The demo trap

One of the ways we've been showing people Anam is to build a version of their own website with our agent embedded in it. It's a mock-up, but it puts the product somewhere they recognise. They can talk to the agent and start imagining how it might fit into their business.

We've built a lot of these personalised demos for particular customers, and they've been working quite well at getting people excited. Alongside that, we've been trying to find a broader demo that gets people interested without being made for them individually. We wanted a major viral moment. We haven't managed it.

That difference has been useful to think about. Something can work well in a customer conversation and still leave you with a difficult question about what to put on your homepage.

Put it in their world#

The customer website demos give us a specific starting point. We're showing an Anam agent in the context of a business the person already understands. The mock-up makes the connection between the technology and their product visible.

I think that does a lot of the work. With a general avatar demo, someone has to understand the technology and then imagine a use for it. With a version of their own website, we've already done some of that imagining for them. It gives the conversation a place to start.

The response has been encouraging, but it tells us about that particular setting. We've chosen the customer and built something for them. Their interest doesn't tell us which demo will make an unfamiliar visitor stop, try it and want to share it.

That's the trap I've been thinking about: taking something that works when it's personalised and expecting the same response when that context is gone.

An agent that knows about Anam#

We've tried several versions of the demo on Anam's homepage. They share the same basic interaction: you talk to one of our agents.

One version knew all about Anam. The idea made sense for the setting. You're on our website, so here's an agent that can answer questions about the company and product while demonstrating the technology.

But we were still struggling to capture people with it. My read is that an agent knowing about Anam only goes so far if the visitor hasn't yet found a reason to be interested in Anam. They still have to decide what to ask and what they want to get out of the conversation.

The personalised demo has a reason for the conversation built into its setting. On the homepage, we were asking the visitor to supply more of that reason themselves.

Give the conversation a game#

We settled on a guessing game. The agent thinks of a person, and you try to work out who it is.

That seemed to work better, although still not as well as we'd hoped. It gave the interaction something to be about immediately. You could try to solve the game while experiencing the agent, without first coming up with a question about our business.

I think that's a useful change in what we're asking of someone. A general invitation to talk gives them freedom, but it also leaves them to decide what to do. A guessing game gives them a role and something to work towards.

It still didn't give us the viral moment we wanted. The improvement was worth keeping, but we hadn't found the broad appeal we were looking for.

What I'm taking from it#

The customer demos and the homepage experiments have made me pay more attention to the reason someone would start a conversation. An agent being available to talk is only part of the proposition. The setting and the activity can give someone a reason to try it.

I also want to be careful about what we conclude from a better demo. Getting a particular customer excited, making a homepage interaction more engaging and creating something people want to share are different outcomes. We've had encouraging results with the first and some improvement with the second. We haven't achieved the third at the scale we'd hoped for.

For the next demo, I want to start with what the person will actually do in it. What makes them want to begin, and what would make the experience worth telling someone else about? The guessing game gave us a better starting point for the first question. We're still working on the second.