Demo Before the Call or During It?

Send it before the call. When you sell AI agents the product is a conversation, and a conversation is something a prospect can have alone at eleven at night without you narrating it. Once the demo works without you in the room, the call stops being a demonstration and becomes a scoping conversation, which is a far better use of thirty minutes for both people on it.

That answer comes with a caveat. Demoing live is not a rookie habit you grow out of. It is the default because it earned that position in software sales, where the product is a dense interface that needs a guide. Owners who move the demo earlier without knowing what they gave up often conclude that pre-call demos "do not work", when what actually happened is that they dropped the parts of the live demo that were quietly doing the selling.

Here is the honest version of both cases, and why the sequencing flips when you sell an agent rather than a dashboard.

The real case for demoing during the call

Start with the incumbent, because the argument for it is stronger than the demo-first crowd admits.

You control the narrative

Live, you decide the order. You open on the strongest behavior, route around the half-finished part, and frame everything before the prospect can interpret it themselves. If the agent answers oddly, you say "we would tune that to your script" in the same breath and the moment passes. Sent ahead of the call, that same odd answer sits on screen unattended, and unattended moments become the whole impression.

You can read the room

You see where they lean in. You notice they went quiet at pricing and lit up at after-hours booking, and that signal shapes the rest of the call and often the proposal. Asynchronously you get open and engagement data, which tells you what happened but not what they felt about it.

You handle objections while they are still small

Every AI agent demo produces the same doubts: will it sound like us, what happens when it gets something wrong, does it connect to our calendar. On a call those are thirty-second answers. Left alone, a prospect answers them privately and badly. The calendar integration objection is the one most often resolved silently and unfavorably.

The demo is a legitimate reason for the meeting to exist

"Can I show you what this would look like for your shop?" is one of the few honest reasons to ask a stranger for time. Give the demo away first and you have to earn the meeting on something else. That is a real cost.

The real case for demoing before the call

The other side changes what the call is for, not just when the demo happens.

The call converts from demonstration to scoping

This is the prize. A prospect who has already used the agent does not need to be shown it again. They arrive with operational questions: what happens to the calls we already answer, can it stop quoting prices, who sets the hours. The meeting goes to their rollout instead of your walkthrough, and lands on scope and price rather than on whether the thing is real.

It filters, and filtering is a feature

Some people open the demo, decide it is not for them, and never book. That looks like lost pipeline and is mostly recovered time, because those calls were heading for a polite "send me some information" anyway. Whoever books afterward has already crossed the belief threshold the first half of a live demo exists to get them across.

Proof arrives while they are still cold

A live demo only reaches people who already agreed to a meeting, which locks your best asset behind your worst conversion step. Sending it first puts the proof where the skepticism actually is. That inversion is the core of the reverse demo method, and it is why the approach works cold.

It survives the person who was not on the call

At a local business the decision often involves a partner or an office manager. A live demo cannot be forwarded. A link can, and the second conversation happens whether you are there or not.

Why before wins specifically for AI agents

In classic SaaS the demo is a tour of an interface, and interfaces need narration because the value is spread across menus a first-time viewer cannot navigate. That is why the live walkthrough became standard. An AI agent has no interface to tour. It has a greeting, a set of questions, and a response to whatever the prospect throws at it. There is nothing to point at, and the entire product surfaces in the first exchange.

That changes who the better operator is. With a dashboard, you are, because you know the map. With an agent, they are, because the thing you most need them to believe is that it holds up against a question you did not plant. A prospect testing alone will ask something rude, something oddly specific, and something you would never have demoed. Passing their test is worth more than passing yours.

A live demo proves the agent works when you are driving. A pre-call demo proves it works when you are not. Only one of those is the thing they are buying.

There is a structural reason too. Tour platforms like Navattic and Storylane assume a product you capture once and reuse. Most AI agencies have no product interface at all, and the thing worth showing is not your build, it is their business running on it. That flips the object of the demo from your software to their operation, which only works per prospect.

What you give up going first, and how to get it back

The losses in the live column are real. They are also mostly recoverable if you design for them.

Narrow the surface

Unattended demos fail on breadth. Every extra capability is another chance for the agent to say something you would have framed away. Build one workflow, the most expensive one for that business, and let the rest wait for the call.

Write the framing into the message

What you would have said out loud has to travel with the link. Two lines is enough: what it is, what it was built from, and what it is not. Saying plainly that this is a preview built from their website, not a system wired into their phones, prevents the trust break that comes from a prospect discovering the boundary themselves.

Move the objection handling, do not delete it

The calendar question, the accuracy question, and the "will it sound like us" question still need answers. Short version in the follow-up, full version on the call. What matters is that nobody sits with an unanswered doubt for four days. Our guide on what to send after a prospect opens your demo covers the timing.

The sequence that actually works

In practice these are not exclusive, and the strongest version is both in a specific order.

Send a narrow, business-specific demo first, and let engagement with it book the meeting. Then on the call, do not repeat the walkthrough. Demo forward: their edge case, the question they raised, a second agent type, the handoff behavior they could not test alone. You keep narrative control and objection handling, but spend them on what is still uncertain rather than on proving the product exists.

When live still wins outright

Keep the demo on the call for warm inbound and referrals, where belief is not the bottleneck. Keep it live when a technical stakeholder wants to interrogate the build, when the prospect has no website to build from, and when the format itself is worth watching, as with a voice agent handled live on a call.

Making pre-call demos survivable at volume

The reason most agencies keep demoing on the call is arithmetic, not conviction. A business-specific demo built by hand runs thirty to sixty minutes per prospect, so the sequencing argument only matters if the demos can exist at volume.

That is the gap Ciela covers. You point it at a prospect's website, it builds a working chat, receptionist, or missed-call text-back agent on that specific business, and you copy the link into whatever outreach you already run in Instantly, Apollo, Clay, or HeyReach. Booking and capture happen inside the demo rather than touching the prospect's real calendar, so nothing is misrepresented while they test it alone. The personalized demo at scale breakdown covers the mechanics.

Frequently asked questions

Should you demo before the sales call or during it?

For AI agents, send the demo before the call. The product is a conversation the prospect can have alone, so once they have used it the call becomes a scoping conversation about rollout rather than a demonstration. Live demos still win for warm inbound, referrals, and any deal with a technical stakeholder who wants to interrogate the build.

What do you lose by demoing before the call instead of during it?

You lose narrative control, real-time objection handling, and the ability to read the prospect's reaction. An odd answer you would have framed away in a live demo instead sits on screen unattended. Most of that is recoverable by narrowing the demo to one workflow and writing the framing into the message that carries the link.

Does a pre-call demo reduce the number of calls you book?

It changes who books more than how many. People who would have taken a polite meeting and then gone quiet tend to filter themselves out, and the ones who do book have already used the thing. That trades call volume for call quality, which is the right trade when your calendar is the constraint.

What should the sales call cover if the prospect already saw the demo?

Scope, not capability. Cover which workflow goes first, where enquiries arrive today, who owns the phone and the calendar, what the agent must never say, how handoff to a human works, and what rollout and price look like. Save live screen time for the edge case that specific prospect raised.

Is a recorded demo video the same as demoing before the call?

No. A recording is still you performing, just asynchronously, and the prospect cannot test it. An interactive demo lets them type their own awkward question and watch what happens, which is the part that converts skepticism into interest. Recordings are a fallback when there is nothing interactive to send.

How do you get a prospect to actually open a pre-call demo?

Lead with the demo instead of burying it under agency positioning, make the reason for it specific to their business, and set an honest expectation such as two minutes and no signup. If the message needs four sentences of setup before the link appears, most people quit before they reach it.

Should you still demo live if the prospect already used the demo?

Only forward, never again from the start. Repeating a walkthrough they already did wastes the goodwill it earned. Demo the part they could not test alone instead: their specific edge case, a second agent type, or the handoff behavior they asked about.

If the sequencing makes sense and the bottleneck is producing one demo per prospect, start with demo software for AI agencies, then check pricing.