Send an Agent - the Better Way to Run 1:1 Meetings

A product walkthrough. Why the most valuable meetings in a company are 1:1s, why teams quietly stopped running them, and how sending an AI agent brings them back - better than the originals.

The best meetings in any company are 1:1s.

One person, full attention, real questions, no audience. Screening a candidate. Training a new rep. Hearing what a customer actually thinks. Every one of these works best as a conversation between two parties, one of whom knows exactly what to ask.

And 1:1s are exactly the meetings teams can't afford to run.

So we compromise. We batch the training into a workshop. We swap the customer conversation for an async video request. We pull an engineer off their sprint to run the screening round. Every compromise trades away the thing that made the 1:1 valuable - the interaction - to make the calendar work.

Send Agent removes the trade. When you log into Waterr, the first page says exactly what it does: send an agent to take meetings and video calls. You create an agent with the right expertise and context, share a link, and the other person jumps into a live video conversation - the agent asks, listens, follows up, and when the call ends, a scored summary lands back with you.

Sending an agent doesn't mean the meeting matters less. It means the meeting actually happens - 1:1, on their time, with full attention - instead of being batched, deferred, or skipped.

Here's why that's better than the meetings we run today, and how little it takes to do.

The 1:1 you can't staff - screening outside your bench

Say you're hiring your first researcher. Nobody on the team is a researcher. Who runs the screening round?

Today the answer is bad in every direction - the founder wings it off a list of questions from a blog post, or an external consultant is hired to sit in, or the screen just doesn't really screen and the decision slides to gut feel. The expertise the conversation needs doesn't exist in the company yet. That's why you're hiring.

With Send Agent, you create the interviewer you don't have. Describe the role, attach the job description, and the agent shows up with a research-literate line of questioning - asks about methodology, digs into past work, pushes back on vague answers, and scores every candidate against the same criteria. The conversation you couldn't staff becomes a conversation you can send.

The same logic runs inside teams that do have the expertise. Every screening round an engineer conducts is an hour of engineering that didn't happen - multiplied across a hiring pipeline, that's weeks of senior time spent asking the same first-round questions. Your engineering team builds the screening agent once - their questions, their bar, their follow-ups - and sends it to every candidate. The engineers stay in the sprint. The bar stays theirs. Every candidate gets the same fair, structured hour, and the humans enter at the round where human judgment actually differentiates.

The 1:1 nobody completes - testimonials beyond async video

Every team knows this loop. You ask a happy customer for a testimonial, send them a Loom-style async video link, and then - nothing. Most never record it.

Not because they're unhappy. Because talking into a void is hard. An async request hands the customer a blank screen and the burden of structure: what do I say, how do I start, was that take good enough? No prompts, no follow-ups, no reaction. It's a monologue assignment disguised as a favor, and most people quietly decline it.

A testimonial is a conversation, not a recording task. Send a testimonial agent instead, and the customer clicks the link and someone asks them things - what were you trying to solve, what changed after, what would you tell someone considering this? They just answer questions, which is easy, instead of composing a monologue, which is hard. Follow-ups surface the specifics that make a testimonial usable - the numbers, the before-and-after - that a self-recorded take never includes. Async video made testimonials cheap to request. An interactive agent makes them likely to happen.

The 1:1 nobody speaks in - training without an audience

Group training has a dirty secret: almost nobody asks questions.

Rooms and workshops run in bulk, and bulk creates an audience. Asking a basic question in front of the team is admitting in public what you don't know - so people don't. They nod through the workshop, the trainer moves at the speed of the average, and the actual learning is deferred to fumbling through it alone later. The format is efficient for the calendar and terrible for the learner.

A 1:1 with an agent is a judgment-free environment - and that changes who speaks. There's no room to read, no colleague to impress, no such thing as a stupid question. The new rep can run the cold-call objection drill four times in a row and flub the first three, because nobody is watching and the agent's patience doesn't run out. The agent adapts to the person in front of it instead of the average of thirty, and every session ends with concrete, private feedback - what worked, what to fix - instead of a certificate of attendance.

The pattern across all three is the same: the 1:1 was always the better format, and the only argument against it was cost. Group sessions, async requests, borrowed engineers - those were workarounds for expert attention being scarce. Send Agent makes that attention mintable.

How little it takes - from nothing to a live agent in minutes

None of this works if creating the agent is a project. So it isn't. Three steps, no engineer, no setup call.

Pick or describe the agent. The meeting-link builder opens with ready-made agents - behavioral and technical interviews, sales roleplay, on-screen IT support, product onboarding, testimonial collection - each arriving complete with a persona, a conversation plan, and scoring criteria. Nothing fits? Describe the meeting in plain language and the studio generates the persona, the conversation flow, and the goals for you. You're not writing a prompt. You're choosing a colleague.

Give it context. Before the link is minted, there's a context box - anything the agent should know walking in. "Screening for a senior ML researcher, focus on evaluation methodology, we're pre-PMF." "This customer just renewed after a rocky first quarter - ask about the turnaround." Same agent, different context, completely different meeting.

Send the link. You get a short link - waterr.ai/xxx-xxx-xxx - and paste it into email, Slack, wherever the conversation already lives. The other person clicks and jumps straight into the call - no account, no login, no app install. They see who they're meeting, grant mic permissions, and they're talking. For structured processes, the scheduled variant emails each participant their own link with a start time, an expiry window, and a one-use limit - fair assessments without link-forwarding leaks.

Then the loop closes. About a minute after the call ends, the conversation is scored against the agent's goals - summary, strengths, growth areas, full transcript - waiting in your dashboard. You weren't in the meeting. You still know exactly what happened in it, in a form you can compare across every candidate, rep, or customer the agent met.

Minutes to create. One click to join. Scored before you've finished your coffee.

Agents as first-class teammates, not offloaded chores

The lazy read of all this is "AI takes the meetings you don't want." That's not what fast-moving teams are actually doing with it.

They're treating agents as first-class members of the team - entities you brief, delegate to, and hold to a bar, exactly like a colleague. The unit of delegation changes: not "send this doc and hope they read it" but "send this agent and know the conversation happened." And the work doesn't just move off a human's plate - it comes back better:

• Screening stops depending on who was free that afternoon - every candidate gets the same structured hour, scored on the same criteria, and bias toward the confident interviewer's gut shrinks.
• Testimonials stop dying in async purgatory - the response rate problem was an interactivity problem, and the agent solves the interactivity.
• Training stops being a broadcast - every person gets the 1:1 attention that was previously reserved for whoever the manager had time for.
• Experts multiply - the engineer's screening bar, the sales lead's objection drill, the researcher you don't have yet: each becomes an agent that runs a hundred simultaneous 1:1s without the expert leaving their work.

The teams moving fastest aren't the ones that automated their meetings away. They're the ones that noticed 1:1 attention was the scarcest resource in the company - and made it something you can send.

The best meetings in any company are 1:1s. Now you can afford to run all of them.

Read next: Plus One for Developers for the flip side - the agent that takes your incoming calls so you stay in flow - and Introducing Plus One for why we built agents that sit in video calls at all.