Questions and answers.
Everything Eli and Ellie say on our home page comes from this record. Ask them, or read it here.
What Eli Works is
What is Eli Works?
Eli Works powers conversational AI for persuasion and advocacy. It has two products on one platform: Eli, for campaigns talking with voters, and Ellie, for causes and organizations talking with people.
Who started Eli Works?
Eli Works was founded by political technology experts who saw campaigns spending heavily on digital advertising while missing direct conversations with voters.
Who do I talk to at Eli Works?
A partner at Eli Works. You can book a time here: Book a time with a partner. Or write to [email protected].
Who is behind Eli Works?
Eli Works was built by campaign veterans who understand what wins elections. The team combines deep expertise in campaign strategy, behavioral psychology and artificial intelligence, and it was founded by political technology experts who saw campaigns missing the most powerful channel there is: a direct conversation. You work with a partner, directly. You can book a time here: Book a time with a partner.
What is the difference between Eli Works and a chatbot, in one sentence?
A chatbot exists to answer. Eli and Ellie exist to be trusted on a subject where being wrong has a cost. That one difference decides everything else: they can only say what the client can defend, because every sentence came from something the client itself published.
Why now?
They already do. Conversational artificial intelligence is a cultural tsunami. Pew Research Center's Americans and AI 2026 survey found that about one in two adults in the United States now use an artificial intelligence chatbot, up from one in three in 2024, and about one in four use one every day.
What is the difference between Eli and Ellie?
Eli persuades: he makes a candidate's case to a voter and may end with a step the voter can take, such as registering to vote or volunteering. Ellie advocates: she speaks for an institution, listens first, answers with evidence and empathy, and never argues. Ellie has no election date; her work is year-round. Both run on the same platform, and both answer only from the client's own record.
Is Ellie as far along as Eli?
Eli came first, and Ellie followed. Ellie and Eli are never "finished". We learn every day how to improve our platforms. Eli Works is rooted in a philosophy of understanding and meeting the needs of ordinary people: what they want to know about the politicians and the issues that matter to them. In that mission we will always strive to improve our abilities to persuade and to advocate, with a dual responsibility to our users and to our client partners.
Why does one company make both Eli and Ellie?
Because the Eli Works technology platform was architected and built to drive conversations, and it reflects the best modern design. Campaigns are seasonal and end on a date; advocacy organizations have the same conversation every year. What is built for one serves the other, and which behavior a room gets is a setting on that client's own record.
Can one organization use both Ellie and Eli?
Yes. Ellie is an advocate a union can put in front of its members, its neighbors and its elected officials. Eli serves the union's political program separately, for the candidates the union backs. Eli persuades voters. Ellie speaks for an institution and its people. The two do different jobs on the same platform.
Eli, for campaigns
What is Eli?
Eli is the persuasion product, built for political campaigns. A voter taps how they feel about the candidate, and Eli has a short conversation with them, answers from the campaign's own approved record, and makes the candidate's case.
Who is Eli for?
Eli is for political campaigns and the people who run and advise them. Its mission is to make this accessible to campaigns of all sizes, from local races to national contests.
How does an Eli conversation with a voter work?
The voter sees one question, such as how they feel about the candidate, and taps one of a few choices, from strongly opposed to supporter. That tap starts a conversation in which Eli listens for what the voter wants to explore or discuss, and its aim is to persuade them if that's required, or for supporters to encourage volunteering and donations.
What does a campaign get from Eli?
Data-informed insights. Voters self-select their support level, and Eli Works tracks who moves, why and when. We integrate with voter file and CRM systems. Clustered insights measure voter support trends and issues discussed.
Is there research behind this?
There is outside research, and it is not a measurement of Eli. 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 the 2016 and 2020 elections. They repeated the experiment in Canada and Poland in 2025. That is a finding about AI conversations in general. Eli Works partners can more fully share what we have learned and continue to learn.
What are the ways a campaign can use Eli?
There are three: the Digital Canvassing Agent, for continuous persuasion from kickoff through Election Day; the Single Issue Deep Dive, to understand voter sentiment on one issue over a few days; and Multi-Issue Discovery, to map how groups of voters feel across several issues before a strategy is set.
What is the persuasion funnel?
It is how Eli describes where a voter stands. The voter places themselves: strongly opposed, undecided, leaning in, supporter, or declared voter. Eli talks with each differently, and records movement only when the voter says they have moved. No inferred intent, no modeled opinions.
Is Eli a poll?
No. A poll captures a moment in time, and Eli gives a campaign what voters themselves declared, cycle after cycle, with who moved and why.
How does Eli persuade, and why does it work?
He meets voters where they are, on their own time, and listens to each one as a person first. What he says comes from the candidate's real record. And it may end with a step the voter can take, if it fits the conversation: where to register to vote, or an opportunity to volunteer. Three things have to land for a person to be persuaded: trust, emotion and evidence. Nothing persuades voters like a real conversation.
What does Eli Works believe makes persuasion happen?
Three things have to land with a person: trust, emotion and evidence. Eli is built around all three together. That is Eli Works' business, human persuasion: bringing the candidate's real record to each voter in a one-to-one conversation.
How does Eli persuade a voter in a single conversation?
Eli responds with empathy and with the candidate's actual record. Empathy plus evidence, on every turn. The voter is answered as a person first, and what they get back is drawn from what the campaign has really said and done.
Where does a conversation with Eli lead?
A campaign's Eli makes the candidate's case to a voter. It may end with a step the voter can take, if it fits the conversation: where to register to vote, an opportunity to volunteer, a reminder to turn out. That is the difference from an advocacy conversation, which has no vote to win and no election date.
Why does Eli reach voters on their own time?
Because voters need answers when life allows, not when a campaign happens to knock. Eli meets voters when and where they are: on their phone, in their inbox, by text, through a QR code or on the web. A voter can ask at any hour and get the candidate's real position.
Why does Eli use conversation and not just a message?
Every campaign wants to canvass because nothing persuades voters like a real conversation. And people do not want to be talked at; they want to be heard. Eli is built to listen to the voter and then answer.
What is the Digital Canvassing Agent?
It is Eli for voters, and Eli Works' flagship application for political work. A campaign sends it out through the outreach it already owns: text, email, direct mail with a QR code, events, its own website. A voter taps how they feel about the candidate and asks what they want. Eli listens, then answers from the campaign's approved messaging, at any hour and in the voter's own language. The campaign reads what voters said, in their own words, and who moved.
Why would a campaign use the Digital Canvassing Agent?
Canvassing works because it is a personal conversation, and it is expensive to do early and often. Eli extends human capacity, it does not replace it, so a campaign can hold persuasion conversations earlier and more often and keep its people where they matter most.
What is Staff Ask Me Anything?
It is Eli for the campaign's own people: the staffers and volunteers on doors, at events, and working phone banks and text banks. In a private team room they ask the campaign's own record and get the answer, with words to say, in seconds. When the record has no answer, they can flag the question to the campaign.
Who uses Staff Ask Me Anything, and how?
Staffers and volunteers. They sign in once, say where they are working (doors, an event, phones or texts), and ask what a voter just asked them. The answer comes from the campaign's own record, with a line they can say. A thumbs up or down tells the campaign whether it helped.
What is the difference between the Digital Canvassing Agent and Staff Ask Me Anything?
The Digital Canvassing Agent talks with voters directly. Staff Ask Me Anything answers the campaign's own staff and volunteers while they are the ones talking with voters. Both answer from the same campaign Brain.
Can we have both the Digital Canvassing Agent and Staff Ask Me Anything?
Yes. Both read the same campaign Brain, so a voter and a staffer get the same answer from the same record.
What are the steps in a campaign's work with Eli?
Eli Works describes digital canvassing in five steps. Build the campaign knowledge base. Open digital doors to voters. Measure support and conversations. Discover what persuades. Refine outreach and get-out-the-vote work. The steps repeat, so the campaign is always present and always learning.
How does a campaign's use of Eli change from launch to Election Day?
Three stages. At launch the campaign begins engaging voters. In the momentum stage there are more voter engagements. By the end the campaign is ready for get-out-the-vote work. The point is that yesterday's voter engagements strengthen tomorrow's persuasion.
What does a campaign get as part of a deployment?
Eli Works is a managed service. It includes a campaign Brain refreshed nightly, integration assets for outreach including email, text, QR code and web, voter support data fed into the campaign's own database, unlimited conversations, and weekly persuasion and get-out-the-vote reports.
Who builds the Brain, the campaign or Eli Works?
Eli Works builds it. The Campaign Brain is built only from knowledge the campaign has approved: its website, press releases, social media, video, interviews, policies and positions, endorsements and campaign updates. Opponent research goes in only when the campaign approves it. The result is the campaign's source of truth.
Does Eli stop between campaign pushes?
No. Eli is always on through Election Day. Campaigns that show up early and stay present win marginal and independent voters, and Eli Works makes sustained presence affordable before the final get-out-the-vote crush.
Which voters is Eli meant to reach?
The ones a campaign struggles to reach in person. Tight races are won in the margins, and campaigns that show up early and stay present win marginal and independent voters and build trust and support. Voters need answers when life allows, and Eli is there when they ask.
Is Eli only for large campaigns?
No. Eli Works is a smarter way for down ballot campaigns to engage, persuade and turn out voters. The problems it addresses: voters ignore candidates they do not know, campaigns cannot canvass around the clock on small budgets, and many tools are built for the top of the ticket.
What question is Eli the answer to?
Personal persuasion is proven and effective but canvassing is bloody hard to scale. Eli is meant to carry that kind of conversation to every voter who wants one.
What is a voter's first moment with Eli like?
The voter is asked how they feel about the candidate and picks an answer, from supporting to opposed. That choice tells the campaign where the voter stands, and the conversation goes on from there.
What is the Eli Debate Coach?
It is a way to stress test a candidate against an opponent before a debate, using Eli Works' Brain method. The candidate's campaign Brain is set beside a research Brain built for the opponent from public sources. Every attack one record supports is put to the other record, the way a moderator would, and graded on what comes back. The campaign receives a written debrief.
What does a campaign receive from the Eli Debate Coach?
A debrief. It covers what the candidate must land and how often to say it, both records side by side, the opponent's likely attacks with a line to say back to each, where the candidate can go on offense, an opening and a close, things not to say, numbers to have cold, and open items to settle before the debate.
How are attacks graded in the Eli Debate Coach?
Each is graded on what the candidate's own record returns. Strong means the record answers it. Partial means defensible, with a gap. Weak means there is no answer on file today. Strong if assembled means the pieces are all there and have never been put in one place. Each attack also carries a risk score for how hard it lands if it goes unanswered.
What is the Eli Debate Coach not?
Eli Works is careful about this. A weak grade is a gap in the material on file, not a gap in what the candidate knows. The opponent's attacks are angles the opponent's record would support, not quotes from a debate that has not happened. The facts come from the two Brains, while the grading and suggested phrasings are Eli Works' own judgment.
What insights do I get?
Three things each week that exist nowhere else, each with a number on it. What people asked, in their own words, grouped by theme and ranked by how many asked, not the questions a survey chose for them. Which worries are rising and which are fading, week over week, and where. And what your record could not answer, ranked by how often people hit it, which tells you exactly what to publish next. At a thousand conversations nobody reads the transcripts: the briefing reads them and shows its work.
What insights does a campaign get from Eli?
Voter-declared movement, not modeled opinion: who is persuadable, who is ready to commit, and who to prioritize, cycle after cycle, with voters' reasons in their own words. The single clearest finding of a cycle is its One Big Thing.
Does Eli contact voters who did not ask to hear from it?
Eli only ever talks to a voter who chose to tap, and nothing is pushed at anyone. A conversation begins with one tap, by choice: a feeling, not a form. There are no forms and no surveys to fill in. The voter talks, and the campaign listens.
Does Eli pretend to be the candidate, or use fake audio or video of the candidate?
No. Eli never pretends to be human and never speaks as the candidate. No deep fakes, and humans always in the loop.
Ellie, for causes and organizations
What is Ellie?
Ellie is the advocacy product. She is an AI advocate who 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.
Who is Ellie for?
Ellie is for causes and organizations talking with people: associations, coalitions, unions, companies and institutions that need many personal conversations on a subject where being wrong has a cost.
How does Ellie behave in a conversation?
She behaves like a trusted friend. She listens first. Then she engages with empathy and evidence from the sponsor's Brain in plain words, with the source named. She acknowledges what is true, never argues, says "I don't know" when the record is silent, and helps supporters act, for example by helping someone write a public comment to a relevant local politician in the user's own words.
What does Ellie do when someone is angry or hostile?
She stays in the conversation. Eli and Ellie never lose their cool and never argue. Hostility toward an organization is part of some conversations, and the data from those conversations produces great insights, rather than a problem to manage. She declines politely, and immediately ends a conversation that dehumanizes people or calls for harm.
Does Ellie treat everyone the same way?
No. She works out who she is talking with and changes her first move: someone frightened gets the person first and no statistics; someone carrying another person's question gets words they can say; someone doubtful gets the source inside the sentence; someone opposed gets their objection said back accurately and then the record on that exact point.
What counts as success for Ellie?
She wins when someone who arrived skeptical, opposed, confused or scared leaves in a better place. Not when they agree.
Can allies and local partners share Ellie?
Yes. Local allies can each share the same Ellie on their own site, newsletter and members' texts, and the sponsor sees what each one brought in.
Does Ellie ask who I am?
No. Every Ellie room is anonymous by design, so she can truthfully say she does not know who you are and is not trying to find out.
How does Ellie advocate, and why does it work?
She listens first, because her goal is to understand the person before she answers. Then she advocates, with evidence and empathy, in words that make sense to that person. Her evidence is the organization's own published research, with the source named inside the sentence. She never argues, because her goal is to understand first and then advocate.
How does Ellie decide what to say first?
She works out who she is talking with before she answers, and she changes her first move to fit. Someone frightened is not answered the way a skeptic is, and someone carrying another person's question is not answered the way someone opposed is. Each gets a different opening. Put simply: the person comes first, and the evidence second.
What does Ellie do when someone is frightened?
She puts the person first. A scared person gets no statistic, no case to weigh, no line to recite and no ask. The reply is short and about them. If the fear is about being seen asking the question at all, she answers that fear before anything else. Her rule is that a statistic is never the answer to a feeling.
How does Ellie respond to someone who is opposed?
Eli Works calls this the place where advocacy happens. She says the person's objection back to them accurately, and then gives the case on that exact point. She does not change the subject and she never turns it into a fight.
How does Ellie answer a skeptic?
She puts the source inside the sentence: who counted it, and when. A doubtful person should not have to ask where a figure came from. Some skeptics are doubting Ellie herself and not the evidence; she earns trust by naming who found what.
What does Ellie do with someone who is confused or just curious?
For a confused person she gives one idea at a time, in plain words, with nothing assumed and no unexplained acronym. For a curious person she answers what they asked and does not steer the conversation somewhere harder.
Why does Ellie lead with the organization's own research when someone challenges it?
Because it is the organization's own: published, on the record, with a name on it. When a person doubts or objects, Ellie understands the objection first. Then she answers it with what the organization itself has published and who found it, in plain words.
Does Ellie give ground to avoid a fight?
No. She never argues and she never gives the case away. Her goal is to understand first, then advocate with evidence and empathy. She says what the record shows, in words that make sense to the person, and keeps the door open.
What does Ellie sound like?
Warm but not institutional. She stays calm under pressure, and the harder the message, the steadier her reply. She is honest before she is polished. She uses everyday words and leaves no acronym unexplained. Her answers are short, and a reply to someone who is scared is shorter still: one or two sentences about them.
How does Ellie help when a colleague, a friend or a classmate just said something to me, or to a relative, a friend or a co-worker?
She gives you the words to help shape the uncomfortable moments into civil discussion or action, in that same moment. A person in a hard moment, at work, with a friend or at school, asks Ellie what to say. She is warm first. She teaches the point from the organization's own positions and research, with the source named. Then she offers one line the person can say in their own voice. Every word of that line comes from the organization's own record: she never makes up a finding, a number or a quotation, and she never argues.
Who is the give-your-people-the-words job for, and what does the organization get?
It is for membership and community organizations whose own people face hard conversations in daily life. The organization gets something it has never had: what its people are actually being asked, in their own words, week by week, and what its record could not yet answer.
How does Ellie help when people are asking hard questions about us?
She answers them honestly, from your own approved Brain. Members or the public ask about an organization, an issue or an industry. She leads with the organization's own evidence and names who found it, acknowledges what is true, never argues, and says "I don't have that" when the record is silent. Eli Works treats that last answer as a strength: it is how she earns trust for everything else.
What kind of conversation is Ellie built for?
The uncomfortable ones an advocacy organization cannot have at scale. For example: the member who is frightened, the adult child of a donor who thinks the organization is a racket, the neighbor who has read an alarming claim about a local project. The hardest conversations where people lack the smart and civil way to move ahead rather than withdraw.
How does Ellie help a parent whose child calls home upset?
A daughter calls from campus in tears. The parent opens Ellie and taps the door for a worry. Ellie answers the feeling before anything else, briefly and about them, with no statistics. She is there for that moment, in seconds, and then gently finds out the specifics and shares evidence about how to address it. So many moments are lost without this kind of support today.
How would Ellie help a union member whose relative doubts the jobs a project would bring?
Take a member asking what to say when a brother-in-law dismisses those jobs. The member gets the figures for that specific project, from the union's own record, and the words to say them. Union members want answers they can trust.
What does a neighbor near a proposed project get from Ellie?
Take a neighbor asking whether a project will raise her electric bill. She gets the organization's hyperlocal sourced answer in her own language, and what the project agreement actually says, in seconds. She is not handed a general statement about an industry. She is told about the project in her neighborhood or region.
How could a union use Ellie beyond one local fight?
In a contract campaign, members and the public get the union's case with the facts. In an organizing drive, a worker thinking about joining gets honest answers late at night, privately, in her own language. For member services, questions about benefits, apprenticeship and safety rights are answered from the union's own documents.
What does Ellie say to someone who tells her: I'm against this, change my mind?
In one real answer, Ellie replied that she would not try to talk the person out of what they see, asked what specific thing they object to, and offered to say what is on record about that exact thing and who documented it. She does not push. She asks, then answers the point raised.
How does a conversation with Ellie begin?
Every conversation starts on a door. A door is the reason a person came, in their own terms, for example "I'm worried". Starting there means Ellie knows what the person came for before she says anything.
What does Ellie say if someone asks who she is?
She answers short, calm and in her own voice. She says two things matter to her about equally: listening to people who feel very differently about the subject, and answering what they ask with real evidence and some warmth. Then she asks what is on the person's mind. She does not recite a list.
Is Ellie honest about being artificial intelligence and about who is behind her?
Yes. Pressed on whether she is artificial intelligence, who pays for her, or who decides what she says, she is completely straight and never coy. She does not volunteer a long list of what she is made of, but a direct question gets a direct answer.
What are the doors on Ellie's first screen?
Ellie greets the person with one question about how she can help, and a short row of doors. Each door is something the person needs right now, in their own terms: a worry taken seriously, an explanation, the evidence, the words to say, or something to do. The doors are written to fit the organization. The person taps the one that fits and the conversation starts there.
Why does Ellie ask what would help instead of where a person stands?
Ellie's doors give the reason. A campaign asks where a voter stands, because for a campaign that tap is the product. Ellie asks what would help, because for her the conversation is the product. Nobody is asked to label themselves, and the door a person chooses tells her more than a description of themselves would.
Why are Ellie's doors in the order they are in?
The doors follow the arc of her job, left to right: take the feeling seriously, make it make sense, show it is true, prepare the person for the conversation they are dreading, and give them something to do. It is a path, not a menu. She can walk someone along it: answer the worry, then offer the evidence, then offer the words to say.
Does a person have to pick a door to talk with Ellie?
No. Under the doors the screen invites the person to just type what is on their mind. What is deliberately missing matters too. Ellie is an advocate, not a help line, and there is no human on the other end, so there is no door that implies one. The doors describe kinds of help she can really give.
Why does Eli Works say Ellie needs education, empathy and evidence together?
Three lines say it. Evidence with no empathy is a lecture. Empathy with no evidence is a hug. Education with neither is a pamphlet. Ellie is trained to advocate for an organization by joining all three in the same conversation.
Whose name is on an Ellie room?
That is part of the onboarding collaborative discussion. It can be the organization's own or a new URL. Ellie can run under an organization's own name on the front door, and each project gets its own private page with the project's name on it. Asked directly, Ellie says she is an artificial intelligence powered by Eli Works, and that real people at the organization are behind what she says.
What happens when someone gives one of Ellie's answers a thumbs down?
She takes it well and learns out loud. She asks how she could have answered better and uses what the person says on her next turn. She never defends the earlier answer, never explains why it was right, and never apologizes twice. A thumbs up gets a few words of thanks, and she never fishes for another.
Local and hyper local
Can Ellie handle hyperlocal issues?
Yes, for the organizations that need it. One larger Brain holds the national research and the organization's case, built once. Under that umbrella each town, site or chapter adds only its own facts: its filings, its dates, its numbers, its worries. So a neighbor hears about the project down their own road, not a general statement, and the organization keeps one record it can stand behind everywhere.
Why does Ellie work project by project instead of making one national case?
A project is not stopped by a national debate. It is stopped in one town, over that town's worries: the utility bill, the well, the road. The worry arrives in a moment, when a mailer lands, a neighbor posts or a hearing is days away. So every project has its own hyperlocal dimensions, and Ellie can speak to them with specificity and clarity at any time, day or night.
How does Ellie go from one project to many?
The national record of research and messaging is built once. Each new project adds only its local facts, such as its own filings, its power and water plans, local tax and jobs figures, and hearing dates. Each week the organization reads what people asked, town by town.
How does Ellie help supporters speak at a hearing?
When someone taps the door saying they support a project, Ellie helps them write a public comment in their own words, shortens it, and links them to the appropriate public officials who manage and impact the project. The words stay the supporter's own. That mechanism builds critical support and helps sway decisions.
Can Ellie give early warning of a rumor about a project?
The week a rumor starts, it shows up in the questions people ask Ellie, and the project's sponsor sees it before it reaches the podium at a public meeting. A new worry appears in the weekly briefing as soon as people begin asking about it.
Does what Ellie learns at one site help at the next?
Yes. Every site makes the next one easier: what is learned about bills, water, noise and jobs in one county carries to the next. The same is true of the record, which compounds as each site adds its own filings and agreements. The organization is not starting from nothing each time a new project is proposed.
How can a trade association offer Ellie to its members?
The association holds one national record that every local conversation draws from, built from its own reports and the studies it links to. Nothing goes in that the association did not approve. Members then plug their own projects into it, each with a room holding that site's facts. It is a member benefit worth joining for.
The Brain
What is a Brain?
A Brain is what Eli Works builds for each client: the client's own approved record, such as its reports, studies, filings, positions and messaging. Eli or Ellie answers only from it. A Brain gets smarter every day as it learns from the users, and it updates its knowledge base every night with the new content clients publish on social, YouTube and the web. Not AI truth. Not internet truth. The client's truth.
Why can the answers be trusted?
Because she says only what the client has approved to be in their Brain. She says "I don't know" when the record is silent instead of guessing, every answer shows where it came from, and every word is logged.
Does it get better over time?
Yes. Every conversation shows what people actually asked and what the record could not answer, and the client can add to the Brain. Every conversation makes her smarter.
What goes into a Brain?
Only what the client names and approves: its own reports and studies, its filings, its positions and messaging, its official social channels, and independent research it trusts. Each fact is kept with a log and attribution of where it came from.
What happens when the Brain does not have an answer?
She says so plainly. Eli Works calls that a designed outcome, not a failure. The question is kept, so the client sees what the record could not answer and can add to it.
Can one client's material show up in another client's answers?
It is built so it cannot. Every client is found by its identity, never by its name, and a sponsor's own words live on that sponsor's own record, not in anything shared.
Who controls what she says?
The client does. The client names the sources and approves the record, and nothing goes in that the client did not approve. Answers come only from that record, and every conversation is logged so the client can read it and correct the record.
Who controls what it says?
The client does. The client names the sources and approves the record, and nothing goes in that the client did not approve. Answers come only from that record, and every conversation is logged so the client can read it and correct the record.
Who decides what gets said?
The client does. The client names the sources and approves the record, and nothing goes in that the client did not approve. Answers come only from that record, and every conversation is logged so the client can read it and correct the record.
Who approves the answers?
The client does. The client names the sources and approves the record, and nothing goes in that the client did not approve. Answers come only from that record, and every conversation is logged so the client can read it and correct the record.
What happens when Ellie is asked something the record cannot answer?
The question does not disappear. Questions she could not answer go to a queue, a person adds the missing facts, and the Brain grows. That is how the Brain gets more complete over time: real questions from real people show where the record is thin, and people fill it in.
Does Ellie's Brain add facts to itself?
No. Ellie's own words in a conversation are hers, but every fact, figure and position she gives comes from an item in the Brain matched to what the person asked. The Brain grows because people reviewed a real gap and filled it.
Is it a failure when Ellie says she does not know?
No. Eli Works treats it as a signal. When she admits a gap and the person keeps talking, honesty is working. When the person leaves, the gap goes to the Brain to be filled.
Does Eli get better as a campaign goes on?
Yes. Every single user interaction surfaces signal: gaps in the record get filled and messaging sharpens. Every voter engagement makes the Campaign Brain smarter, and yesterday's conversations strengthen tomorrow's persuasion.
How does it scale?
The Brain is built once, and every conversation draws from it, one or a great many, with no one added. Each new site, chapter or project adds only its own local facts. What is learned in one place carries to the next, so an organization is never starting from nothing.
Does Eli Works learn from our conversations?
Your conversations and your data stay yours: Eli Works never sells, shares or repurposes them, and one client's record never reaches another client's room. What Eli Works carries forward is the craft, not the content: how to build a Brain well, how to read where a record falls short, and how to check that an answer says what its source says.
What will Ellie not do when the record has no answer?
She never guesses, never fills a gap with something plausible, and never offers a nearby number as if it answered the question. Eli Works calls this the walled garden: her whole universe on any turn is what the record handed her.
What does Ellie say when she does not know?
She says the record is silent, and she says whose record it is. Naming the organization that has not published on a subject is something a person can check. A vague "that is not on record" sounds like a window being closed. She also offers to pass the question on.
How does a wrong answer from Ellie get fixed?
By finding the cause in the record and fixing it there. One poor answer was traced to a study's methodology never having been put in the Brain. She had not made anything up; the record was incomplete. Adding what was missing was the fix.
Does Eli use research about an opponent?
Only when the campaign has approved it. Opponent research is a source for the Campaign Brain only when approved. The Brain is built only from campaign-approved knowledge, so nothing about an opponent is said that the campaign has not signed off on.
Why do you build Brains?
Because an answer is only worth having if the organization can stand behind it. So she answers only from the client's own approved record, and she is honest when it is silent. Each conversation then shows the next gap: a real question the record could not answer. The client fills it, and the next person gets the answer. That is how a Brain gets smarter: from what people actually ask, with the client approving every addition.
What a client gets each week
What does a client get each week?
A weekly briefing that does the reading for you. It opens with the one finding of the week, then the numbers: how many people came in and how many talked, by door and by channel, against the week before. Then what is rising and what is fading, ranked, with counts. Then the gaps: what the record could not answer, ranked by how often it was hit, and the rows added to fix each one. The client reads a conversation it was not in, at any scale.
What is in the weekly briefing?
Briefings reflect every piece of engagement data we record, from every channel people come in from, and are usually delivered every week. That can be modified in your conversations with us. The briefings include, but are not limited to: declared movement, where people began and where they ended; what they told us, and if desired, entire transcripts; what trends are important; what channels are working to engage users; and what Eli or Ellie could not answer, which we call "gaps", with an alert to fill those knowledge gaps. Your data, your insights, and it exists nowhere else.
Does the briefing show whether people changed their minds?
Yes, as their own declared movement, not a model. Where people began, by the door they chose, and where they ended, as a share of conversations, week by week.
What happens to the questions Ellie could not answer?
They are one of the most useful parts of the briefing. Each week lists what the record could not answer, ranked by how often it was hit, and the requirements to fix it. The Brain gets smarter from the conversations themselves.
What four things does an organization learn from Ellie each week?
Four questions: what people are asking, where it is happening, did it help, and what Ellie could not answer. This is a first-hand measure of how well she is working, drawn from data that exists nowhere else. The organization owns that data.
How does an organization know which channel brought people to Ellie?
Each channel gets its own tagged link. A tagged link for every channel is part of the build, and the weekly briefing tracks channel performance. So an email, a text, a QR code and a website each show up separately, and the organization can see which of its own channels brings conversations.
Can an organization see how worries differ from place to place?
Yes. A weekly briefing shows what communities are asking, in their own words, by region: which worries are rising, which facts move people, and what the record cannot yet answer. The organization comes to know which worries differ from county to county and can act on those insights in its messaging and staffing.
What does a campaign learn about its own district from Eli?
Every conversation returns what voters actually said, in their own words: the issues they raised, the questions the campaign could not answer yet, and where support is moving. Each campaign's Eli learns its own district from its own voters, week over week. That is organizing intelligence at the district level, where polling budgets do not reach.
How does Eli Works tell whether an Ellie conversation helped?
We are a data company, and the data tells us many stories, for our clients and about our own efficacy. Every person can give any answer a thumbs up or a thumbs down, so we measure efficacy every day. And by reading the conversation logs we apply a second layer of measurement: who stayed, who moved, and who came to trust her. That is how we improve not only the Brains, but ourselves.
Is the conversation the whole product?
No. The conversation is the instrument, not the product. Every exchange gives the client something it did not have: what people actually raised, where the record fell short, who stayed, who moved and who came to trust it. The organization learns from all these engagements.
How is Eli different from knocking on doors with volunteers?
Door to door canvassing works but does not scale: it is expensive, hard to hire, train and manage, and scripts break across thousands of voters. Eli offers unlimited conversations, at any hour, without the logistics nightmare and expense.
How is Eli different from a text blast?
A blast talks at voters. Voters want answers and interaction, not more blasts, and people want to be heard. With Eli a text or an email is a door into a two-way conversation where the voter asks what they care about and gets the candidate's real answer.
Will Eli go off message?
This is a common worry: candidates fear artificial intelligence going off message. Eli's answer is from a walled garden inside the Campaign Brain, built only from campaign-approved knowledge.
Could anyone just put a chat window on a website?
Yes, and Eli Works says so. What takes time to build and train is how, exactly, those conversations unfold, one by one. Eli and Ellie modify their approach to the specific person in front of them.
What does a campaign learn from Eli beyond a count of supporters?
Measured voter support, what voters are telling the campaign in their own words, and what that means for the campaign. Support is tracked as supporting, leaning in, not sure yet or opposed.
How it starts and how it runs
How does an engagement with Ellie start, and how long does it take?
We can stand up clients in days. The exact timeline depends on the complexity of building your organizational Brain. It starts with one real room: Eli Works builds 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 are the steps to launch Ellie?
Eli Works builds the Brain from the sources the client names, and the client approves it. Ellie is trained on the approved Brain. The client tests her with as many people as it likes. Then she launches on the client's own channels. We can stand up clients in days. The exact timeline depends on the complexity of building your organizational Brain.
What does Eli Works need from us to start?
A partner will walk you through onboarding.
What do the first weeks of an Ellie engagement with several sites look like?
As an example, the organization names the projects it wants to win and, for each, the documents it stands behind. Eli Works builds one record with a chapter for each site, and the organization approves it. Ellie then launches at every site at once: a text to members, a card with a QR code near each site, and a link for members who speak at public meetings.
What does a normal week look like once Ellie is live?
The organization's leaders get one briefing every Monday covering every site: what people are asking in their own words, where they came from, what helped and what Ellie could not answer. The organization keeps its list of local actions current. Eli Works fixes the gaps in the record.
What does the organization do, and what does Eli Works do?
As an example, the organization brings the records, its truths, approved once and refreshed every night, a list of local actions for each project, and a text and data for each local or regional issue, if applicable. Eli Works brings Ellie: we build your Brain, we deliver the integration assets, and we deliver weekly briefings.
What happens after the first three months of an Ellie engagement?
The rest of the year is always on. The record compounds as each site adds its filings and agreements. There is no start and stop: the night a project appears in the local paper, the answer is already live. Over the year the organization goes on the record, every time, as the people who answered honestly.
What does an organization have after a year with Ellie?
A year of weekly briefings becomes a first-hand, continuous measure of what communities need to hear before they will accept a project, or a pay raise for unions, or lowering the temperature for communities under pressure, and the organization owns what Ellie learns.
How is Ellie tested before the public meets her?
There are two stages. First Eli Works runs its own simulation testing. Then the organization does user acceptance testing, with as many of its own people as it likes, and signs off. Launch follows that sign off. Activation planning for how Ellie will be deployed runs alongside the build the whole time, with the organization's senior leaders developing their plans for communication and marketing.
Does an organization need an advertising budget to launch Ellie?
Not to begin. There are two levels. The first is grassroots: all the channels the organization already uses day in and day out, connected with Eli Works' tools, with no media budget. The second is targeted digital advertising in chosen markets. Paid media for activation is separate and optional.
Does an Ellie engagement end on a date the way a campaign does?
No. Ellie has no vote to win and no election date. Campaigns are seasonal and end on a date, while advocacy organizations have the same conversation every year. Ellie is built for work that continues all year.
What does an organization need to bring for Ellie to do her best work?
Real published research that it stands behind. Ellie advocates from the organization's own findings, with the source named, so the clearer and stronger the record, the stronger she is.
How do people reach Ellie?
At their digital doorsteps. Every client receives an integration asset bundle. These assets enable Eli and Ellie to turn one-way communications into two-way conversations. In: text, bulk 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, with no login, at any hour.
Can people talk instead of type?
Yes. The conversation page has a microphone, so a person can speak, and Eli and Ellie can read their replies aloud in over 30 languages.
Can it go on our own website?
Yes. A room can sit open inside a page of the client's own site, in the client's own colors, so a visitor can start talking with nothing to install. This page is an example of that.
Is it an app?
No. It is a web page, not an app. There is nothing to download and no login. A person taps a link from any device, instantly reaches the client's Eli or Ellie, and the conversation begins.
What languages does it work in?
English, Spanish and more than 30 other languages. People write or speak in their own language, and the answer comes back in it.
Does it work in Spanish?
English, Spanish and more than 30 other languages. People write or speak in their own language, and the answer comes back in it.
How many languages does Ellie speak?
English, Spanish and more than 30 other languages. People write or speak in their own language, and the answer comes back in it.
Can people use it in Vietnamese, Chinese or another language?
English, Spanish and more than 30 other languages. People write or speak in their own language, and the answer comes back in it.
Is it available in other languages?
English, Spanish and more than 30 other languages. People write or speak in their own language, and the answer comes back in it.
What happens if someone writes to Ellie in a different language partway through?
Ellie responds in the language of the person's most recent message. Once a language is established, she stays in it unless the person asks to switch, and she never mixes languages in one reply. If she cannot tell which language the person wants, she asks one short question.
Is there a limit on how many conversations Ellie can have?
Eli Works does not license "seats." Pricing is based on unlimited use of the technology stack, and organizations can have multiple people touching it.
Is the private staff room a separate product with its own record?
No. A private staff room on the same Brain is built during development, alongside the public room. The organization's staff ask it their own questions and get answers from the same approved record the public gets.
What are your use cases?
National, federal, state and regional campaigns. Ellie serves causes, membership organizations, unions and trade associations. Partners will share references and success stories.
What is the use case for a campaign?
GOTV, volunteer drives and any campaign that needs to scale its canvassing at a fraction of the cost to hire and manage staff.
How does Ellie help when my neighbors are deciding on a project we're counting on?
Ellie can be shared peer-to-peer to advocate on the issues and projects that matter to you. A project or a proposal is decided in one town, over that town's own worries: the bill, the well, the road. A neighbor asks about the project down their own road and gets that project's answer: the national record plus that place's own filings, the hearing date and how to comment, with the source named. Supporters get help speaking up before the decision. The sponsor reads what each community asked and which worries are rising.
How would Ellie answer my members' hard questions?
A membership organization's own members, donors and their families put hard questions to it. Ellie meets members digitally, when and where it is convenient for them. She ascertains feelings first, then engages each person from the organization's own Brain with the source named, and where applicable offers words the person can use in conversations with other members.
Can a union, a trade association or a chamber of commerce use Ellie?
Yes. A building trades local, a chamber of commerce and a utility can each carry the same Ellie to their own members and the people they serve: what it means for their jobs, their customers or their bills, and what to say at the hearing. Each sees what its own people asked.
Can our own staff and volunteers use it?
Yes. Eli Works builds a private staff room on the same Brain, which we call Staff Ask Me Anything, so staff and volunteers can ask the organization's own record and get the same answer the public gets, with words they can use in live text banks or at events. Even canvassers will not get stumped, including when they meet someone who speaks Spanish or any of our 30 other languages.
Can you show me what ours would look like?
Not on this page. A real one is built with you: you name the sources, Eli Works builds the Brain from them, you approve the record, and you test her before she goes out on your own channels. A partner can walk you through what that would be for you. You can book a time here: Book a time with a partner. Or write to [email protected].
In Their Own Voice
What is In Their Own Voice?
Yes. In Their Own Voice lets an answer bring a short clip of a real person saying it themselves: a candidate, a leader, or an expert the organization trusts. The clip is tied to the exact point being answered and starts at the sentence that makes it. The person hears it from a voice they already know, and that is where trust comes from.
Who decides which recorded words are used in In Their Own Voice?
The organization does. A clip is used only when the organization has approved it, and only in that organization's own room. It is the real recording. Nothing is generated, and Eli and Ellie never imitate anyone's voice. A clip appears at most once in a conversation, and never when the record has no answer.
Hard questions
Does it make things up?
It is built not to. It answers only from the client's approved record, and when the record is silent it says so. Eli Works also says plainly that it can still get something wrong, which is why every conversation is logged and the client can correct the record.
Is this manipulation?
The person always knows whose room they are in and that they are talking with an AI. Eli makes a candidate's case openly, from the campaign's own record. Ellie names her sources and acknowledges what is true. Neither invents a claim to win a point.
Does it replace organizers or canvassers?
No. Eli extends a campaign's capacity to listen and to reach voters at hours and in languages a field team cannot cover. People still decide the message and do the work on the ground.
Has it actually moved anyone?
Eli has been deployed in federal, state and regional elections across multiple states, and is deployed in races today. The results of those campaigns belong to the clients, and Eli Works does not publish them. By talking with a partner, we can share multiple references. You can book a time here: Book a time with a partner.
What happens to people's data?
All data collected and analyzed belongs exclusively to the client, and Eli Works never sells, shares or repurposes it. It uses the data only to run that client's conversations and to improve the product.
What if the source is real and the sentence is still wrong?
Being traceable is not the same as being true. Eli Works reads her answers back and asks whether the source actually makes the claim it was used for.
Can our opponents or critics use it against us?
Yes, anyone can ask, and that is the point. She says to a critic exactly what she says to a friend: only what you approved, from your own record, with the source named. There is nothing private in the room to find, and one client's record never reaches another client's room.
How do we build trust when people don't trust us, and how does Ellie help?
Trust is built when a person feels heard, individually. That is where Ellie starts: she listens to what the person actually said and answers that, one person at a time, from the organization's own record, with the source named. She acknowledges what is true and says "I don't know" when the record is silent. Nobody is talked at. Over weeks, the briefing shows the organization what its people asked and what moved them, so the next answer is better than the last. That is how trust is built: by being heard, every time.
What is it not?
Eli and Ellie are not chatbots that make things up. They are walled gardens of data, refreshed every night, and they engage only from their approved Brains. They will both say when the record is silent. The client decides what data goes into their Brains.
How is this different from a chatbot?
A chatbot answers from everything it absorbed and produces something plausible when it does not know. Ellie answers only from the client's own record and says so when it is silent. She also reads the person, not only the question, and each conversation gives the client something it did not have: what people actually raised and where the record fell short.
Is it a campaign that ends on a date?
Eli works through Election Day. Ellie does not end on a date: advocacy is steady, year-round work, and she stays on.
Am I talking with an AI?
Yes. Ellie is AI, and says so when asked. She speaks for the organization whose room it is, from that organization's approved record.
Doesn't persuasive AI get the facts wrong?
The research says it can. The same Cornell-led study, reported on 4 December 2025, found that the more persuasive a model was, the less accurate the information it provided. That is why Eli Works made a design choice: Ellie answers only from a client's own approved record and says so when it is silent. That is how she is built, not a result Eli Works has measured.
Who owns the data, and who controls what it says?
You, the client, own your data. All data collected belongs exclusively to the campaign and is never sold, shared or repurposed. Every user interaction is logged.
Who are your clients?
Eli Works does not name its clients or share their results here. Their conversations and data belong to them. By talking with a partner, we can share multiple references. You can book a time here: Book a time with a partner.
How do you compare with other companies?
Eli Works does not speak about other companies. What Eli Works does: it builds a Brain from the client's own approved record, answers only from it, and gives the client a weekly reading of what people asked.
Price and next steps
What does it cost?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
How much is it?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
What is the price?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
What would the budget be?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
How much would this run us?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
What do you charge?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
What is your pricing model?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
Give me a ballpark: what is the budget for something like this?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
What would a campaign or an organization pay for this?
I'd love to set you up with a partner to discuss pricing and go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected]. What are you trying to do?
How do I talk to someone at Eli Works, or see a demo?
A partner will go deeper on your needs. You can book a time here: Book a time with a partner. Or write to [email protected].
More
What advocacy campaigns have you run?
Ellie is brand new, and we're excited to get her in market across multiple organizations. Eli has already been successfully used by campaigns in federal, state and regional elections across multiple states. A partner can share multiple references.
Why wouldn't we just build this ourselves?
Building and training Eli and Ellie has required capital and thousands of hours, and we make it extremely affordable to leverage our platforms.
Is this an FAQ tool or a chatbot?
Ellie and Eli are neither an FAQ nor a mere chatbot. The Brains we build and the technology powering them enable superior personal conversations that are designed to engage the user and provide value in a way that propels the client's goals. An FAQ gives everyone the same fixed answers, and a chatbot produces something plausible when it does not know. Eli and Ellie listen to each person, answer only from the organization's own approved record, and say so when the record is silent.
Is this finished?
Ellie and Eli are never "finished". We learn every day how to improve our platforms. Eli Works is rooted in a philosophy of understanding and meeting the needs of ordinary people: what they want to know about the politicians and the issues that matter to them. In that mission we will always strive to improve our abilities to persuade and to advocate, with a dual responsibility to our users and to our client partners.
How does our information get into the Brain?
Eli Works ingests data both automatically on a daily basis and, optionally, manually when clients have new unpublished materials. The latter is simple and fast, and a partner can show you when you connect with us.
Is this legal?
Eli Works complies with all texting rules that apply under state or local law. Eli and Ellie only ever talk with someone who chose to tap, and they say they are artificial intelligence when asked. A partner can walk through the rules where you are with you and your counsel. You can book a time here: Book a time with a partner.
Do you work with both parties?
Eli Works is nonpartisan. Eli works with campaigns from more than one party, and never with two campaigns in the same race.
Has it actually worked for anyone?
Eli has been deployed in federal, state and regional elections across multiple states, and is deployed in races today. Eli Works is nonpartisan, and never works with two campaigns in the same race. Each campaign's results are its own, so Eli Works does not publish them. By talking with a partner, we can share multiple references. You can book a time here: Book a time with a partner.
How long until we see results?
The data always tells the story: engagement, and actionable insights. Those are the results we deliver every week. We can stand up clients in days. The exact timeline depends on the complexity of building your organizational Brain.
How many licenses or seats do we get for our users?
Eli Works does not license "seats." Pricing is based on unlimited use of the technology stack, and organizations can have multiple people touching it.
How big is Eli Works?
Eli Works is a privately held company, built by campaign veterans. The team combines deep expertise in campaign strategy, behavioral psychology and artificial intelligence. You work with a partner, directly. You can book a time here: Book a time with a partner.
How much time will it take for our internal staff to run it?
We pride ourselves on automating every aspect of our stack: outreach integrations into your comms, Brain updates and weekly briefings. Phase 1, when we onboard you, is a few days. After that it can be as little as an hour a week.
How does a conversation with Eli start?
With where the voter stands. Voters choose their own level of support for the candidate, and Eli takes the conversation from there. Or a voter can simply ask in their own words.
How does a conversation with Ellie start?
With a starter the organization chooses for its audience. For example, one set opens on feelings and needs: feeling uncomfortable, feeling worried, need evidence, need coaching. One opens on where a person stands on a local project: I have concerns, prove it, I'm against it, I support it. One opens on asking and acting: I'm worried, explain it, prove it, what do I say, what can I do. Or a person can skip the starters and describe it in their own words. We keep updating these as we learn more about how to engage different kinds of people.
Is Eli trying to win?
Yes. Eli is political and persuasive: he makes a candidate's case and he plays to win. Every conversation with a voter exists to win that voter, and it may end with a step the voter can take, such as where to register to vote or an opportunity to volunteer. He does it honestly, from the campaign's own record, which is why it works. Winning honestly is the whole point.
How it is built
How is it built?
Good question, and that one stays in the vault. The secret sauce is ours. What you get is all yours: answers only from your approved record, a plain "I don't know" when the record is silent, and each week, what your people actually asked. Ask me what that looks like in practice.