X.Ai’s Amy — A brilliant MVP experiment (and no Ai at all) ?

Amy has been making a lot of headlines, for quite some time. So who is she? Here’s a quote from May 2014:

Amy, a personal assistant that has been built to help you schedule meetings […]

Getting copied in triggers Amy (full name: Amy Ingram, apparently) to read the email, look for date and time and place suggestions and then continue to the conversation directly with the other person to find a time and place that works for everyone — and then put the detail into your diary.

[…] “Anything which a human PA could add to a traditional meeting invite she can do,” co-founder and CEO Dennis Mortensen tells me.

The New York-based startup that makes Amy, X.ai (“Ex-dot-A I”), is today announcing $2.1 million in funding to develop the product, with the seed round coming from a top-shelf list of investors: IA Ventures, Softbank and Lerer Ventures.

Meanwhile, Amy is still be in a closed beta. You have to register your interest for now, and Mortensen says that a new intake will be made on August 1.

source: X.ai Lands $2.1M To Build ‘Amy’, An AI-Powered Personal Assistant For Meetings on TechCrunch.

Here’s a typical email that Amy sends:

On 10 Feb 2015 20:24, “Amy Ingram” <amy@x.ai> wrote:

Hi Pejvan,

Happy to get something on Jonathan’s calendar.

Does Thursday, Feb 12 at 11:00 AM EST work? Alternatively, Jonathan is available Thursday, Feb 12 at 4:00 PM EST or Friday, Feb 13 at 11:00 AM.

What’s the best number for Jonathan to call?

Amy

Amy Ingram | Personal Assistant to Jonathan
x.ai — artificial intelligence that schedules meetings

I registered my interest back in Feb, to test drive this “software Personal Assistant”, and here’s the email confirmation I received:

— — — — — Forwarded message — — — — —
From: Dennis Mortensen <xyzxyz@x.ai>
Date: 11 February 2015 at 22:56
Subject: Waiting for x.ai, a personal assistant who schedules meetings for you.
To: xyzxyz@gmail.com

I really appreciate you signing up — and I am thrilled about the incredible demand we’ve seen for our artificial intelligence meeting scheduling solution!

We’re adding busy people as fast as we can every week, but drop us a kind note on @xdotai so we can find you and move you up the list. […]

Cheers,
Dennis and the x.ai team :-)

PS. You can read more about how Amy works here — and learn why GigaOm called Amy the best scheduling tool ever.

Now, initially, I thought it’s pure genius, though limited software. However, I’ve grown increasingly skeptical with AI-based “software”:

1. First of all, it’s been in private beta forever now (since early 2014).
2. For a beta product, it work surprisingly well. In fact, it just works perfectly fine. For every occurence I’ve been involved a meeting scheduling experiment with “Amy”. I mean 100% of the meeting occurrences are just perfectly organised, and it’s impossible to spot any “machine talk” at all. It just feels like it’s a human responding to your emails.
3. It can takes several hours for Amy’s email to be received, not very compatible with AI software and Cloud computing outputs
4. It takes forever to get a “private beta” access to Amy. This seems incompatible with the first point: if the software is already working so well, scaling it for more users in the age of Cloud computing should take minimal efforts — in the order of weeks — not years.
5. In March 2015, they have announced the new assistant called “Andrew”. The difference with Amy? Just the email address and the fact that it’s a male PA.

So the conclusion that I’ve drawn after several months of waiting to get a beta access is that currently Amy is not an AI software, but it’s simply a brilliant form of LeanStartup MVP experiment, with real humans reading and responding to emails.

So what’s an MVP experiment? It’s an experiment conducted by creating a minimal (or pseudo) product to validate an idea or a concept.

The case of Amy at @xdotai, the experiment is very conclusive: they have received loads of requests from potential users but the company behind it has probably not developed a single line of code to get to this result. And really, how much work do you think there is in setting up Andrew if you already have Amy? Does that even require an official announcement?

The only flaw in the experiment is that there is currently no charges associated with the service, so the day they’ve developed it and are ready to sell it, they might be up for a surprise on that front.

It’s a nice experiment, but it’s not a product yet, and I am quite disappointed about that.