Counterfeit contact comes for the phone bank
“So if we’re calling a Chinese person that speaks Mandarin, we can have an AI that sounds Chinese over the phone and can speak Mandarin.”
That is Cameron Bonesso, president of an Ottawa campaign firm, talking to Canadian Affairs in March. He offered it as a selling point: a way to make voter contact “far better.” A synthetic voice, engineered to pass as the person on the other end of the line, so they pick up and stay on long enough to be counted. The campaign never has to find a single Mandarin-speaking volunteer. It only has to buy the voice.
Nothing in the article treats that as a strange thing to say.
The article it appears in is headlined “As volunteers vanish, Canadian political campaigns turn to AI.” Vanish. As though volunteers were a migratory bird, or a glacier: something that leaves on its own schedule while we watch, regretful but powerless.
I’ve spent more than ten years running campaigns that depend on volunteers. Volunteers do not vanish. A campaign loses them, or never earns them, and there is always a reason, and the reason is almost always the campaign.
Here is what the article gets wrong about its own subject. It treats the voter-ID call, the one where you phone thousands of people to ask who they’re voting for, as a data-collection task. Collect enough IDs, win the day. By that logic, a machine that collects IDs faster is pure upside.
That’s a key part of any campaign I’ve supported, but the reporting is missing the point. The call was never only about the ID. The call was the recruitment funnel. The volunteer dialling those numbers is being trained while they dial: learning the script, learning to hear a soft yes, learning that they can do this, watched by a lead who is deciding whether to ask them back, and then whether to ask them to run the next phone bank. The ID goes in the database. But the organizer actually gets made making the phone calls. The call runs both ways.
Automate the call and you keep the part the CRM can see. You lose the part it can’t. You stop growing a volunteer base. You stop teaching people how to talk about your message.
On the other hand, an AI dialler genuinely makes more calls per hour than a nervous volunteer on their first-ever shift. But a campaign that needs an AI dialler has already lost the thing the human phone bank was for. The efficiency is real. It is also the symptom. You cannot buy your way out of a problem whose definition is “we have stopped being able to ask people for help.”
Let me concede something the article never reaches for. A voter-contact call is a para-social encounter, and it always has been. A stranger picks up the phone; another stranger asks them a question they did not invite. The connection is brief and asymmetric; the volunteer will not remember this voter by Thursday. All of that is honest. It works because both people know what it is: a human took an evening, learned a script, and chose to spend it reaching toward people they will never meet. Even through the awkwardness, even through a no, the voter hears that someone bothered.
The AI call keeps the shape and removes the person. That is what turns an honest para-social call into a parasitic one: it harvests the feeling of someone reaching out while no one reaches. The deceit is the design, not a side effect. This is counterfeit contact: a call manufactured to be taken for the real thing. The Mandarin-speaking voice from the top of this piece is its sharpest form. Call that one voicejacking: counterfeit contact tuned to be mistaken for a particular person or identity, here a neighbour who shares the voter’s language. An authentic call asks for a moment of a voter’s attention and offers a real person in return. Counterfeit contact asks for the same and offers a performance of one.
The counter-example is recent, and I was inside it. I spent seven months on Avi Lewis’s NDP leadership campaign: interim manager first, then operations and communications, then operations through the closeout I’m still doing now. Avi made an early strategic choice to run on “big organizing”: push authority and resources down to local volunteers, equip them with templates and training, and let them act without waiting on head office. Over those seven months the campaign onboarded 8,400 volunteers, logged more than 450,000 phone calls across 6,500-plus caller-hours, and signed up members in 338 of 343 federal ridings. solidarity.tech, the relational-organizing CRM we ran on, held the records and sent the reminders. The calls were made by people.
The number that matters in that paragraph is 2,800: the relational organizers the campaign built, volunteers who got trained, asked back, handed a phone bank of their own, and who in turn pulled thousands more people in. You do not get 2,800 organizers from a faster dialler. You get them from a campaign that treats the phone call as the place an organizer is made. Counterfeit contact logs the call and makes no one.
Look at who Canadian Affairs asked.
Two of its three sources sell the thing the article is about. Bonesso runs a firm that sells voter contact; in the same piece he says it will be “rolling out” AI calling “in the next few months.” He is not a campaign that lost its volunteers and reached for a machine. He is the machine’s vendor. He has no volunteer shortage to grieve; he has a product to ship. Read the article with that in mind and it stops being reporting and becomes a sales call with a byline.
The second source is Jeff Ballingall, introduced as a “veteran political strategist,” who hands the article its tidy thesis: “authenticity is really going to be what’s gold here.” Ballingall founded Ontario Proud and Canada Proud. Those were never grassroots that thinned out. They were manufactured from the first post: Facebook pages built to look like a constituency, funded by real estate developers and construction firms, the spending routed back to his own company. He does not organize a base. He manufactures the appearance of one, for whoever is paying. In January, Canada’s National Observer reported that the same firm ran a campaign to save the CBC and Canada Proud’s campaign to defund it. Same shop, both sides, whoever pays. He is the article’s authority on authenticity.
A campaign that turns to an AI dialler has lost something real: the volunteers, the base, the reason it could once ask people for help. Bonesso and Ballingall never had that, and have stopped pretending to. They do not organize constituencies; they sell the look of one.
The third source, Chris Tenove, is a UBC democracy researcher with nothing to sell. He says micro-targeting has diminishing returns, that the real problem is “information pollution,” and (the last line of the article) that “we need more civil, face-to-face political engagement.” The one person in the room without a product gets the truest sentence, and it runs last, where closing lines go to be forgotten.
This is how astroturf gets normalized: through lazy sourcing, the kind no one has to sign their name to. Quote the vendors as experts, give the academic the caveats, and a product roadmap reads as an inevitable fact of nature: one more AI turn nobody can stop, and the question of whether anyone wants it never gets asked. Astroturfing used to be a scandal; treated as a trend, it becomes something to adjust to rather than object to. And what we would be adjusting to is larger than one article.
A democracy is legible in one specific, precious way: you can tell what people want by what they show up for. Doors knocked, rooms filled, calls made by people who chose to make them. That is the signal. Counterfeit contact jams it the way counterfeit money does. A forged bill does its damage long after it leaves your hand: once enough of it circulates, every bill is suspect, and people stop trusting the currency itself. A counterfeit call cheats the voter who picks up, and it teaches everyone that the voice on the phone might be bought, the canvasser might be a monthly subscription, the show of support might be an ad buy. The campaign loses the ability to know its own base. So does everyone watching. For whoever can afford the tooling, that illegibility is what the money is for.
One detail, and then I’ll leave the byline alone. The reporter who filed this, Sam Forster, is the author of a 2024 book for which he disguised himself as a Black man: dark foundation, an Afro wig, contact lenses over his blue eyes, hitchhiking across the United States to document racism. Black scholars called it dehumanizing. It is not my subject and I won’t pretend it is. But the instinct travels. Synthetic ethnicity as a way to get the story; voicejacking as a way to sell the call. The Mandarin-speaking AI went into his article without a raised eyebrow. By then it must have looked like method.
Canadian Affairs published this article with an audio version: a synthetic ElevenLabs voice that will read the whole thing to you, if you’d rather not read it yourself. A story about campaigns replacing human voices with manufactured ones, delivered, at your option, in a manufactured voice.
So here is the reframe. The volunteer did not vanish. The volunteer was never going to dial ten thousand numbers. A machine was always going to be cheaper at that. What a campaign loses, when it pays for counterfeit contact, is its reason to want the volunteer at all: the doorstep conversation, the nervous first shift, the person who came to make calls and stayed to run the riding. The phone bank was the campaign meeting the people it claims to speak for. Automate that and you have not found an efficiency. You have quietly changed what kind of thing you are: from a campaign, which needs a base, into an audience-acquisition funnel, which needs a budget. And an audience does not knock doors, or govern, or hold you to anything once you’ve won.