Persuasion and advocacy, one real conversation at a time.
Eli Works powers conversational AI for persuasion and advocacy. Two products on one platform: Eli, for campaigns talking with voters, and Ellie, for causes and organizations talking with people.
A voter taps how they feel about the candidate. Eli has a short conversation with them, answers from the campaign's own approved record, and makes the candidate's case.
Talk with EliEllie speaks for an institution. She listens to one person's worry, answers from the sponsor's own evidence, and names her source. She never argues.
Talk with EllieEli, for campaigns
Eli is for political campaigns and the people who run and advise them. The voter sees one question, such as how they feel about the candidate, and taps one of a few choices, from strongly opposed to declared voter. That tap starts a short conversation: Eli asks why, listens, and answers from the campaign's approved messaging.
The campaign gets voters placing themselves by support level, a record of who moves, why and when, files of voters by support level for follow-up, and what voters said, clustered. Eli records movement only when a voter says they have moved.
On research: Cornell-led researchers reported on 4 December 2025, in two papers published in Nature and in Science, that in the 2024 United States election an AI model arguing for one candidate moved likely supporters of the other 3.9 points on a 100-point scale, an effect the Cornell Chronicle called roughly four times larger than traditional ads tested in 2016 and 2020. That is a finding about AI conversations in general, not a measurement of Eli. The same research found that the more persuasive a model was, the less accurate its information, which is why ours answers only from a client's own approved record.
Ellie, for causes and organizations
Ellie is for associations, coalitions, unions, companies and institutions that need many personal conversations on a subject where being wrong has a cost.
- She listens first.
- She answers from the sponsor's evidence, in plain words, with the source named.
- She understands first, then advocates. She never argues.
- She says "I don't know" when the record is silent.
- She helps supporters act, for example by helping someone write a public comment in their own words.
Eli came first, and Ellie followed. Neither is ever finished: we learn every day how to improve our platforms.
The difference
Eli persuades. He makes a candidate's case to a voter, and it may end with a step the voter can take, such as registering to vote or volunteering. Ellie advocates. She speaks for an institution with evidence and empathy, and her work is year-round, with no election date. Both run on the same platform, and both answer only from the client's own record.
The Brain
For each client we build a Brain: the client's own approved record, such as its reports, studies, filings, positions and messaging. Eli or Ellie answers only from it. Nothing goes in that the client did not approve. When the record is silent she says so instead of guessing. Every answer shows where it came from, every word is logged, and every conversation shows what to add next.
How people reach her
Through a link: in a text, an email, a QR code on a mailer, a social post, or on the client's own website. One tap and they are talking. It is a web page, not an app. No login. Any hour, in English, Spanish and more than 30 other languages.
What a client gets
A weekly briefing: what people actually asked, in their own words, which worries are rising, and what the record could not yet answer. The client reads a conversation it was not in. The data belongs to the client and is never sold, shared or repurposed.
How it starts
With one real room. We can stand up clients in days, and the exact timeline depends on the complexity of building your organizational Brain. We build the Brain from the sources the client names, the client approves it, the client tests her, and then she goes out on the client's own channels.
What it is not
It is not a chatbot that makes things up. It answers only from the approved record and says so when the record is silent. It can still get something wrong, which is why every conversation is logged and the client can correct the record.
Talk with us
We would love to set you up with a partner to talk about what you are trying to do. Book a time with a partner, or write to [email protected], or ask Eli or Ellie yourself.