The asset you should not risk
Ask a service business owner which asset they would least like to lose and the answer is rarely a van. It is the phone number.
That number is printed on vehicles, etched into yard signs, listed on a Google Business Profile with years of reviews attached, saved in thousands of customers' contacts under your business name, sitting in directory listings you no longer control, and embedded in a decade of word-of-mouth referrals. Some of those surfaces you can update in an afternoon. Most of them you cannot update at all, because they live in other people's phones.
So when a service business considers adding an AI receptionist, the first serious question is almost never about the AI. It is: do I have to give up my number?
The answer is no. As of September 2026, there are two established ways to keep it, with different risk profiles and different end states, and the choice between them is genuinely a judgement call rather than a technicality. This guide covers both, the risks specific to each, how tracking numbers fit alongside your main line, and exactly what to test when the change goes live. The platform referenced is Run with Jarvis.
Two routes, one customer-facing result
Forwarding. Your number stays exactly where it is, with your existing carrier, on your existing account and bill. You set a forwarding rule so inbound calls to that number are sent to the answering platform. The customer dials the same number they always have and gets answered by the receptionist. Nothing about ownership changes.
Porting. The number itself moves from your current provider to the new one. After the move, the number lives with the platform that answers it. Your old provider no longer carries it, and you stop paying them for that line.
Both produce the same experience for the caller. The difference is everything behind it.
Forwarding is reversible in minutes. If you decide against it, you change the forwarding rule back in your carrier portal and the world is as it was. Nothing was moved, nothing was risked, and no third party had to approve anything. That reversibility is the whole argument for it.
Porting is cleaner as an end state. The number lives where it is answered, there is one provider instead of two, one bill instead of two, and no forwarding rule sitting in a portal that somebody could change by accident — which, for what it is worth, is one of the more common causes of a business quietly losing its calls.
What forwarding costs you is a small amount of clarity and a dependency. You are paying the old carrier for a line that now does nothing except pass calls onward, and the forwarding rule is a link in the chain that can break or be misconfigured. What porting costs you is the irreversibility of the move itself and a process you do not control the pace of.
The sequence most businesses should follow is therefore the obvious one: forward first, prove it on real calls, port later if you want the tidier end state. There is no operational penalty for staying on forwarding indefinitely, and some businesses do.
What a port involves, in honest general terms
It is worth being straightforward about the limits of what can be said here. Porting requirements and timelines are set by the carriers on both ends, vary by number type and account type, and change. Anyone — including a software vendor — who gives you a firm duration or a step-by-step without looking at your specific account is guessing. Get the specifics from your current provider and the receiving provider.
What is generally true, and worth knowing before you start, is this.
The request has to match the account. Carriers validate a port against the details they hold — typically the account name, the service address and the authorised contact. If the request says anything different from their records, it fails validation, and the failure usually arrives as an unexplained rejection rather than a helpful explanation. This is the single most common cause of a delayed port, and it is entirely avoidable by checking your own account details before submitting rather than from memory.
The old account must stay open until it completes. A number is attached to an account. Closing the account, or letting it lapse for non-payment, can release the number — and a released number can be gone permanently. This is the risk that actually matters.
The timeline is not yours. It involves at least two carriers and their processes. Plan around a window rather than a date, and do not schedule anything that depends on the port completing on a particular morning.
Some numbers are more complicated than others. Numbers that are part of a bundle, tied to other services, recently transferred, or on certain account types can carry extra conditions. Ask rather than assume.
Number portability itself is regulated in the United States, and the Federal Communications Commission at fcc.gov is the authority on those rules. That is useful context, not an operational substitute for talking to your carrier.
| Forwarding | Porting | |
|---|---|---|
| Where the number lives | Stays with your current carrier | Moves to the new platform |
| Reversible | Yes, in minutes, by you | Not simply — it is another move |
| Third-party approval needed | No | Yes, both carriers |
| Ongoing cost | Two providers for one line | One provider |
| Main failure mode | Forwarding rule broken or changed | Account mismatch, or old account closed early |
| Good for | Starting, testing, low-risk trials | The settled end state |
| Effect on the caller | None | None |
The risk list, in the order things actually go wrong
Do not cancel the old service early. Whatever else you take from this guide, take this. The number is released from the account; close the account prematurely and you may lose the number. Cancel only after the move is confirmed complete and you have tested inbound calls on the new path.
Do not change account details mid-process. Renaming the business on the account, updating the service address, or switching the authorised contact while a port is in flight is a reliable way to fail validation. Do it before, or do it after.
Keep the billing current. A lapse for non-payment during a move is the worst possible timing. This sounds trivial and is exactly the kind of thing that happens when a card expires.
Write down what is published where. Before changing anything, list every surface your number appears on: website, Google Business Profile, vehicle wraps, yard signs, directories, review sites, paid ads, invoices, email signatures, social profiles. You are not necessarily changing them — the point of keeping the number is that you should not have to — but you need the list to test against, and you will find at least one surface with a different, older number on it. That discovery is worth the exercise on its own. Local-profile specifics are covered in the Google Business Profile guide.
Do not let two systems answer. A transitional state where the old phone system and the new platform are both live on the same number produces unpredictable behaviour — sometimes one answers, sometimes the other, and nobody can reproduce it. Decide which one answers and make the other passive.
Remember texts. Voice and messaging are not automatically the same path. A number that forwards calls correctly can silently drop texts, and customers increasingly reply to a call with a text. Test messaging explicitly.
The same general migration discipline applies here as when changing any system of record — the risks cluster around the handover rather than the destination, which is the theme of the field service software migration guide.
Tracking numbers are a different thing entirely
This is the distinction that causes the most confusion, and getting it wrong causes real attribution damage.
Your main published number is an identity. It is on the truck, the profile, the signage and the customers' phones. It should be treated as permanent, moved rarely and carefully, and never used as a marketing experiment.
A tracking number is an instrument. It is a number assigned to a specific channel — a paid campaign, a particular landing page, a directory listing — so that a call arriving on it tells you where it came from. Dynamic Number Insertion swaps the number shown on a web page based on how the visitor arrived, which is what makes per-channel attribution possible. Tracking numbers are meant to be created, rotated and retired freely.
Two rules follow. Never put a tracking number on a permanent surface — a vehicle wrap or an etched sign — because retiring it later orphans that surface. And never route your main number through a temporary tracking setup, because the day that setup is dismantled your identity number stops working.
There is also a reporting consequence worth naming: your main number will always be your largest single "channel", and it is unattributable by nature, because it collects referrals, repeat customers, vehicle sightings and anyone who saved you years ago. That is not a measurement failure to be fixed; it is the correct behaviour of a brand number. What tracking numbers can and cannot tell you is in the call tracking and attribution guide, and whether you need them alongside an AI receptionist at all is worked through in the call tracking with an AI receptionist guide.
On Run with Jarvis, DNI and channel attribution sit in the Pro tier and above, while the answering, booking and dispatch that your main number needs are in Core — so a business that only wants its existing number answered well does not need the attribution layer to get there. Tier contents are on the pricing page.
Cutover day: what to test
A short, unglamorous checklist that finds almost everything.
From a phone not on your account, from outside your building, dial the main number. Let it run without intervening. Confirm it is answered, and answered by what you expect.
Check the call produced a record. A call that was handled but left no trace is a call you cannot review, count or recover. This is the check most people skip.
Text the number. Confirm the message lands somewhere a human will see it. Then reply from that side and confirm it reaches the phone you texted from.
Make an outbound call and send an outbound text. Confirm the number displayed to the recipient is the one you want customers to see and call back. A new setup that sends from an unexpected number trains customers to save the wrong one.
Test every other published number. The tracking numbers, the old second line, the number on a five-year-old directory listing. These are where breakage hides, precisely because nobody dials them.
Confirm out-of-hours behaviour. Ring it in the evening. After-hours is when most service businesses lose calls, and it is the path least likely to have been tested.
Keep the results written down with the date. That record becomes the baseline for the next time somebody asks whether something changed. What the first weeks after a change should look like more broadly — including what to watch and what to fix — is covered in the first 30 days onboarding guide.
What this is actually worth
It is easy to treat number continuity as a technical detail. It is closer to the centre of the decision than that.
The reason a service business can add an answering layer without a marketing cost is precisely that the number does not change. No re-wrapping vehicles, no re-printing signage, no updating a profile with reviews attached to it, no stranded referrals, no customers dialling a dead line. The asset keeps working and only the thing behind it improves — which is the same logic that makes the economics work at all, set out in the AI employee versus hiring cost math.
The opposite decision — taking a new number because it is simpler to set up — quietly costs a share of your inbound volume for years, and the loss never appears anywhere as a line item. It shows up as calls that stopped arriving from sources nobody was tracking.
What to automate, what to keep human
Automate the answering, the record, the notifications and the testing reminders. Keep human the decisions: whether to forward or port, when to cancel the old service, which surfaces get updated, and who owns the process end to end. Those are twenty-minute decisions with multi-year consequences, and they need a person who knows what is published where.
One structural note: if the answering, the booking, the messaging and the attribution live in one system, then a number change is one change to verify. If they are four separate tools, the cutover has four places to break and four logins to remember — which is the practical version of the argument in the all-in-one versus point solutions comparison.
Where to start
Start by forwarding, because it is reversible and it answers the only question that matters — does this work on your real calls, with your real customers, in your real trade — without risking the asset.
Before you change anything, list every surface your number appears on and check your carrier account details against what you believe them to be. Then forward, then run the cutover checklist, then leave it alone for a few weeks and read the call records.
If it is working and you want the tidier end state, talk to both carriers about porting then, with no time pressure and nothing depending on the date. Plans and minutes are on the pricing page — three flat monthly tiers, no setup fee, month-to-month. To talk through your specific setup before touching anything, get in touch.
