Key facts
- Client
- A truck dispatch service run from India, serving US and Canada carriers.
- Goal
- At least 1 new active, paying dispatch client per month.
- Main rule
- Optimize for signed carriers, not for the number of calls.
- Channels
- AI voice calls (only with consent), email, and human transfer.
- Expected
- 1 to 2 signed carriers per month from a list of about 1,000 carriers (planning assumption, not a result).
- Approval
- Owner approval is needed before any agreement is sent.
- First step
- A prototype in about four weeks, before a full build.
Problem
What did the business ask for?
The business asked for an "AI client hunting agent" that works from its existing carrier list. It wanted one new paying dispatch client each month, run fully from India.
The business told us:
- They dispatch for owner-operators, small carriers and truck owners.
- They work with all common truck types: dry van, reefer, flatbed, step deck, power only, box truck and hotshot.
- They already have a carrier list and a CRM / dispatch software.
- They want AI voice calls, personal emails and automatic follow-ups for "no answer" and "I need more time".
- They want help with onboarding and dispatch agreement signing, with owner approval.
- They want every call, email, follow-up and conversion tracked.
- They want a working prototype before they commit.
How it works
How does an AI voice agent for truck dispatch work?
An AI voice agent for truck dispatch is software that calls carriers, talks in natural speech, answers simple questions and books the next step. Read more about what AI voice calling agents can do.
Our proposed design has six steps:
- Clean the list. Remove duplicates, check each MC/DOT number with FMCSA, and remove opt-outs and do-not-call numbers.
- Pick the channel. Carriers with recorded consent can get an AI call. All others start with email, or a call from a human dispatcher.
- Reach out. Send a personal email or make the call.
- Qualify. Ask the checklist questions below and score the carrier.
- Hand over. Send hot leads to the owner, with a short summary.
- Onboard. Collect documents, prepare the agreement, get owner approval, send for e-signature.
Every action goes into the CRM with a time, a channel and a result.
What does the AI say on the call?
The AI says it is an AI assistant in the first sentence. It offers a human at any time and stops calling at once if the person says no.
Sample opening script (draft):
"Hi, this is Ava, an AI assistant calling for Sharma Dispatch Services. Am I speaking with someone from [carrier name]?
I'm calling because you asked us to get in touch about dispatch help. This call may be recorded.
You can say 'person' at any time to talk to our team. If you don't want calls from us, just say 'stop' and we will not call again.
Do you have two minutes? I'd like to ask what truck you run and which lanes you like."
Call rules in the design:
- Disclose "AI assistant" at the start. Never claim to be human.
- Human transfer on request. If no human is free, book a callback.
- "Stop", "remove me" or similar: end the call, mark do-not-contact in the CRM, and block all channels.
- No pricing promises beyond the approved rate sheet.
- After the call: a short summary, the next step and a follow-up date go into the CRM.
Which carriers count as qualified?
A qualified carrier has active authority, a truck type and lanes the team can serve, and a real need for a dispatcher. The agent checks these points before it sends a lead to the owner.
| Check | How the agent checks it |
|---|---|
| MC / DOT number is valid | FMCSA QCMobile API (free web key) or SAFER lookup |
| Operating status | "Allowed to operate" and authority status from FMCSA data |
| Authority age | Date from FMCSA records. New authorities (under about 90 days) may need different broker setup. |
| Truck type and count | Asked on the call or email: dry van, reefer, flatbed, step deck, power only, box truck, hotshot |
| Lanes | Asked: home base, preferred regions, US only or US + Canada |
| Current dispatcher | Asked: none, unhappy, or happy (happy = follow up later) |
| Paperwork ready | Asked: insurance in place, factoring company or not |
FMCSA data is public. But a public phone number is not consent to call with an AI voice.
What is the email and follow-up plan?
The agent follows a fixed plan for each carrier, so no lead is forgotten. When a carrier says "I need more time", the agent asks for a date and follows up on that date.
| Day | Channel | Message |
|---|---|---|
| 0 | Short intro: who we are, the truck types and lanes we cover, one clear question | |
| 1 | AI call (consent only) or human call | Intro call using the script above |
| 3 | Answer the most common question (fees, how loads are found) | |
| 7 | AI call (consent only) or human call | Second try at a different time of day |
| 14 | "Is now a bad time?" message with a one-click reply | |
| 30 | Short check-in, then pause | |
| 90 | Re-start only if the carrier did not opt out |
Each email has the sender's name, a real postal address and a working unsubscribe link. Emails come from a separate, warmed-up sending domain to protect the main domain. This is a standard step in AI lead generation.
Expected outcomes
How many carriers do you need to sign one per month?
Expected outcomes
In our planning model, a list of about 1,000 carriers gives an expected 1 to 2 signed carriers per month. Every rate below is an assumption to test in the prototype.
| Stage | Assumed rate | Expected count |
|---|---|---|
| Carriers on the list | (assumption) | 1,000 |
| Valid after FMCSA check and opt-out / DNC removal | 60% | 600 |
| Reached (real conversation or email reply) in the month | 15% | 90 |
| Interested in dispatch help | 25% | about 22 |
| Qualified (authority, truck, lanes, need) | 40% | about 9 |
| Signed and first load booked | 15–20% | about 1–2 |
Assumptions: The list is fairly recent. A good part of the list has given consent for calls. The offer and fee are competitive. The owner calls hot leads back within the same day. If the reach rate is lower, the plan needs more carriers on the list, not more calls to the same people.
How does onboarding and e-signing work?
The agent collects the documents and prepares the dispatch agreement. The owner approves it, and only then does the agent send it for e-signature.
Documents the agent collects:
- MC authority letter
- W-9 form
- Certificate of insurance
- Notice of assignment (NOA) and factoring details, if the carrier uses factoring
Steps: the agent checks that each file is present, prepares the agreement from the owner's template, and sends the owner a summary to approve. After approval, it sends the e-sign request, tracks the signature and marks the carrier "Active" in the CRM.
How is it run from India?
The agent runs in the cloud, so the owner can manage everything from India. It uses US and Canada phone numbers and only calls inside the carrier's local calling hours.
- Time zones: US Eastern is 9.5 to 10.5 hours behind India, and Pacific is 12.5 to 13.5 hours behind. The agent plans calls by the carrier's own time zone.
- Calling hours: US calls stay between 8 a.m. and 9 p.m. local time (TCPA rule). Canada uses CRTC hours: 9 a.m. to 9:30 p.m. on weekdays and 10 a.m. to 6 p.m. on weekends.
- Numbers: local US and Canada numbers with caller ID that matches the business. Calls go through a carrier that signs calls under STIR/SHAKEN.
- Owner view: a morning summary in India time with hot leads, documents waiting and agreements to approve.
What US and Canada rules apply?
AI voice calls and cold emails are regulated in both countries. The design follows these rules by default. This is not legal advice. The client should confirm the plan with a US and a Canadian lawyer.
United States
- TCPA and AI voices: In February 2024, the FCC ruled that AI-generated voices are "artificial" voices under the TCPA (FCC press release). Calls to mobile numbers with an artificial voice need prior express consent. Many owner-operators use a mobile phone as their business phone.
- Do Not Call: The FTC says most business-to-business calls are exempt from the National Do Not Call Registry (FTC TSR guide). But a one-truck owner-operator on a personal mobile is a grey area, and some states have their own rules. The design scrubs against the DNC list and keeps an internal do-not-call list.
- Caller ID: The FCC requires STIR/SHAKEN caller ID authentication on IP networks (FCC call authentication).
- Email: CAN-SPAM covers business email too. It needs honest headers and subject lines, a postal address and an unsubscribe that is honored within 10 business days (FTC CAN-SPAM guide).
- SMS: if SMS is added later, it needs A2P 10DLC registration and consent.
Canada
- CRTC rules and National DNCL: Telemarketers must follow the Unsolicited Telecommunications Rules, including calling hours and an internal do-not-call list. The National DNCL does not apply to calls to businesses, but the internal list still does (CRTC rules). Automatic calls with a recorded or artificial voice for selling need express consent under these rules.
- CASL (email): Commercial emails to Canadian addresses need express or implied consent, sender details and an unsubscribe (Government of Canada CASL consent guide). CASL applies even when the sender is outside Canada.
What this means in the design: AI voice calls only go to numbers with recorded consent. Everyone else gets email (where allowed) or a human call. Every opt-out blocks all channels at once.
What does it cost to run?
Running costs come from a few usage-based parts. We give exact numbers after the prototype, when real call minutes and email volume are known.
| Cost part | What it pays for |
|---|---|
| Phone numbers and call minutes | US and Canada numbers, outbound minutes |
| Voice AI | Speech-to-text, the AI model and text-to-speech, per minute |
| AI model use | Writing emails, summaries and scoring |
| Email sending | Sending tool, extra domains and warm-up |
| E-signature tool | Sending agreements for signature |
| CRM | Your existing tool (no new cost if it has an API) |
| FMCSA data | Free (QCMobile API web key) |
| Build and maintenance | Our fee: prototype, full build, monitoring and tuning |
What will this agent not do?
The agent will not call anyone without consent, pretend to be human, or send an agreement without owner approval. It also cannot promise a signed carrier.
- It does not buy or scrape new contact lists.
- It does not negotiate rates or book loads in this phase.
- It does not hide that it is an AI.
- It cannot make an uncompetitive offer win. Pricing and service quality still decide.
- Voice quality on poor mobile signal, and some strong accents, may cause errors. A human backup is always there.
Risks and open questions: how much of the current list has consent; the CRM's API access; how old the list is; and the lawyer review of the call plan.
What is the prototype-first plan?
We start with a small prototype on a consented part of the list, then decide together on the full build.
- Week 1 – Discover: review the list, CRM access, scripts and consent records.
- Week 2 – Design: set the call script, email plan, checklist and approval flow.
- Weeks 3–4 – Deploy prototype: run on about 100 to 200 carriers, with live CRM updates and a daily summary.
- Evolve: compare real rates with the model above, fix weak steps, then scale.