The question I get most
"Which AI agent should I get?" I hear some version of that question from small business owners most weeks, and it almost always comes with a screenshot of a product that promises to replace three employees.
My answer is usually a question back. Which job in your week, if it got done every single time, would you notice on your bank statement?
That's the frame for this piece. It isn't a vendor ranking. It's a list of ten jobs that an AI agent can do for a small business today, as of September 30, 2026, with two or three real products for each, the prices those products publish on their own pages (I checked every one today), the setup work, the legal lines and the part a person should keep.
Every price below links to the page I read. Prices change without notice, so screenshot the page on the day you sign.
Disclosure up front. Frenchy Digital builds custom AI agents, and where one of our product pages matches a job below, I link it. I also tell you, with arithmetic, why for most single jobs a $30 to $300 a month tool is the better buy than anything we build.
Agents are jobs, not products
The model most owners bring to this is shopping. There's a category called "AI agent," you pick the best one, and it does a bit of everything.
That model is wrong, and it's expensive. An AI agent is software that takes an input (a call, a chat, an email, a receipt), decides what to do with it inside rules you set, and does something: answers, books, drafts, files. The useful ones in 2026 are narrow. They do one job, inside one system, with one clear handoff to a person.
Think of it like hiring for a restaurant. You don't hire "a worker." You hire a host, a line cook and a dishwasher, and you know exactly what each one is not allowed to do (the dishwasher doesn't comp tables).
So the better model is a job description. For each agent below I'll tell you what the job is, what it costs to fill with an off-the-shelf tool, how long it takes to train, what can go wrong, and where the job description ends.
Therefore you should expect to buy two or three narrow tools, not one big one, and to spend more of your time on the rules than on the software. Here's the whole list on one screen before we go through it.
| Agent | Example tools checked | Entry price found | Main legal or trust risk |
|---|---|---|---|
| 1. AI receptionist | Upfirst, Smith.ai, Jobber Receptionist | $24.95/mo (Upfirst Starter) | Disclosure; outbound calls need TCPA consent |
| 2. Website chat support | Tidio Lyro, Intercom Fin | $32.50/mo (Lyro add-on) | Bot disclosure; wrong answers on policy |
| 3. Appointment scheduling | Calendly, HighLevel | $10/seat/mo (Calendly Standard) | Double bookings; reminder texts need consent |
| 4. Lead follow-up | HighLevel AI Employee, QuickBooks Customer Agent | $97/mo plus $50 add-on (HighLevel) | TCPA consent for texts and AI calls |
| 5. Review requests and replies | NiceJob, Google Business Profile | $0 (reply in GBP) to $125/mo | FTC fake review rule; Google gating ban |
| 6. Email inbox triage | Fyxer, Gemini in Gmail | $7/user/mo (Workspace Starter) | Full mailbox access; wrong sends |
| 7. Receipt and bookkeeping capture | Dext, QuickBooks agents | $25.21/mo (Dext Business, annual) | Miscategorized tax data |
| 8. Quote follow-up | Jobber Grow, HighLevel | $105/mo (Jobber Grow, annual) | Consent for texts; pricing promises |
| 9. Internal knowledge assistant | Claude Team, Gemini in Workspace | $20 to $25/seat/mo (Claude Team) | Confidential data in prompts |
| 10. Content drafting | Jasper, Claude Pro | $20/mo (Claude Pro) | False claims; undisclosed AI endorsements |
One more thing before the list. I refuse to print any vendor's resolution, deflection, accuracy or time-saved figure as fact. Several of the tools below publish one; I note them where they appear and explain why they don't count.
Agents 1 to 3: the front door
The first three agents answer the people who are already trying to reach you: by phone, on your website, or to book a time. For most service businesses this is where the money is, bc these are customers who have already decided to contact you.
Agent 1, the AI receptionist
What it does.Answers your business line when you can't, in a natural voice. It greets the caller, answers common questions from facts you give it (hours, service area, what you do and don't do), takes a message with the details you need, books a slot if it's connected to your calendar, and transfers or texts you when something is urgent.
Real products and prices, checked today. Upfirst publishes four plans: Starter at $24.95 a month for 30 calls ($1.50 per extra call), Premium at $59.95 for 90 calls, Pro at $159.95 for 300 calls and Scale at $299 for 600 calls, with lower rates on annual billing. Smith.ai offers a free AI receptionist tier with 25 calls a month and $3 per extra call, then Pro at $150 and Enterprise at $500 a month, with included calls varying by volume tier. If you already run a field service business on Jobber, its Receptionist add-on is listed at $29 a month on top of your plan.
The realistic setup work.Porting or forwarding the number takes minutes. The work is the one-page brief the agent reads: your services, prices you're willing to quote (if any), your service area, what counts as an emergency, who gets the transfer and when. Then call it yourself ten times with ten different problems. Budget about a week of evenings.
The risks. Inbound answering is the easy case. The hard case is when someone decides the receptionist should also call leads back. An AI-generated voice on an outbound call is an "artificial" voice under the Telephone Consumer Protection Act, per the FCC declaratory ruling adopted February 8, 2024 (summarized by Hunton Andrews Kurth), which means those calls generally need the called person's prior express consent. Also tell callers they're speaking with an AI at the start of the call. I'm not a lawyer; this is the map, not advice.
What a human keeps.Emergencies, upset callers, anything that sounds like a complaint, and any price or promise the brief doesn't already contain.
Our matching page. Our custom AI receptionist is for businesses whose call flow the tools above can't follow. For most, they can.
Agent 2, website chat support
What it does.Sits in the chat bubble on your website and answers questions from your help pages, policies and product catalogue. Good ones hand the conversation to a person when they don't know, rather than guessing.
Real products and prices. Tidio has a free plan and paid plans from $24.17 a month; its Lyro AI Agent is sold from $32.50 a month for 50 Lyro conversations. Tidio also says Lyro can "solve up to 67%" of customer problems. That's the vendor grading its own homework, so it doesn't go in any maths here. Intercom charges $0.99 per Fin outcome on top of seats that start at $29 per seat per month on Essential, $85 on Advanced and $132 on Expert.
Notice the two pricing shapes. Tidio sells a bundle of conversations; Intercom bills per outcome. Per-outcome pricing is fair when volume is low and can surprise you when it isn't: 1,000 Fin outcomes in a month is $990 before seats.
The realistic setup work.The agent is only as good as the pages it reads. Most small business websites have a thin FAQ and a returns policy written in 2019. Plan on writing or rewriting ten to twenty help articles before launch. It's the most useful afternoon of writing you'll do this year, with or without an agent.
The risks. California's bot disclosure law (Business and Professions Code section 17941) bars using a bot online to mislead someone about its artificial identity to influence a sale, unless the bot is disclosed clearly and conspicuously. Label it. The other risk is a confident wrong answer about a refund or warranty, which a customer will screenshot.
What a human keeps. Refunds, exceptions to policy, complaints, and any change to an order or account.
Our matching page. Custom customer support agent, for when support has to read and write in your own order or ticketing system.
Agent 3, appointment scheduling
What it does.Books, reminds, reschedules and fills gaps in your calendar without the back-and-forth. At the simple end it's a booking page with rules; at the agent end it reads an email or chat and proposes times itself.
Real products and prices. Calendly is free for one event type, $10 per seat a month on Standard and $16 on Teams, and lists an AI scheduling assistant called Callie as an add-on for Standard and Teams (its price wasn't on the page I read). HighLevel includes appointment scheduling alongside its Conversation AI and Voice AI in every plan, starting at $97 a month.
The realistic setup work. The software is the easy part. Write down your booking rules first: buffers between jobs, travel time, which services need a deposit, how far ahead people can book, and who is allowed to override. An agent enforces your rules exactly, including the bad ones.
The risks.Double bookings when a second calendar isn't connected, and reminder texts. Text reminders to customers should go only to people who gave you their number for that purpose.
What a human keeps. Deposits, cancellation fees, and any booking outside the rules.
Our matching page. AI appointment scheduler, for scheduling across several staff calendars and a system of record the tools don't connect to.
Agents 4 and 5: leads and reviews
The next two agents work on people who've already touched your business: the lead who filled in a form and the customer who just finished a job. Both are where the legal lines are sharpest, so read the risks twice.
Agent 4, lead qualification and follow-up
What it does.Replies within a minute to a new web form, text or ad lead, asks the two or three questions that tell you whether it's a fit, and either books a call or passes a short summary to you.
Real products and prices. HighLevel lists plans at $97, $297 and $497 a month, with AI Employee add-on plans at $50 to $97 a month. Intuit's QuickBooks Customer Agent, part of the agent set Intuit launched on July 1, 2025 per Accounting Today, manages leads and customer workflows inside QuickBooks Online, which starts at $38 a month at regular price on Intuit's pricing page.
The realistic setup work. Decide what a good lead looks like in writing: budget, location, timeline, service. Connect every lead source (site form, Facebook, Google) so nothing arrives outside the agent. Then write the handoff: what the summary to you says and where it lands.
The risks.This is the agent most likely to get a small business sued, bc it texts and calls. Texting a lead who asked to be contacted is one thing. Calling back with an AI voice needs prior express consent under the FCC's February 2024 ruling. And a purchased list is not consent. Keep outbound AI voice off until you've spoken with counsel about your consent language.
What a human keeps. Every outbound AI voice call decision, every quote, and every lead the agent marks as unsure.
A custom lead generation agent only makes sense when leads flow through systems no tool connects to; the approval step matters more than the software either way.
Agent 5, review requests and replies
What it does. After a job, sends the same polite review request to every customer, then drafts replies to the reviews that come in for you to approve.
Real products and prices. NiceJob lists Starter at $75 a month and Pro at $125 a month, with AI-powered review responses on Pro. You can also reply to reviews yourself for free inside your Google Business Profile, where Google checks replies before they post publicly, and draft them with any general assistant.
The realistic setup work. Connect your job or invoicing system so the request fires on completion, write one request message, and write three reply templates (happy, mixed, unhappy) for the agent to adapt.
The risks. Two rules make this the easiest agent to misuse. The FTC's final rule on fake reviews, announced August 14, 2024, bans fake reviews including AI-generated ones, bans incentives conditioned on positive or negative sentiment, and gives the FTC civil penalty authority. Separately, Google Maps policyprohibits offering incentives for reviews and selectively soliciting positive reviews. So no "were you happy? then leave us a review" filter, and never let an agent write a review.
What a human keeps.Every public reply, especially to a negative review. A reply is permanent and it's your name on it.
Agents 6 and 7: inbox and books
These two agents don't face customers at all. They give you back time, which is harder to put on a bank statement and easier to underestimate.
Agent 6, email inbox triage
What it does. Sorts incoming email into buckets, drafts replies in your voice, and surfaces the few messages that need you today.
Real products and prices. Fyxer has a free tier, then Pro at $32 per person a month on annual billing or $49 monthly, and Team at $64 or $99. If you're on Google Workspace, Gemini in Gmail is included from the Starter plan at $7 per user a month at regular price, with broader Gemini access on Standard at $14 and Plus at $22.
The realistic setup work. Low. Connect the mailbox, run in draft-only mode for two weeks, and correct its drafts so it learns your tone.
The risks. It reads everything: contracts, bank notices, employee issues. Read the vendor's data terms before you connect. And the classic small business fraud, the "our bank details have changed" email, is exactly the kind of message an eager assistant might answer politely. Keep the security basics from our AI agent security controls in place before you connect any inbox.
What a human keeps. Sending anything that commits money, price or scope, and every request to change payment details.
Our matching page. AI business assistant, for an assistant that works across email, calendar and your own systems together.
Agent 7, receipt and bookkeeping capture
What it does. Reads receipts and bills from photos or email, pulls out the vendor, date, amount and tax, suggests a category and pushes it into your accounting software.
Real products and prices. Dext's Business planis $25.21 a month billed at $302.50 a year, for 250 documents and 5 users. Inside QuickBooks Online, Intuit launched an Accounting Agent and a Payments Agent on July 1, 2025. Intuit says the Payments Agent gets businesses paid faster by a number of days; that's Intuit's own figure and I don't repeat it as fact.
The realistic setup work. Set up your chart of accounts properly first (or have your bookkeeper do it), then map your top 20 suppliers to categories so the agent has examples. Forward receipts to a capture address for a month and check every one.
The risks. A wrong category repeated 400 times is a tax problem. The capture is reliable; the judgement is where errors hide.
What a human keeps. Categorization review, reconciliation, paying bills and anything filed with a tax agency. Your accountant should see the books before year end regardless.
Agents 8 to 10: sales and knowledge
The last three agents help you sell and help your team know things. They're the most useful for businesses with a few employees and the least useful for a solo operator.
Agent 8, the sales assistant for quote follow-up
What it does. Chases quotes that went quiet. After a set number of days it sends a reminder by email or text, answers simple questions about the quote, and tells you who replied.
Real products and prices. Jobberincludes automated quote follow-ups by email or text on its Grow plan (from $105 a month billed annually) and Plus (from $280). HighLevel's workflow automation, from $97 a month, can run the same sequence for businesses outside field service.
The realistic setup work.Pick the cadence (for example two, five and ten days), write each message, and decide what the agent is allowed to say when someone asks for a discount. The answer should usually be "I'll ask the owner."
The risks. Texts need consent, and an agent that improvises a discount has just set your price. Keep the pricing out of its reach.
What a human keeps. Discounts, scope changes and the final yes.
Our matching page. AI sales assistant, for sales follow-up tied into your own CRM and quoting.
Agent 9, the internal knowledge assistant
What it does.Answers your staff's questions from your own documents: the procedures manual, product specs, price list, the way you like jobs written up. It stops the owner being the only person who knows where things are.
Real products and prices. Claude's Team plan lists standard seats at $20 to $25 a month, with Pro for an individual at $20 a month. Google Workspace Plus, at $22 per user a month, extends Gemini across Docs, Drive and the rest of Workspace, which matters if your documents already live there.
The realistic setup work.The hard part is that most small businesses don't have the documents. Writing down the ten procedures people ask you about most is the real project. Once they exist, loading them takes an hour.
The risks. Staff pasting customer files, payroll or medical information into a tool whose data terms nobody read. Decide what goes in before anyone starts.
What a human keeps. Anything customer facing, and HR or payroll questions.
Agent 10, marketing and social content drafting
What it does. Drafts social posts, email newsletters, service page updates and ad variations from your notes and past content.
Real products and prices. Jasper Pro is $69 per seat a month, or $59 on annual billing, with marketing agents and brand voices. A general assistant such as Claude Pro at $20 a month does most of what a small business needs if you give it good examples.
The realistic setup work.Collect ten pieces of your best past writing and a one-paragraph description of your voice. Without those, you'll get the same bland copy as everybody else.
The risks.Claims you can't back up ("best in town," "guaranteed"), invented statistics, and fake testimonials, which the same FTC rule covers. Every fact in a post should be one you could prove.
What a human keeps. Final approval of everything that publishes.
Which one first
Start with the agent where one missed task is a sale you can count, then measure it for 30 days before buying anything else.
Let's do the arithmetic out loud, bc it's simpler than it looks. Here's what it would cost to fill all ten jobs with the cheapest sensible tool I priced today, one seat each.
| Job | Cheapest sensible tool we priced | Monthly |
|---|---|---|
| Receptionist | Upfirst Premium (90 calls) | $59.95 |
| Website chat | Tidio Lyro AI Agent | $32.50 |
| Scheduling | Calendly Standard, one seat | $10.00 |
| Lead follow-up | HighLevel Starter | $97.00 |
| Reviews | NiceJob Pro | $125.00 |
| Email triage | Fyxer Pro, annual rate | $32.00 |
| Bookkeeping capture | Dext Business, annual rate | $25.21 |
| Quote follow-up | Jobber Grow, annual rate | $105.00 |
| Knowledge assistant | Claude Team, one standard seat (top of range) | $25.00 |
| Content drafting | Jasper Pro, annual rate | $59.00 |
| Total | All ten, one seat each | About $570.66 |
About $571 a month, or roughly $6,848 a year. That's the ceiling, and nobody should buy it. Several of these tools overlap (HighLevel already includes scheduling, Jobber sells a receptionist add-on), and most small businesses need two or three jobs filled, not ten.
Now the break-even, which is the part that matters. Consider a plumbing business where a typical job is worth about $400 in gross profit. This is a scenario, not a client. Upfirst Premium costs $59.95 a month. Divide: $59.95 over $400 is about 0.15. The receptionist pays for itself if it saves roughly one job every six or seven months.
Compare email triage for the same owner. Fyxer Pro at $32 a month pays for itself if it saves about 40 minutes a month valued at $50 an hour ($32 over $50 is 0.64 hours). Probably true. But you'll never see those 40 minutes on a bank statement, which is why I put it later.
So the order I'd suggest for most service businesses is:
- 1.The phone or the quote, whichever you drop more often. Receptionist if calls go to voicemail; quote follow-up if you send quotes and hear nothing back.
- 2.Scheduling, if booking still happens by phone tag.
- 3.Review requests, sent to every customer, with you approving replies.
- 4.Then the back office: inbox triage and receipt capture.
- 5.Website chat, the knowledge assistant and content drafting last, unless your business lives online.
An online store flips this list: chat support first, bc that's where its customers ask questions. The rule doesn't change, only the answer.
Why one at a time? Because of a concept economists call the confounding variable. If you switch on a receptionist, a review tool and a follow-up sequence in the same month and sales go up, you have no idea which one did it, or whether it was the weather. You'll keep paying for all three forever out of fear of turning off the one that worked.
It's the same reason a doctor changes one medication at a time. Not bc two changes are dangerous, but bc two changes make the result unreadable.
What 30 days of measuring looks like
Consider a two-van landscaping company that picks the receptionist first. This is a scenario to show the method, not a client story.
Week zero, before anything is switched on, the owner counts missed calls from the phone log: say 14 in five working days. That number is the baseline. Nothing the vendor says replaces it.
Weeks one to four, the agent answers every call the owner doesn't. Each Friday the owner reads the call summaries and marks three things: calls that were real new work, calls that booked, and calls the agent handled badly. The last column is the one people skip, and it's the one that tells you whether to keep going.
At day 30 the question is not "did it work?" in general. It's narrower: did it book at least one job I would otherwise have lost, and did it embarrass me more than once? If the first answer is yes and the second is no, it stays and you pick agent number two. If not, cancel it. Most tools on this list publish monthly billing, a free tier or a free trial, so the downside is roughly one month's fee.
That's the asymmetric bet in plain terms. The most you lose is roughly $25 to $160 and a few evenings of setup. The upside is a job or two a month that used to go to voicemail. Few decisions in a small business are that lopsided.
For a fuller breakdown of usage costs and what drives them, our AI agent cost guide goes line by line.
When custom makes sense
For one simple job, a custom build almost never makes sense. I'll show you why with our own prices.
A Frenchy Digital custom agent is $5,000 for the agent plus $5,000 setup, so $10,000 to start, and more if it needs many integrations. After launch it's $2,500 a month. You pay the usage (model tokens, telephony, third-party APIs) directly at the providers' rates. Year one comes to $10,000 plus 12 times $2,500, which is $40,000 plus usage. Each later year is $30,000 plus usage.
Now divide. Upfirst Pro at $159.95 a month is $1,919.40 a year. $40,000 over $1,919.40 is about 20.8. Our build would have to be worth nearly 21 times what that tool is worth to you in year one. For answering calls and taking messages, it isn't. I'd tell you to buy the tool.
Try it the other way round. At $2,500 a month after launch, a custom agent has to produce about $30,000 a year of value just to cover its running fee, before the $10,000 start and before usage. For a business that nets, say, $400 a job, that's 75 extra jobs a year, or about six a month, that would not have happened otherwise. For a phone line that misses a few calls a week, that's not a realistic bar.
Even the whole ten-tool stack above, at about $6,848 a year, is roughly a sixth of our year one.
To be clear, that's not modesty. It's the maths. A custom build earns its price in three situations:
- The workflow crosses several of your own systems: for example, a call that checks inventory, quotes from your own pricing rules, books a crew and writes back to a system no tool connects to.
- The rules are the product: your intake order, approval steps or compliance checks are specific enough that a configurable tool can't follow them.
- You need to own it: we transfer full source code and IP, so the agent is an asset of the business rather than a subscription.
Our only real agent case study is the Beyond Points AI build, where a browser agent transfers points and books travel with every irreversible step (transfer, book, pay, cancel) behind a user confirmation gate. That's the shape of work where custom makes sense: real money, several systems, and rules that had to be exact. If your job looks like that, our AI agent creation service is the place to start. If it doesn't, the custom versus SaaS decision guide will probably confirm you should buy.
What a human keeps
Every agent on this list can prepare work. A person approves anything that spends money, sets a price, gives advice or goes public.Here's the boundary in one table.
| Agent | May act alone | Needs a person to approve | Never touches |
|---|---|---|---|
| Receptionist | Answer, take details, book open slots | Emergencies, angry callers, pricing questions | Legal, medical or financial advice |
| Chat support | Answer from your published policies | Refunds, exceptions, complaints | Changing an order or account without a person |
| Scheduling | Book, remind, reschedule inside rules | Overbooking, special requests | Deposits and cancellation fees |
| Lead follow-up | Reply to inbound inquiries | Any outbound AI voice call | Contacting people who never consented |
| Reviews | Send the same request to every customer | Every public reply | Writing, buying or filtering reviews |
| Email triage | Sort, label, draft | Sending anything that commits you | Payments and bank detail changes |
| Bookkeeping | Capture receipts, suggest categories | Categorization and reconciliation | Filing taxes or paying bills |
| Quote follow-up | Send scheduled reminders | Discounts and scope changes | New prices you did not set |
| Knowledge assistant | Answer staff questions from your documents | Anything customer facing | HR, payroll and client files it was not given |
| Content drafting | Draft posts and emails | Everything before it publishes | Testimonials, reviews, fake endorsements |
If you're unsure where a new task sits, ask one question: if the agent gets this wrong, can I undo it by tomorrow without anyone outside the business noticing? If not, a person approves it.
Three rules come up across this list more than any others. I'm not a lawyer and this isn't legal advice, but these are the ones to raise with yours.
- Outbound AI voice calls need consent: The FCC ruled on February 8, 2024 that AI-generated voices are artificial voices under the TCPA, so calls using them generally need prior express consent. This touches agents 1, 4 and 8 the moment they call out.
- No fake, filtered or bought reviews: The FTC's final rule announced August 14, 2024 bans fake and AI-generated reviews and sentiment-conditioned incentives, with civil penalties. Google Maps policy also bans review incentives and selective solicitation. This touches agents 5 and 10.
- Say it's a bot: California's section 17941 requires clear and conspicuous disclosure when a bot interacts online to influence a sale. Other states have their own rules. Disclose at the start of every chat and call, whatever state you're in.
Red flags
Walk away from any tool or agency that does one of these.
- It promises to replace staff, or quotes a resolution or accuracy rate with no way for you to check it on your own traffic.
- It offers to make outbound AI voice calls to a lead list without asking where your consent came from.
- It offers review filtering, review writing, or rewards for five-star reviews.
- It hides pricing until a demo and then quotes a per-outcome price with no cap.
- It won't show you its data terms before you connect your inbox, phone or books.
- It has no clear way for the agent to hand off to a person.
- An agency quotes a build cost range with no stated method, or cites the untraceable '95% of AI pilots fail' figure to sell you a rescue.
Limitations
Every price here comes from each vendor's own pricing page on September 30, 2026. Several pages show introductory discounts (QuickBooks and Google Workspace both did today); I used regular prices where the page gave them. Jobber, Dext, Fyxer and Jasper figures marked annual are annual-billing rates.
I did not run a trial of any product for this article. Nothing here measures quality, only what each vendor publishes. I chose two or three tools per job to show the price range; many good tools aren't named, and naming one is not an endorsement.
The FCC ruling is summarized from a law firm's write-up rather than the order itself. The QuickBooks agent launch date comes from Accounting Today's report. Calendly's Callie add-on price wasn't shown on the page I read. I also couldn't load OpenAI's pricing page, so ChatGPT's business plans aren't priced here.
The break-even examples are scenarios with made-up job values, framed as such. Plug in your own numbers.
Numbers I refused to print.
Tidio's "solve up to 67%" for Lyro, and Intuit's claims that its Payments Agent gets businesses paid days faster and that its agents save hours a month. All three are the seller measuring itself.
The "95% of generative AI pilots fail" figure, the "85% of AI projects fail" line and "AI cuts support costs by 30%." They show up in almost every article like this one. The first rests on a narrow study of custom enterprise tools, the second misquotes a 2018 prediction about something else, and the third has no traceable source. None applies to a plumber buying a $60 receptionist.
Any missed-call or call-abandonment benchmark. They're everywhere in receptionist marketing and I couldn't trace one to a method I'd trust. Count your own missed calls for a week instead; your phone bill already has the number.
Three things this week
You can pick and test your first agent in a week.
- 1.Count what you drop. For five working days, tally missed calls, unanswered quotes and unanswered website messages. The biggest number is your first agent.
- 2.Write the one-page brief for that job: facts the agent may state, what it must hand to you, and what it must never say. That page is also your test script.
- 3.Start a trial of one tool from this list, disclose the AI to customers, run it on real traffic for 30 days, and compare the tally to week one.
If the tool can't follow your brief after 30 days, that's the moment to talk about custom. Until then, buy the $60 thing.
Want One Agent Built Around Your Workflow?
Book a discovery call with Frenchy Digital, a senior-led Black-owned Los Angeles agency. We map the job, the systems it touches and the approvals a person keeps, and send a fixed-price phased proposal within 5 business days.
Want One Agent Built Around Your Business?
Book a discovery call. We map the workflow, the systems it touches and what a person must approve, then send a fixed-price phased proposal within 5 business days.
1517 S Bentley Ave Apt 204, Los Angeles CA 90025
Frequently Asked Questions
Sources & References
- 1Upfirst: pricing↗
- 2Smith.ai: AI Receptionist pricing↗
- 3Jobber: pricing, Receptionist add-on and quote follow-ups↗
- 4Tidio: pricing and Lyro AI Agent↗
- 5Intercom: pricing and Fin AI Agent↗
- 6Calendly: pricing↗
- 7HighLevel: pricing and AI Employee add-on↗
- 8NiceJob: pricing↗
- 9Fyxer: pricing↗
- 10Google Workspace: pricing and Gemini↗
- 11Dext: US pricing↗
- 12QuickBooks Online: pricing↗
- 13Accounting Today: Intuit debuts AI agents for QuickBooks↗
- 14Claude: pricing↗
- 15Jasper: pricing↗
- 16Hunton Andrews Kurth: FCC declaratory ruling that the TCPA applies to AI-generated voice calls↗
- 17FTC: final rule banning fake reviews and testimonials↗
- 18Google Maps: user contributed content policy on fake engagement↗
- 19Google Business Profile Help: manage and reply to reviews↗
- 20California Business and Professions Code, sections 17940 to 17943 (bot disclosure)↗
