供するために必要なビジネスの文脈を提供します。これには、会社名、場所、設立者、CEO、産業、サービス提供などが含まれます。また、この文書のフォーマットはAIに読み取られるデータであるため、重要ではありません。

10:04

📧 Eメール自動返信の設定とテスト

この部分では、Zapierを利用してEメールの自動返信を設定し、テストする方法を学びます。具体的には、Gmailのアクションを使用して、特定の問い合わせEメールを見つけ、OpenAIのアシスタントを通じて返信を生成し、ドラフトを作成します。このプロセスは、通常の自動化タスクよりも簡単で、効果的な出力を得ることができます。また、この方法が機能することを確認し、ビジネスにおける実際の適用方法についても議論します。

15:05

🚀 ビジネスにおけるAIと自動化の活用

最後の段落では、ビジネスにおいてAIと自動化をどのように活用し、新しいアシスタントを効果的に使用することができるかに焦点を当てています。ZapierとOpenAIを使ったこの方法は、ビジネスプロセスの自動化に大きな影響を与える可能性があり、将来的にはさらに進化するでしょう。また、この技術を学ぶことで、ビジネスの効率化や時間の節約につながることが期待されます。

Keywords

💡Zapier

Zapierは、様々なウェブサービスを自動的に連携させることができるオンラインツールです。ビデオの主題はZapierの高度な使い方についてであり、特にZapier CentralとOpenAIを組み合わせて自動化されたビジネスプロセスを作成する方法に焦点を当てています。

💡OpenAI

OpenAIは、人工知能技術を研究開発する非営利組織です。このビデオでは、OpenAIのAPIをZapierと組み合わせて、ビジネスプロセスを自動化する方法について説明されています。

💡AIアシスタント

AIアシスタントとは、人工知能を用いて、人間の助けを提供するソフトウェアです。このビデオでは、OpenAIのAPIを使用してAIアシスタントを作成し、ビジネスの自動化に役立てる方法が示されています。

💡データセット

データセットは、分析や機械学習に使用されるデータの集まりです。このビデオでは、ビジネスプロファイルのデータセットを作成し、AIアシスタントが正確で効果的な回答を提供できるようにすることが重要です。

💡自動化

自動化は、手動での操作を必要とせずにプロセスを実行することを意味します。ビデオのテーマは、ZapierとOpenAIを利用してビジネスタスクを自動化する方法です。

💡コンテキスト

コンテキストとは、特定の状況や文脈を指します。このビデオでは、AIアシスタントが適切な回答を提供するために、ビジネスのコンテキストを理解することが重要です。

💡API

API（Application Programming Interface）は、ソフトウェア間で通信を行うためのインターフェースです。ビデオでは、OpenAIのAPIを使用してAIアシスタントをトレーニングし、Zapierの自動化プロセスに組み込む方法が説明されています。

💡GPT-3

GPT-3は、OpenAIによって開発された高度な自然言語処理AIモデルです。このビデオでは、GPT-3を使用してAIアシスタントを構築し、ビジネスタスクを自動化する方法が紹介されています。

💡Zapier Central

Zapier Centralは、Zapierのプラットフォーム上で自動化を管理し、より高度な自動化プロセスを構築するための機能です。ビデオでは、Zapier Centralを使用してAIアシスタントをトレーニングし、自動化プロセスを改善する方法が説明されています。

💡ビジネスプロファイル

ビジネスプロファイルは、企業に関する情報を含むドキュメントやデータセットです。このビデオでは、ビジネスプロファイルを作成してAIアシスタントに提供することで、より正確な自動化結果を得ることができます。

💡メール自動返信

メール自動返信は、受信したメールに対してプログラムによって自動的に返信する機能です。このビデオでは、ZapierとOpenAIを利用してメール自動返信を設定し、ビジネスの効率を向上させる方法が紹介されています。

Highlights

Creating a new AI assistant without the need for W2 or employee insurance.

Leveraging Zapier's automation platform with AI to improve business processes.

Using OpenAI's assistance feature to create a more effective AI assistant.

Creating a business profile dataset for better AI responses.

Automatically responding to emails using the AI assistant.

The importance of context in AI models for effective outputs.

Custom instructions to improve the performance of AI models like ChatGPT.

Preparing a Google Doc as a context source for the AI assistant.

Including company information and services in the AI assistant's data source.

Setting up email response guidelines within the AI assistant's configuration.

Exporting the Google Doc as a PDF for use with AI platforms.

Using OpenAI with Zapier to reference the AI assistant in automations.

Creating a behavior in Zapier to respond to emails using the AI assistant.

Utilizing filters to identify specific inquiry emails for the AI assistant.

Testing the AI assistant's ability to draft email responses based on inquiries.

The potential for AI to automate email responses, reducing manual labor.

The cost-effectiveness of using AI models like GPT-3.5 and GPT-4.

Transcripts

in today's video you're going to create

your new assistant don't worry this new

assistant doesn't require a W2 or

employee Insurance we're going to learn

how to leverage zap your central with

creating behaviors and adding data

sources so we can start actually

understanding how this can apply to our

business and the labor associated with

our business therefore in order to do so

correctly we're going to have to use an

AI provider so we're going to use open

Ai and its assistance feature we're

going to create a data set that is our

business profile so we can get better

and more effective answers and to

Showcase his power we're going to go

ahead and automatically respond to this

email let's jump into today's video to

see the rest of the video click the link

down below it's a $50 pay no I'm just

joking so in this video this is going to

be probably one of the more powerful

videos I do on this channel especially

since basically no one's doing any

videos on zap your central it almost

like I don't think any even know maybe a

lot of people don't know about it right

now but we're going to go ahead and I'm

just going to show you a really cool way

so that when you start actually

leveraging this feature it gives you

more effective outputs because one thing

as we know when dealing with J gbt any

AI model is a huge thing that happens is

that it's not it doesn't understand

context in the sense of like why you're

using it in the specific type of output

you're looking for as we know custom

instructions can help with Chad gbt

there's different ways we can go about

it so that the outputs are more

effective in this video we're going to

learn how to get more effective outputs

of Zap your Central and more

specifically how this would apply to our

automations this is my Twitter

description down below let's jump in

today's video right off the bat go ahead

and check out that video right there if

you don't even have an idea what zamb

your central is it's going to show you

how to create behaviors add data sources

everything that this can do up to this

point then come back to this video on

this video though we're going to just go

over how to make these things even more

effective so we got our assistant your

new assistant maybe it's Jim maybe it's

Mary whatever it may be we're going to

go ahead and train this new assistant

with a data source data source is going

to be very simple create a Google doc

Microsoft doc and let's proceed the way

I want you to think of this Google doc

is it's going to be basically the

context that we provide these Bots we

get more effective outputs so provide

all relevant information that your

assistant would know in the cont context

that this was an employee such as

company name location maybe stuff like

founded CEO industry isn't as relevant

but for me I'm just going to leave it

there mission statement services offered

this is probably pretty important so for

example we got solar panel installation

Energy Efficiency Consulting sustainable

urban planning Corbin do you run this

company no I do not this is all a chat

gbt output I also want to notate that

the actual formatting of this document

is irrelevant for I mean it's relevant

if you want to see like it look cooler

or more pretty so if I bold this for my

eyes to make it like more clear to me

you can do that but when a AI model

reads this they don't care they're just

reading the data right so let's keep on

coming down here if this is relevant for

your use case you would provide key

clients maybe an award probably super

important would be contact information

such as your phone number email website

now we can kind of get into more of the

proctoring of how we want to structure

our responses depending on the context

so maybe your assistant is handling

email maybe your assistant handling

Excel sheets this is where you're going

to give the kind of the structuring of

how you want to input data how you want

to respond to emails you know everything

above the board here is our email

response guidelines to ensure

consistency and professionalism in our

communication follow these guidelines

when responding emails promptness

probably not too relevant honestly so we

can delete that tone relevant Clarity

relevant accuracy relevant escalation

possibly relevant probably not too much

in this context

so you know tone Clarity accuracy maybe

you want to add another part here that

says you know reference this email at

the end and then provide the email

whatever it may be leave it at that so

this is like what we're really looking

for here is that you are operating a

business so you know how you operate

with clients you know from past data and

past conversations past experiences what

typically gets the better responses and

if you don't then keep it General AB

test proceed in that manner then we got

frequently asked questions which can be

relevant for the context of your

business so such as how do I get a quote

for a solar panel installation what

warranty do you offer on your renewable

energy sources can your Energy Solutions

be implemented in existing buildings you

know stuff that comes up very very often

within your business once you kind of

created a docet like this and it has a

sufficient amount of data this is aerial

at 11 font I would say the max amount of

data you probably want to provide a bot

in this context is between 5 to 10 pages

Reed let's go ahead and then export this

as a PDF so I'm going to come up here to

file download PDF so we have the

information about our business Corbin

let's just jump over here and just start

going no no no we got to actually set it

up so that we can reference this

information in the context of artificial

intelligence so we're going to use open

aai you actually probably already know

what open a is if you have no clue about

zappier and open Ai and how they work

together you can check out the playlist

at the end here the first two videos are

like combined or over an hour and 10

minutes so you can get everything you

need to know about this industry but hit

create also side note I've done a ton of

videos when it comes to assistance API

in the specific context of zapier no

code friendly

solutions that sounded like a slogan

let's get the name here I'm going to go

ahead and name it the fake business here

instructions are going to be relevant

but let's just come down here real quick

to the files this is what we care about

so we're going to upload our data the

data we're uploading is that PDF we just

created together in Google doc couldn't

save don't worry uh we're going to go

ahead and add some instructions here

we're going to say follow the and then I

like to usually just give like copy

this and I'll add to quotation marks for

context on how to respond we'll leave

you at that for now now depending on

your use case this is going to give you

discretion on what model to use if

you're building out an API honestly I

would try both depending on how

effective the outputs and inputs are

just as a quick shorthand 3.5 is cheaper

than gbt 4 gbt 4 is for more the way I

like to phrase it is basically use gbt 4

outputs if a human's going to look at it

if a human's not going to look at it and

it's more of a data manipulation data

input output you'd use 3.5 for example I

can see a context here where you would

fine-tune a API assistant like this for

Excel sheets and data inputting and

outputting 3.5 could probably work in

that context but this matter we're going

to go ahead and just jump up to gbt 4

Turbo keep in mind as well in the

industry with apis and accessing AI

providers there is a downtrend when it

comes to cost other than like Cloud 3

that is more expensive than gbt 4 but

for the most part a downtrend in

November of last year when it came to

accessing an endpoint of gbt 4 it's

actually 66% cheaper now super cool

let's go ahead and hit save side note in

order to access files we do need to have

code interpreter and retrieval on okay

all right so we have our assistant here

reason we use this assistant is for the

fact that now we can reference this

assistant in zapier coming over to

zapier here let's create our Behavior

we're going to hit create behavior and

here we go let's do an event and this

could be any context but I usually like

to do the event of basically we can

command it I think that's probably the

most powerful point of this feature and

we're going to say the command is

respond

email I think that's good add trigger so

anytime I hit or anytime I type in

respond email this is going to run this

following flow so we're going to add an

action we're going to do an action of

Gmail we're going to say find email for

reference this is the example data point

coming in here this dude I don't know

who this guy is but uh he looks pretty

cool I like that sweater maybe we could

be friends is basically asking how do I

get a quote on solar pan panel

installation I'm going over here let

going and proceed so the search thing

we're going to do is typically when it

comes to receiving inquiries over email

most of the times you'll have a piece of

fix text in there so are you receiving

from Yelp Google Business whatever it

may be you'll have a piece of fixed text

in the subject l line therefore making

this entire process a lot easier when it

comes to making sure that we only find

specific emails that are inquiry emails

knowing this I'm going to reference this

tutorial right here this tutorial right

here is going to show you how to

basically grab emails that maybe aren't

perfectly structured that maybe come in

like this and you can go ahead and learn

how to use AI in order to identify

between a you know MailChimp

subscription and a actual lead email so

use that tutorial right there for the

purposes of this tutorial I'm just going

to show you how to respond and you know

use the chat bot so I'm going to just go

ahead and do this for right now

reference that tutorial up there if you

want to see how to use AI to identify or

alternatively use filters add action and

the next action here is going to be a

Chad gbt block so do chat gbt we'll do

conversation or we'll do conversation

with assistant that's the whole point of

what we just did there so we'll do

conversation with assistant we're going

to go ahead and have it generate the

message the assistant is going to be a

specific field here we you might have to

give the exact ID here unless it shows

up does show up we're going to click our

assistant so we're going to have to have

ai generate the value for this field and

everything else we're going to have ai

do except the model should be very

specific we should have that up to gbt 4

now here's the first thing that you'll

realize not Vision we don't need this to

see anything here's the first thing

you'll realize when leveraging Central

is it doesn't necessarily allow you to

reference data points and chain it that

is where this comes into play so this

can this can get a little confusing but

you have to make sure your instructions

are very specific on the way it handles

this kind of information that's why

we're basically leaving the message as

have ai generate the field but it has

context from the assistant and the

instructions we set up within the

assistant's API okay let's go ahead and

add an action here we're going to say

send email or create a draft send email

will automatically send the email to the

receive to the basically the person that

sent the email to us purposes this video

so we can kind of look at it we're going

to say create a draft

and then add action all right so this is

where the fun comes into play this may

take a couple more times to go can I

update this name I cannot update this

name we're going to go ahead and begin

here when I say uh apostrophe or

quotation marks find the email and

understand the type subject

line and body once you understand both

use we say conversation we'll just type

it out

conversation with assistant action to

write to

generate a subject line and body then

once generated output

place in the

create draft action two things first

thing this is my first time actually

proctoring in this context so we'll see

if this even works second thing I may

actually have to add more information to

the assistant API to make it more

specialized for email responsing so it

can't it maybe not be able to be a broad

version of the data let's find out

though let's test test this Behavior

what we're expecting here is we're

expecting a draft to show up that

responds to this

inquiry and let's see what happens I

want to also point out as well if this

works this is actually going to be

pretty crazy because alternatively this

would required a formatter block uh in

the context of an automation it would

have required a lot more blocks a lot

more work in order to get this kind of

output so we got the conversation with

the assistants complete let's see if it

can actually set up a draft here and we

can gut check it with our data source to

make sure it's actually the correct

answer okay I'm going to be genuinely

impressed if this works let's

see reload

this got a

draft to get a quote for the solar pan

installation fromation we have two

options contact them through the website

call their customer service line let's

see if that's what we put into our draft

or data scrolling down here over here

y'all okay initial

inquiry this the first time I'm doing

this y'all all right this is like so we

got how do I get a quote for solar panel

installation how do I get a quote for

solar panal installation you can request

it quote by contacting through

contacting us through their website or

calling our customer service

line draft contact through the website

call customer service line okay okay

okay it worked now to be honest with

y'all maybe we need to add a little bit

more information like hey when

responding to inquiries provide the

customer service line like the number or

provide the website but holy smokes

y'all this actually just

worked okay this worked it was able to

reference the data yeah no this worked

y'all so now that this works which is

like extremely cool if I want to go go

ahead and just refresh a chat here and

and if you referenced that other video I

referenced earlier now what you can do

is we can set up logic where okay first

off I want to point out something as

well which is kind of insane so you know

how I said earlier that you would need a

formatter or block like if you're

familiar with this channel you know what

I'm talking about you would need a

formatter block to split the data points

between the subject line and the body

and how to like structure everything

with this conversational with Central

what they're doing is they're adding

like a when I reference a Chad gbt block

and conversation of assistant they're

adding another layer of artificial

intelligence to like

basically smooth it over to make sure

that everything comes out more correct

in a way like this is big y'all if you

have little to no knowledge of how to

talk to these models this seems like a

very easy way to start doing this okay

right bot

command let's go ahead whatever we can

give it bot command the way we

references it though we're going to

respond email so I may need to do more

videos on this because for example what

I could do is I'm starting my workday I

have 10 unread emails I'm like you I

don't want to respond to any of that

already lasered in through the API

assistant who I am purpose everything

done some work on it all I would need to

do now and I could do a loop function to

go through the emails is simply go to my

central zampier click respond to email

and then we can set up a flow where it's

checking every single

email has

context and then responds to every

single email with drafts or

automatically okay yeah yeah this is it

it seems like they'll as well you may

need to actually Proctor and say do AC

respond email and then it will keep

going of everything we just set up

that's impressive y'all uh yeah this is

this is huge this basically I mean

imagine two years from now this if we

can do this now I mean we could

basically what I just described that

workflow we could do as well which would

be actually pretty crazy good let me

know if you want to see more videos like

this make sure you leave a like it let

me know on this channel if you want to

see more stuff like this it's completely

free I'm going to leave a playlist at

the end here showing everything we can

do with zapier I've kind of like moved

it around more some more recent videos

are on the top you get all the context

of how to start leveraging artificial

intelligence and automations in your

business but today's video we just

created your new assistant and with

basically probably a four or five hours

of just going at it you could probably

create something very effective for your

business this is also free as of now I

have no clue how they're going to charge

in the future I'll see you in the next

video that's a playist I was referring

to when it comes to zapier and

artificial intelligence and how to

leverage it that's a random video that's

my face make sure to click subscribe

click notification Bell leave a comment

all that good stuff bye-bye