Author: Pete

You joined the Teams call, someone said “we’ll send notes after,” and you already know nobody will. You want Otter in the room so you can actually listen. Here is how to set up Otter AI for Microsoft Teams, and what to do when it will not join.

The short answer: connect your Outlook calendar to Otter, turn on auto-join for the meetings you want, and ask your Teams admin to allow guests to join. Otter’s notetaker enters Teams meetings as a guest, and that one admin setting is where most setups fail.

What you need before you start

• An Otter account (a free plan exists, but check Otter’s current limits on its pricing page before you rely on it).
• A Microsoft 365 Outlook calendar. Otter says it supports the online Microsoft 365 calendar, not on-premises Exchange (Otter calendar help).
• A Teams meeting link in the calendar event. Otter shows its toggle only on events with a valid Zoom, Google Meet or Microsoft Teams link.
• Someone who can change Teams meeting settings, if you are not the admin yourself.

Connect Otter to your Microsoft calendar

1. In Otter, open Apps (or Integrations) and click Add or Connect next to Microsoft.
2. Sign in to your Microsoft account and accept the permissions.
3. Wait for your events to appear on the Otter home calendar.

If Microsoft says you need admin approval, that is your IT team’s call. Otter itself suggests asking the admin to grant the calendar permissions first, which saves a lot of back and forth.

Let the notetaker into Teams meetings

This is the step people skip. Otter’s notetaker cannot sign in with your company account. It joins as an anonymous guest, so your organisation has to allow that. According to Otter’s Teams setup guide, an admin needs to:

1. Open the Teams admin center, go to Meetings, then Meeting settings, and turn on “Anonymous users can join a meeting.”
2. Go to Meetings, then Meeting policies, and turn on “Anonymous users can join a meeting unverified.”

Two optional settings make it smoother: letting people in your org and guests bypass the lobby, and setting external bots to “Do not detect bots.” Skip them if your security team says no. You will just admit the notetaker by hand.

Choose which meetings Otter joins

On the Otter home calendar, each Teams event has a toggle. Otter’s auto-join options let you pick all meetings with a video link, only meetings you host, only external meetings, only internal ones, or only the ones you toggle yourself.

My advice: start with “only meetings I select.” Add one meeting, see how it behaves, then widen it. If you forget, you can add the notetaker mid-meeting from the microphone icon on the event.

“Notetaker failed to join”: what is going on

If you see “This meeting is not being recorded. Notetaker failed to join,” Otter says the cause is usually a Teams policy that blocks anonymous guests. Check this in order:

• The two anonymous-guest settings above are on.
• The meeting is hosted by your organisation or one where guests are allowed. An outside organiser’s settings apply there, not yours.
• The lobby is not holding the notetaker. Someone with permission needs to admit it.

What people in the meeting will see

Otter says the notetaker cannot join invisibly. It shows up in the participant list. Tell people before you hit go, and follow your company’s rules and the local recording laws. In practice, one line in the invite works: “I will have an AI notetaker in this call so I can focus on you.”

How to use Copilot in Teams for meeting notes instead

If your company already pays for Microsoft’s own tools, you may not need a guest bot at all. Microsoft’s intelligent recap needs a Teams Premium or Microsoft 365 Copilot licence, and transcription has to be turned on for the full recap (Microsoft Learn). Copilot also does not work in meetings hosted outside your own organisation (Microsoft Support).

So the split is simple. Internal meetings and a Copilot licence: use the built-in recap. Client calls hosted by someone else, or no licence: Otter is the option you control yourself, as long as guests are allowed. For a wider comparison, see our honest comparison of AI note takers.

Turn the notes into something useful

A transcript is not a result. After the call, paste the Otter summary into your chat assistant and ask for owners and deadlines. Our guide on summarizing meeting notes with ChatGPT has a prompt for it, and the AI meeting minutes generator page covers the formal version. If you live in Google’s apps, Gemini for meeting notes is the other route. Planning the next call? Try the AI meeting agenda guide.

Quick answers

Does Otter work with Microsoft Teams?

Yes. Otter lists Microsoft Teams next to Zoom and Google Meet as a supported meeting platform, once your calendar is connected and guests are allowed.

Do I need to be the meeting host?

No. You can set auto-join for meetings you host or for any calendar event with a Teams link. The meeting’s own guest settings still decide whether the notetaker gets in.

Is it free?

Otter has had a free plan, but limits change. Check Otter’s pricing page today before you plan a whole workflow around it.

Try it this week

Pick one recurring Teams meeting, connect the calendar, and test it once. Then use the time you saved to learn the next tool. The free Launchpad on the RobotsHelp home page walks you through practical AI for your actual job.

You want to learn Gemini properly, but a search for a Gemini AI course gives you a wall of paid bundles, certificates and YouTube playlists. Which one is worth your evenings?

The short answer: start with Google’s own free material. The “Google Workspace with Gemini” learning path on Google Skills is built for people who work in Gmail, Docs and Sheets, and it asks for about 30 minutes a week. If you also want a structured, certificate-style course, Google AI Essentials on Coursera takes roughly 8 hours. Whichever you pick, the learning only sticks when you use Gemini on a real task from your own job as you go.

The free Gemini AI course from Google

Google Skills (the new home of Google Cloud Skills Boost) has a no-cost Google Workspace with Gemini learning path. It lists 11 activities and covers Gemini in Gmail, Docs, Slides, Sheets, Meet, Drive and Vids. Google’s Workspace team describes it as a public, no-cost pathway of 30 to 45 minute activities, and has suggested spreading it over eight weeks at about 30 minutes a week.

There is also a shorter option on the same site: AI Boost Bootcamp: How to Use Gemini at Work 101. It is listed at about one hour. Its page says it covers writing prompts, project management, presentations, meetings and a first look at AI agents.

My take: if you have a free evening, do the bootcamp first. If you want to go deeper inside one app, say Gmail or Sheets, pick only those activities from the longer path. You do not have to finish all 11 to get something out of it.

One honest limit. Google Skills labs can use credits, and the site itself shows notices when labs have problems. Do the reading and video parts first, and treat the hands-on labs as a bonus.

Google AI Essentials: the certificate route

Google AI Essentials is a five course beginner specialization on Coursera. Google’s own page says it takes under 10 hours of self-paced study and teaches you to use generative AI tools at work, write effective prompts and use AI responsibly. You practise with a conversational tool like Gemini, but it is a general AI course, not a Gemini manual.

On cost: Google’s page says it costs $49 on Coursera, and Coursera lists it as a subscription with a 7-day free trial in the US and Canada. Prices differ by country and change, so check the enrolment page before you pay. Coursera also lets you apply for financial aid.

Worth it? Only if a certificate matters to you, for example for a CV or LinkedIn. If you just want to get faster at your job, the free path above covers more of Gemini’s actual buttons.

Which Gemini course should you start with?

A 2 week plan that fits around a job

Courses fail when they sit in a browser tab. Here is a plan built around your real work.

Days 1 to 3: choose one task you repeat

Pick something weekly that you slightly dislike: a status update, a long email thread, a meeting you have to write up. Note how long it takes you today. That is your own baseline, so you can see later whether Gemini really helps.

Days 4 to 8: the Gemini at Work bootcamp, then one app

Do the bootcamp in one or two sittings. Then take the one Workspace app you use most and do its activities from the learning path. After each lesson, try the idea on your task. If you have not used Gemini at all yet, our Gemini tutorial for beginners walks through the first conversation.

Days 9 to 12: put it inside your inbox and documents

Try it where you already work. Our guides on using Gemini in Gmail and using Gemini for work show practical examples you can copy and adjust.

Days 13 and 14: keep what worked

Write down the two or three prompts you reused. Save them somewhere you can find them. Compare the time on your baseline task. If it did not save time, drop it and move on. That is a result too.

Do you need a paid Gemini course?

Not first. The free Google material covers the product basics and prompting. Pay for something only if you need what the free path cannot give you: feedback on your own work, a fixed schedule, or a group that keeps you going. Before paying anyone, check what is in the syllabus, when it was last updated, and whether it teaches tasks you do. Tool names and menus change often, so a course recorded a year ago may show screens that no longer exist.

Gemini, Claude or ChatGPT: do you need to choose?

Not to start learning. The prompting skills carry across tools. If you are still deciding which to pay for, read our comparison of Claude vs Gemini Pro at work. We also wrote about the Claude AI course and the ChatGPT certification course if you want to compare learning paths.

Frequently asked questions

Is there a free Gemini AI course?

Yes. Google offers a no-cost Google Workspace with Gemini learning path and a short Gemini at Work bootcamp on Google Skills. Check each page for current availability, since Google updates them.

How long does it take to learn Gemini?

You can be useful with it after an hour or two. Getting comfortable with it in your daily apps takes a couple of weeks of using it on real tasks, which is why the plan above is two weeks.

Do I need to be technical?

No. Everything above is aimed at people who work in documents, email and spreadsheets.

Where to go next

Courses show you the buttons. A short weekly lesson helps you keep going. Launchpad is free, and you only need your email.

Prefer a plain link? Join Launchpad on robotshelp.me.

You keep hearing that everyone at work is getting “AI certified,” and you wonder if you should be too. Good question. Here is the short answer: yes, a ChatGPT certification course exists, but what you can actually get today is smaller than the headlines make it sound.

Robotshelp Launchpad is free, and it’s where I’d start if you want to use ChatGPT at work this week.

The quick answer: what a ChatGPT certification course is today

OpenAI runs two separate things, and people mix them up.

The first is OpenAI Academy. It offers free, self-paced courses for anyone with a ChatGPT account. The entry course is called AI Foundations, and OpenAI’s help center estimates 60 to 75 minutes. If you pass the assessment with 80% or more, you earn a badge. Finish every course in a pathway and you can get a certificate of completion.

The second is OpenAI Certified, the formal credential. It runs through Coursera inside ChatGPT, and right now it is available only to ChatGPT Enterprise and Edu workspaces, by invitation. You can’t simply sign up for it as an individual.

OpenAI is also clear that the Academy badges and certificates are not certifications and do not guarantee you a spot in a future certification. Sources: OpenAI Academy courses (OpenAI Help Center) and OpenAI Certified app (OpenAI Help Center).

Which ChatGPT course should you take first?

Pick by what you need, not by the name on the certificate. Based on OpenAI’s own course list:

• AI Foundations (60 to 75 minutes): the basics of AI, clearer prompts, adding context, checking outputs and using AI responsibly. This is the right first step if you’re new.
• Applied AI Foundations (75 to 90 minutes): breaking a job into steps and building a repeatable workflow with review points. Better once you already use ChatGPT a little.
• Agents and Workflows (75 to 90 minutes): handing structured tasks to an agent, with boundaries and checkpoints.

There are also courses for educators and college students, and API-focused ones for developers. Those are a different lane.

Is a ChatGPT certificate worth it for your career?

Honest take, in my own words: a completion certificate shows you finished a course. It doesn’t prove you can do the work. A hiring manager will care far more about whether you can show a real task you sped up.

That doesn’t make the course useless. An hour of structured practice on prompts and reviewing outputs is a fair price for a clean start. Just don’t pay for a promise. The official OpenAI Academy courses are free, so if a site charges you for a “ChatGPT certification,” check what you are actually buying and who issues it.

How to spot a certification that isn’t from OpenAI

Lots of other providers sell courses with “ChatGPT” in the title. Some are good. Before you pay, ask three things:

1. Who issues the certificate, and is that the company on the badge?
2. What will I be able to do at the end, in one sentence?
3. Does the course say whether the credential is official, or just a completion certificate?

If the answers are vague, skip it. Time spent on a real task beats a PDF.

Build proof you can use, not just a certificate

Whichever course you take, finish with something you can point to. Pick one boring task from your week, like replying to client emails or summarizing a long document, and run it through ChatGPT with the habits the course teaches: clear instructions, the right context, and a human check at the end.

Write down the before and after. How long did it take you before, and how long now? That one honest line is worth more in an interview than a badge.

Where to go from here

Do AI Foundations first, since it’s free and short. Then pick one real task and practice it for a week. If you want guided practice that stays in the real world of work, with no jargon, that’s what I’m building at Robotshelp.

Join Launchpad, free

You need a header image for a client deck, a simple diagram for a proposal, or a visual for a post, and the designer is booked until next week. You can make a decent first version yourself in about five minutes with Gemini.

The short answer: open Gemini, choose Images in the sidebar, and describe what you want: the subject, what it is doing, the setting and the style. Then refine it with follow-up messages instead of starting over. The rest of this page shows how to do that for real work, and where it falls short.

Create your first image in Gemini

These are the steps from Google’s own help page for the Gemini app on a computer.

1. Sign in to the Gemini website with your Google account.
2. Open the sidebar at the top left and choose Images.
3. Pick a template or type your prompt.
4. Look at the result, then keep chatting to change it.
5. Hover over the image you like and download the full size version.

Google also says you can export an image to Docs if you want to keep working on it there. Details can change, so check the official Gemini image help page if a button looks different on your screen.

How to write a prompt that gives a usable image

Google’s tips are simple. Start with a verb like draw, generate or create. Name the style, for example photorealistic, watercolor, cartoon illustration or charcoal drawing. Then describe the subject, what it is doing and the setting.

For work images I add two more things: where the image will be used, and what to leave out. “Wide banner for a slide” gives you a different composition than “square profile picture”, and “no text” saves you from garbled words.

That one prompt covers the verb, the style, the subject, the setting and the layout. It is the same structure every time, only the details change.

Refine instead of restarting

The first image is rarely the final one. Gemini keeps the conversation, so you can ask for one change at a time: “make the lighting warmer”, “remove the plant on the right”, “change the style to a flat illustration”. Google describes this as local edits, meaning quick and specific changes to parts of an image.

One change per message works best. If you pile five changes into one request, you usually get a mix where two of them land and three do not.

If you already have a photo and want to change it rather than create a new picture, that is a different job. I wrote it up in how to use Gemini to edit photos.

What Gemini is good for at work

Google lists several strengths of its current image model: better text in images, following more complex instructions, keeping a character looking the same across several images, and more accurate diagrams and infographics. In practice, these are the work uses I would try first:

• Slide visuals. A custom picture beats a generic stock photo, and you control the empty space for your title.
• Simple diagrams and infographics. Good for a first draft of a process or a comparison. Always check every label and number yourself.
• Social and newsletter images. Describe the mood and the format, and ask for no text so you can add your own.
• Mockups and concepts. Show a client a rough direction before you pay for final artwork.

Limits you should know before you rely on it

Be honest with yourself about three things.

Always read the text. Text rendering is better than it used to be, but a misspelled word on a client slide is your problem. Read every word of the image.

Quotas and plans differ. Google notes that you can download at 2K resolution with a Google AI plan and at 1K without one, and that paid subscribers can redo an image with a stronger model for extra detail. If you hit your daily limit on the standard model, that redo is not available either.

Work accounts follow different terms. Google says users with a work or school account may be subject to different terms depending on the license. Check with your admin before you use it for company material. Also, do not use images that could violate someone else’s copyright or privacy, and read what Google says about using generated content before you publish it.

My own rule: use Gemini for the first draft and for anything internal. For a brand logo or a hero image that stays on your site for years, have a person finish it.

Gemini or another tool?

Gemini is convenient if you already live in Google’s apps. It is not the only option. If you use ChatGPT daily, read how to use ChatGPT to create images. For a wider comparison of tools, see the best AI tool for image generation. And if you are new to the app itself, start with the Gemini tutorial for beginners or how to use Gemini for work.

A five minute routine you can repeat

1. Write one sentence: where will this image be used?
2. Use the prompt pattern: verb, style, subject, setting, layout, what to leave out.
3. Generate, then change one thing per message, two or three rounds at most.
4. Check text, hands, logos and anything factual.
5. Download the full size version and drop it into your slide or post.

Once that routine is a habit, a visual stops being a task you postpone and becomes something you do between two meetings.

Want more practical walkthroughs like this one, one real task at a time? Join Launchpad below.

The meeting ended ten minutes ago. Someone has to send the minutes, and you know it will be you. An AI meeting minutes generator can write the first version for you, and you only have to check it.

The short answer: record or transcribe the meeting, paste the transcript into ChatGPT, Claude or Gemini, and give it a fixed template for decisions and action items. The prompt below does that. It is free to try with the free plans of these tools.

Minutes are not the same as meeting notes

Notes are for you. Minutes are the official record: who was there, what was decided, who does what, and by when. People who were not in the room should be able to read them and know where things stand. That is why a general “summarize this” prompt often disappoints. It gives you a nice paragraph, but no owners and no deadlines.

If you only want a quick summary or a note taker that joins your calls, see our guides on the best AI note taker for meetings and Gemini for meeting notes. This page is about the formal minutes document.

Step 1: Get a transcript

The AI can only write minutes from what it can read. You have three simple options:

• Your meeting tool. Google Meet has a “Take notes for me” feature, and Microsoft Teams has Copilot recaps. Both depend on your plan and your admin settings, so check the official pages: Google Meet help and Microsoft Support.
• A recording. Record the call (with everyone’s permission) and turn the audio into text. Our guide on how to transcribe audio with AI shows how.
• Your own bullet points. Even five rough lines typed during the call work, as long as you tell the AI that this is all it has.

Before you upload anything, ask yourself if the meeting contained client data, HR topics or anything confidential. If yes, check your company’s AI policy first, or remove names and numbers.

Step 2: Paste this prompt

This is the template I would use. Copy it, paste the transcript at the bottom, and send.

You are my meeting secretary. Turn the transcript below into formal meeting minutes.

Use exactly this structure:
1. Meeting title, date, and attendees (only names that appear in the transcript)
2. Purpose of the meeting (one sentence)
3. Decisions made (one line each)
4. Action items (task | owner | deadline). If the owner or deadline was not said, write "not stated".
5. Open questions to follow up
6. Next meeting (only if mentioned)

Rules: do not add anything that is not in the transcript. Keep each line short. If something is unclear, flag it with [check] instead of guessing.

Transcript:
[paste here]

Two parts of the prompt matter most. “Not stated” stops the AI from inventing an owner. The [check] flag makes unclear moments visible, so you do not send out a confident mistake.

Step 3: Check it in two minutes

Never send AI minutes unread. Go through this list:

1. Are all decisions really decisions, or just ideas someone floated?
2. Does every action item have the right person next to it?
3. Are the deadlines the ones people agreed on?
4. Is anything in there that nobody said?

Fix what is wrong, then send. Checking a draft is usually quicker than typing the minutes yourself, and the document looks the same every week.

Which free AI tool should you use?

All three can do this job. Use the one you already have open. If your transcript is very long, Claude and Gemini are often comfortable with big documents, but limits change by plan, so test with your own file. If you want a direct comparison, read ChatGPT vs Claude vs Gemini.

Turn minutes into follow-up

Minutes are only useful if the actions happen. Once you have the action list, ask the AI for a short follow-up email per owner, with only their tasks in it. If you want the result as a longer report, here is how to turn meeting notes into a report. And if the next meeting needs a plan, try an AI-written meeting agenda.

Common mistakes

• No template. Without a structure, you get a different format every time.
• Trusting names. Speech-to-text often gets names wrong. Check them.
• Recording without asking. Tell people. In many places it is also a legal matter.
• Sending it raw. The AI writes the draft. You are still the one who signs.

Frequently asked questions

Can I generate meeting minutes with AI for free?

Yes. The free plans of ChatGPT, Claude and Gemini can turn a transcript into minutes. Free plans have usage limits, so a very long transcript may need to be split in two.

What should meeting minutes include?

Date, attendees, purpose, decisions, action items with owner and deadline, open questions and the next meeting. The prompt above covers all of them.

Is it safe to paste a meeting transcript into an AI tool?

It depends on what is in it. Leave out confidential details, or use the tool your company has approved.

Want to learn more ways to use AI for real work, in short practical steps?

It is the 28th, your receipts are in three places, and the bank export looks like a wall of numbers. You do not need a new accountant tonight. You need a routine that stops the pile from growing.

AI bookkeeping for small business works best as a helper for the boring parts: reading receipts and invoices, sorting transactions into categories, and flagging what looks odd. A person still approves the books. Used that way, AI can cut the typing and leave you with a short list of decisions.

Want to learn practical AI like this, step by step? Join the free Launchpad here.

What AI can do for your books (and what it should not)

Bookkeeping is mostly repeating small jobs. That is why AI fits it. These are the jobs where it helps most:

• Reading documents. A photo of a receipt or a PDF invoice becomes a row: date, supplier, amount, tax.
• Suggesting categories. “Adobe, 54.99” goes to software, “Shell, 62.10” goes to travel. You confirm or fix it.
• Matching. Which invoice does this bank payment belong to?
• Spotting oddities. The same supplier charged twice, or a payment far above normal.
• Explaining numbers. Paste a monthly summary and ask what changed compared with last month.

Here is what it should not do alone: file your tax return, decide how a tricky item is treated, or change your books without anyone looking. Tax rules depend on your country and your situation, so keep a qualified accountant for those questions.

Step 1: Get your records into one place

AI cannot sort what it cannot see. Pick one folder, shared inbox or accounting app as the home for every receipt and invoice. Then send paper receipts there as photos the day you get them.

The IRS lists the reasons good records matter: they help you monitor the business, identify income, track deductible expenses and support what you report on your tax return (IRS: Recordkeeping). That is the standard your AI routine has to meet. Rules differ outside the United States, so check your own tax authority.

Step 2: Let AI read and categorize

You have two routes, and both are fine for a small business.

Route A: the AI features in your accounting app

Many accounting apps now suggest categories and read receipts. This is the easiest start, because the books and the AI live in one place. Check the app’s own help pages for what it does with your data and what it costs.

Route B: your own chat assistant

Export a month of bank transactions as a spreadsheet, remove account numbers, and ask a chat assistant to propose categories. Try a prompt like this:

Here are my business bank transactions for October (CSV below).
Add a column with a category from this list only: software, travel, meals, supplies, advertising, fees, other.
Add a column "unsure" with yes or no. Say yes when the description is vague.
Do not guess. Return a table.

The “do not guess” line matters. You want unsure items marked, not hidden. Review every row marked yes yourself.

If you want to go further and skip the typing, you can automate the receipt step. Our guide to an n8n workflow for automatic invoice processing shows Gmail invoices landing in a sheet, and AI invoice processing software helps you pick a ready-made tool.

Step 3: Reconcile with a human check

Reconciling means checking that your books match the bank. AI can propose matches, but you decide. A simple routine that takes about twenty minutes a week:

1. Open the list of unmatched transactions.
2. Accept the matches that are obviously right.
3. Look at every item the AI marked unsure.
4. Look at the five biggest amounts, whatever the AI says about them.
5. Save a note on anything you changed, so you can explain it later.

Spot checks beat blind trust. If you find a mistake, add a rule to your prompt, for example “payments to this supplier are always software”.

Keep your data safe

Bank and invoice data is sensitive. Before you paste anything into a chat tool, read its privacy and data settings, and remove account numbers, names of private customers and anything you do not need. If a client’s data is in the file, check your contract and local privacy rules first. When in doubt, use the AI feature inside your accounting app, where the data stays in one system you already trust.

How long should you keep the records?

Scanning receipts does not mean you can throw the originals into the bin without thinking. The IRS says you must keep records as long as needed to prove the income or deductions on a tax return, and that in the common case this is 3 years from filing, with longer periods in some situations such as 6 years for under-reported income of more than 25% of gross income (IRS: How long should I keep records?). Those are United States rules. Ask your accountant what applies where you live, and keep digital copies somewhere backed up.

A simple first week

• Day 1: choose one home for all receipts and invoices.
• Day 2: run last month’s transactions through the categorize prompt.
• Day 3: review every “unsure” row and fix your category list.
• Day 4: try the AI features in your accounting app on one week of data.
• Day 5: do the twenty minute reconcile routine and write down what you changed.

After a week you will know which part saved time and which part you still want a person to handle. For the wider picture of what to hand over first, see AI automation for small business.

Common questions

Can AI replace my bookkeeper?

For the typing and sorting, it can take over a lot. For judgment, such as unusual items, year end and tax questions, a qualified person is still the safer choice. Many owners keep a bookkeeper for a monthly review and let AI do the daily sorting.

Is it safe to paste my bank data into a chat assistant?

Only after you check the tool’s data settings and remove what you do not need. Account numbers and private customer details stay out.

Does it cost a lot to start?

You can try the chat route with a free plan and a spreadsheet. Accounting apps with AI features usually charge a monthly fee, so check the current price on their pricing page.

Ready to build practical AI habits like this one? Join the free Launchpad.

Your inbox fills up with the same five questions. Where is my order, what are your opening hours, can I get a refund, do you ship abroad, how do I reset my login. You answer them well, and you answer them again tomorrow. That is the moment most small teams start looking at an AI chatbot for customer support.

The short answer: a chatbot works well when it answers only from a short, written set of facts you gave it, and hands the customer to a person the moment it is unsure. Start with your top ten questions, test it yourself before any customer sees it, and keep a clear “talk to a human” option. That is the whole recipe.

What an AI chatbot for customer support actually does

It sits in a small chat window on your website, or inside your help desk, and answers visitors in plain language. You feed it your own material: a FAQ page, shipping and returns policy, price list, opening hours. It then replies using that material and tries not to wander off it.

It is not the same as an AI agent that works behind the scenes on your inbox. If you want that wider setup (sorting tickets, drafting replies for you to approve), read our guide to automating customer support with AI agents. This page is narrower: the chat window your customers talk to, and how to put one live without embarrassing yourself.

Step 1: write down your top ten questions

Open your last 50 support emails or messages. Mark every question that has a clear, factual answer. Most small businesses find that a handful of questions make up the bulk of the pile.

Write each question with its answer in two or three sentences, in the tone you would use with a customer. Put it all in one document. This document is the single most important part of the project. A mediocre tool with a good document beats a great tool with a vague one.

Be honest in the document. If the refund window is 14 days, write 14 days, and write what happens on day 15. A bot cannot be more accurate than the page you gave it.

Step 2: decide what the chatbot must never answer

This is the step people skip, and it is the one that protects you. Make a short “hand off” list. For example:

• Anything about a refund beyond what your written policy says
• Complaints, angry messages and anything that mentions a lawyer
• Medical, legal or financial advice
• Questions about a specific order that the bot cannot look up
• Any message where the customer asks for a person

Why it matters: a company is responsible for what its chatbot says. In 2024 a Canadian tribunal ruled that Air Canada had to honour a bereavement fare the airline’s own chatbot had described wrongly to a customer. The airline argued the bot was separate from the company. The tribunal did not accept that. You can read the story at Ars Technica. The lesson is simple: the bot speaks in your name, so limit what it is allowed to promise.

Step 3: choose a type of tool

Skip the long feature tables. There are really three kinds, and the right one depends on how much you want to set up yourself.

A chatbot built into your help desk or shop platform

If you already use a help desk or an online shop system, check its settings first. Many now include an AI answer feature that reads your help articles. It is the least work, and it usually hands off to your existing inbox cleanly.

A stand-alone chatbot builder

You upload your document or point it at your website, and it gives you a small snippet to paste into your site. Good for sites without a help desk. Check that you can see every conversation afterwards, because you will learn more from those logs than from any review.

A custom build with a workflow tool

If you want control over what it can look up, such as order status, a workflow tool can connect a language model to your own data. It takes longer and needs care. Our page on a AI agent builder that needs no coding is the place to start, and the guide to conversational AI agents for businesses explains the moving parts.

My own view: for a team of one to ten people, begin with the first or second kind. Build custom only when you know from the logs which questions need real data.

Step 4: test it like a difficult customer

Before it goes live, spend 30 minutes trying to break it.

1. Ask each of your ten questions in three different wordings.
2. Ask something that is not in your document. The right answer is “I am not sure, let me get a person”, not a confident guess.
3. Ask for a discount, a refund outside the policy, and something slightly rude.
4. Type in another language if your customers do.
5. Ask for a human and see how fast you reach one.

Write down every wrong or odd answer. Almost every fix is a change to your document, not to the tool. If you want ready phrasing for tricky replies, our ChatGPT prompts for a customer support agent give you a starting point for the tone.

Step 5: go live small, then read the conversations

Switch it on for one page, such as your contact or shipping page, rather than the whole site. Put a clear line in the chat window: “I am an AI assistant. I can help with common questions, or connect you to a person.” Telling people it is an AI is the honest thing to do, and in some places disclosure rules are tightening.

For the first two weeks, read every conversation. Look for three things: questions it could not answer (add them to the document), answers that were wrong (fix the document), and customers who gave up (check your hand-off path). After two weeks you will know whether it saves you time. Count the conversations it closed without you, and the ones that still landed in your inbox.

When a chatbot is the wrong tool

If most of your messages are individual, emotional or complicated, a bot will frustrate people. The same goes if you get fewer than a few questions a day: a good FAQ page and a quick personal reply may serve you better. Also, if people mostly phone you rather than type, look at an AI voice option instead, such as in our guide to an AI voice agent for small business.

A chatbot is not there to replace the relationship with your customers. It is there to give you your afternoon back, so that the questions that need you get your full attention.

Quick checklist before you launch

• One document with your top ten questions and honest answers
• A written list of what the bot must hand off
• A visible way to reach a human
• A line that tells people they are talking to an AI
• A test session with wrong, odd and rude questions
• A habit of reading the conversations every day for two weeks

Learn the practical side, one small project at a time

If you want more step-by-step guides like this for everyday work, join the free Launchpad on the RobotsHelp home page. You enter your email and you are in.

It is 4:45 on a Friday, you are under a client’s sink or in a meeting, and the phone rings for the third time. By Monday that caller has booked someone else. Missed calls are quiet losses, and that is why small business owners start looking for an AI receptionist.

Here is the short answer: the best AI receptionist for a small business is the one that handles your three most common call types (hours and prices, booking, and “please call me back”) and hands everything else to a person. Pick by your call mix, test it with real calls, and keep a human fallback. A tool that sounds great in a demo but cannot book into your calendar is the wrong tool.

What an AI receptionist actually does

An AI receptionist answers the phone with a synthetic voice, understands what the caller wants, and responds from the information you gave it. Good ones can answer common questions, collect a name and number, and book a slot in a connected calendar. They do not run your business. They cover the gap when nobody can pick up.

It is a close cousin of the AI voice agent. If you want the wider picture first, read our guide on an AI voice agent for small business. This page is narrower: how to choose a receptionist for the front desk job, and how to compare it with the other options.

AI receptionist vs answering service vs a person

An AI receptionist is not the only way to stop missing calls. Here is the honest comparison.

A human answering service

A real person answers in your company name and passes on a message. It handles odd, emotional or complicated calls well. The trade-off is that it usually costs more per call, and the person only knows what is in the script you wrote for them.

An AI receptionist

It answers every call, at any hour, and does not get tired on a busy Monday. It is strongest on repetitive questions and booking. It is weakest on anger, unusual requests and heavy accents or bad phone lines. Treat it as a first line, not the last.

A person on your team

Still the best for relationships. Also the one thing that cannot answer at 7 pm. Many small teams end up with both: AI picks up, a human handles anything the AI marks as important.

My own view: if fewer than a handful of calls a week are missed, a simple voicemail greeting and a text back may be enough. An AI receptionist earns its place when missed calls clearly cost you jobs.

What to look for before you choose

Skip the feature lists. Check these six things, in this order.

• Calendar booking. Does it book into the calendar you already use, and does it avoid double bookings?
• Handoff. Can it transfer a call to your phone, or send you a text with the caller’s details when it gets stuck?
• Your own answers. Can you give it your hours, prices, service area and the questions you hate repeating?
• Call recordings and transcripts. You need to read what it said. Without this you cannot fix mistakes.
• Language and voice. Does it handle the language and accents your callers actually use?
• Clear pricing. Per minute, per call or monthly? Ask what happens when a month is busy.

The call mix test: pick by what your phone really does

This takes ten minutes and beats any review site. Look at your last 20 calls and sort them into three piles.

1. Information: “Are you open Saturday?” “Do you cover my area?”
2. Booking: “Can I come in on Thursday?”
3. Everything else: complaints, quotes, suppliers, existing jobs.

If most calls sit in piles one and two, an AI receptionist will help a lot. If most are in pile three, you need a person or a good answering service, with AI as backup at most.

How to try one this week without risk

1. Write a one page cheat sheet. Hours, services, prices you are happy to share, address, and what to say when someone asks for a quote.
2. Start with after hours only. Forward calls to the AI only when you are closed or busy. Your normal day stays untouched.
3. Make ten test calls yourself. Ask odd questions. Interrupt it. Say your name badly. Note where it fails.
4. Read the transcripts for a week. Fix the cheat sheet where the AI guessed.
5. Decide after two weeks. Count bookings it took and calls that needed rescuing. Keep it, tune it or drop it.

If your main goal is booking, the next read is our walkthrough on an AI agent for appointment booking. If callers mostly ask the same questions, see how to automate customer support with AI agents.

Rules you should not skip

Two things matter more than the brand you pick.

Tell callers it is an AI. People trust a business more when it is honest, and in some places disclosure is a legal matter. Say it in the greeting.

Know the rules on AI voices. In the US, the FCC confirmed in 2024 that the “artificial or prerecorded voice” rules of the TCPA cover AI voices that sound human. That is mainly about outbound calls and consent, so do not use a receptionist tool to cold call people. Read the FCC’s own page: FCC confirms TCPA applies to AI technologies that generate human voices. If you are outside the US, check your own rules on call recording and consent before you switch on transcripts.

Mistakes that make callers hang up

• A greeting that runs longer than ten seconds.
• No way to reach a human. Always offer one.
• Letting it quote prices you have not approved.
• Never reading the transcripts after week one.
• Promising “24/7 service” when nobody checks the messages until Wednesday.

FAQ: AI receptionist for small business

Is an AI answering service the same thing?

People use the two names for the same idea. “Answering service” often means a human team, and “AI receptionist” means software. Check which one a vendor really sells before you compare prices.

Will it replace my receptionist?

For most small teams, no. It takes the repetitive calls so the person can focus on the callers who need a human.

How much does it cost?

Pricing differs a lot between tools and changes often. Ask for the monthly price at your real call volume, and ask what is included. We do not quote numbers here because they go out of date fast.

Can I build my own?

Yes, with a voice tool and an automation platform, but a ready made receptionist is usually the faster start. Build only when you need something very specific.

Your next step

Do the call mix test today. It costs nothing and tells you whether you need an AI receptionist, a human service or just a better voicemail. When you want help turning AI into a real result at work, join Launchpad below.

Once calls are covered, the next pile is usually receipts and invoices. Here is a safe way to start with AI bookkeeping for small business.

An agency just sent you a proposal with a monthly fee, a list of “AI-powered” promises and a contract that runs for six months. You are not sure whether you are buying real work or a nice-looking tool you could run yourself. Fair worry.

Here is the short answer: hire an AI lead generation agency when you have budget, an offer that already sells, and no time to run outreach. Do it yourself when you are still finding out who buys, or when the money is tight. Most small teams do best with a middle path: DIY first, agency later, with their own data in hand.

What an AI lead generation agency actually does

Strip away the buzzwords and an agency does four jobs. It builds a list of people who might want what you sell. It writes messages to them. It sends those messages through email, LinkedIn or ads. And it hands you the replies that look interested.

AI changes the speed of the first two jobs. Tools can draft a first version of a message in seconds and sort a list by simple rules. They do not change the hard parts: knowing who your buyer really is, and having an offer worth replying to. An agency that skips those two is only sending more messages faster.

Agency vs DIY: the honest comparison

When an agency makes sense

• You already close deals from the leads you get today, and you know what a lead is worth to you.
• Your time is the bottleneck, not your budget.
• You need volume across several channels at once.
• You want someone else to deal with deliverability, list upkeep and follow-ups.

When doing it yourself is the better move

• You are still testing who buys. Paying someone to scale a guess burns money.
• Your offer needs your own voice. Founder-written messages often beat polished ones when the audience is small and specific.
• You want to understand the numbers before handing them to someone else.
• Your lead volume is low. A few well-researched messages a day can be enough.

My own view: if you cannot describe your ideal customer in two sentences, an agency will not fix that. It will just guess in public, with your name on it.

Questions to ask before you sign

A good agency answers these plainly. A weak one dodges.

1. Who writes and approves the messages? You should see them before they go out, at least at the start.
2. Where does the contact list come from? Ask how people were found and whether they can be removed on request.
3. What do you count as a lead? A reply, a booked call and a qualified buyer are three very different numbers.
4. Whose accounts and domains are used? If they send from your domain, a mistake lands on your reputation.
5. What happens if it does not work? Look at the notice period and what you keep: lists, scripts, results.
6. Can I see what you did for a client like me? Ask for specifics you can verify, not a screenshot of a happy reply.

The rules you stay responsible for

Hiring someone does not hand them your legal duties. In the US, the FTC’s CAN-SPAM rules ban false header information and misleading subject lines, require a way to opt out and a valid postal address, and say you cannot have another party send email on your behalf to an address that has opted out. You can read the summary in the FTC’s guide for commercial emailers. Rules in other countries differ, so check yours.

Mailbox providers have their own demands too. Google’s email sender guidelines set requirements for authentication and unsubscribing, and senders that ignore them risk landing in spam. Ask any agency how they handle this before they touch your domain.

How to do AI lead generation yourself this week

You do not need an agency to find out if the idea works. Here is a small test you can run in about a week, using tools you probably already have.

1. Write down your buyer. One job title, one type of company, one problem. Ask ChatGPT or Claude to poke holes in it and list what that person worries about on a Monday morning.
2. Build a list of 50 people by hand. Use public sources like company sites and professional profiles. Small and accurate beats big and sloppy.
3. Draft one message per person. Let the AI write a rough draft from two facts you found, then rewrite it in your voice. Never send raw output.
4. Send a few a day and track replies. Note which opening line gets answers. After a week you have real numbers, which is exactly what an agency would otherwise charge you to find out.
5. Automate only what already works. Once a message gets replies, connect a simple workflow. Our guides on using AI for lead generation and on an AI agent that qualifies leads show how.

If your CRM is a spreadsheet and a mess, start with CRM automation for small business. For the bigger picture, see AI automation for small business.

What it costs, and how to compare offers

Prices vary widely, so I will not quote a number. Compare offers by what you get per month: how many contacts, how many messages, which channels, and who does the work. Then divide the fee by the leads you actually expect to close. If the agency cannot help you work that out, that tells you something.

Also watch for long lock-in contracts at the start. A short pilot of a month or two shows you the real quality before you commit.

A middle path that works for most small teams

Run the DIY test first. Once you know which message and which audience get replies, you have something an agency can actually scale. You will also brief them better, judge their reports properly, and spot padding. If the test works well and you enjoy it, you may find you do not need them at all.

Quick answers

Is an AI lead generation agency worth it for a small business?

Sometimes. It pays off when you already know what a lead is worth and lack the time to run outreach. It is a poor fit when you are still unsure who your buyer is.

Can AI replace a lead generation agency?

For small volumes, yes. AI covers drafting and sorting well. You still supply the target customer, the offer and the judgment on what to send.

What is the biggest risk of hiring one?

Sloppy sending in your name. Poor lists and spammy messages hurt your domain reputation, and the legal duties stay with you.

Want to get comfortable with these AI tools first, before spending on anyone? Start free.

If you are also thinking about answering new enquiries faster, our guide to the best AI receptionist for small business shows how to test one with a short call.

Your studio runs on spreadsheets, a shared inbox and one person’s memory of which consultant owes you a drawing. You search for project management software for architects and find a dozen tools that all claim to fix it. Here is the short answer: you need two layers, not one tool. One layer tracks fees, hours and invoices by phase. The other tracks who does what by when. AI then helps with the writing around both.

Want to start using AI on the admin side of your practice? Join the free Launchpad at RobotsHelp.

Project management software for architects: the two layers

Most “best of” lists mix two different kinds of software. A practical comparison from Arkscape splits them the same way: for fees, timesheets and invoicing it points to Monograph, BQE CORE or Deltek Ajera (all quote only, per Arkscape), and for who is doing what by when it points to Basecamp, Asana, ClickUp and similar task tools.

• Practice management (money layer): phase budgets, timesheets, invoices, staffing forecasts. This is where you see that schematic design has burned 80% of its fee.
• Task and coordination (work layer): boards, deadlines, owners, comments. This is where “send revised plans to the structural engineer” lives.

A two-person studio can often live on the work layer plus a time tracker. A firm that bills by phase and hires staff will want the money layer sooner than it thinks.

What to look for before you compare tools

1. Phases that match your contracts. If you bill by design phase, the tool should budget and report by phase, not just by project.
2. Timesheets people will actually fill in. The best budget report is useless if hours are guessed on Friday afternoon.
3. Accounting sync. Several architecture tools connect to QuickBooks Online. Check that yours matches the books your accountant uses.
4. Consultant and document handling. RFIs, submittals and drawing issues need a home. Some firms add a dedicated document tool for that.
5. Price you can verify. Many architecture-specific tools publish no price. Monograph’s own cost comparison lists BQE Core and Deltek Ajera as quote-only, BigTime Essentials from $20 per user per month, and Harvest Teams at $9 per seat per month billed annually. Prices change, so confirm on the vendor’s page.

The main options, in plain words

These descriptions follow the vendor-neutral comparisons from Arkscape and Scoro. Both are published by companies that sell competing tools, so treat them as a starting list and run your own trial.

Monograph

Cloud practice management for architecture and engineering firms: timesheets, phase-based fees and budgets, invoicing and staffing forecasts. Arkscape describes it as a fit for small to mid-size firms that bill by phase and keep their books in QuickBooks Online.

BQE CORE

Time and expenses, billing and project management, with optional accounting, CRM, HR and payroll modules. It suits firms that want fewer systems, and Arkscape notes it can take real time to set up.

Deltek Ajera

Project accounting and scheduling aimed at small architecture and engineering firms, with quote-only pricing.

Basecamp, Asana, ClickUp

General task tools for the work layer. They do not know what a design phase is, but they are quick to set up and cheap to try. Many studios start here and add a money layer later.

Where AI helps an architect’s project management

None of the tools above writes your meeting minutes or turns a messy site-visit note into an action list. A chatbot such as ChatGPT or Claude can. These are the jobs worth handing over this week.

Turn a consultant meeting into minutes and actions

Paste your rough notes and ask for decisions, open questions, owners and dates. Our guide to Gemini for meeting notes shows the full workflow, and the same idea works in any chatbot.

Draft the phase checklist and the project brief

Give the chatbot your scope, phases and client type and ask for a checklist per phase that you then edit. The reusable brief in how to use AI for project management keeps every answer consistent, and how to use AI to create a project plan covers the plan itself.

Write the status email nobody wants to write

Feed it three bullet points on progress, the one risk and the one decision you need from the client. Ask for a short, polite update. You still read it, change what is wrong and send it yourself.

Summarise a long document before you read it

Specs, tender documents and long email chains can be summarised first, so you know where to look. For contracts, use AI only to flag questions for your lawyer. Our piece on AI contract review software explains where the limits are.

Reusable prompts

If you want a starting library, see ChatGPT prompts for project management. Change the wording to your own phases and clients.

What AI should not do in your practice

• It does not replace your fee and hours data. Numbers from your timesheets and invoices stay in the money-layer tool. A chatbot can explain a report, but it should not invent one.
• Check confidentiality first. Before pasting client details, look at your contracts and your AI provider’s data settings. When in doubt, remove names and addresses.
• Review every output. A wrong date or a wrong scope line in a client email costs more than the minutes you saved.

A simple way to choose this month

1. Write down the three questions you cannot answer today. For example: which phase is over budget, who has the latest drawings, what did we promise the client on Tuesday.
2. If the questions are about money, trial one practice management tool with a real project for two to four weeks. If they are about tasks, trial one task board.
3. Start the AI side with one job: meeting minutes. Do it for a week and measure how long it takes you now.
4. Add the next tool or the next AI job only when the first one sticks.

FAQ

What is the best project management software for architects?

There is no single winner. Architecture-specific tools such as Monograph, BQE CORE and Deltek Ajera focus on phases, fees and billing. General task tools such as Asana or Basecamp focus on coordination. Pick by the problem you have today, and trial before you pay for a year.

Do small architecture studios need practice management software?

Not always. A solo architect can run on a task board and a simple time tracker. The need grows once you bill by phase, have staff to schedule and want to see profit per project.

Can ChatGPT replace project management software?

No. A chatbot does not store your budgets, timesheets or invoices. It is useful for writing minutes, checklists and emails around them.

Is it safe to paste project details into an AI chatbot?

That depends on your client contracts and the provider’s data settings. Remove names, addresses and anything confidential, and check the terms before you use real project data.

Your next step

Pick one admin job you repeat every week and hand the first draft to AI. If you want help with that, join the free Launchpad and start with a short, practical lesson.