The Alexa team has been busy since our last post, with a series of announcements and changes that are sending waves across the development community. We wanted to take a minute to throw out some hot takes about what we've seen.
Some of these are more substantial than the others, and we fully intend to dive deeper in those cases and do full writeups. But at a very surface level, here's how things look from our perspective:
Skill Builder
First off, lets jump into the big change to the developer console. Dave Isbitski has a post up about Alexa's brand new Skill Builder UI. This is essentially a revamp of the "interaction model" screen of the development console, with an interface much friendlier to new users. It does a lot of cool things, like automatically tagging slots in your utterances and providing a drag and drop interface for a new shared intent/utterance json file.
For experienced developers, though, the biggest change is likely going to be their new Dialog Model and Dialogue directives, which provide a way to essentially let Alexa auto-pilot the gathering of bits of information, moving the simple back-and-forth of multi-turn dialogues out of your own server side code. This is a neat idea, and we have a ton of questions about how it will play out in practice. (What happens if a user purposely breaks out of the Dialog Model mid-conversation? Can we piggy-back other concepts onto this feature? etc).
For those who liked the old interface, it's still around. The Skill Builder UI is entirely optional at this point. It's worth noting, however, that using the old interface does not provide you with a way to describe a dialog model, so you're handicapped in that way. There are also a couple caveats that come with the dialog model - you may not use the Yes/No built-in intents, and the Literal is strictly verboten - which may slow its adoption a bit.
(For an even deeper dive, check out Paul Cutsinger's 3-part series on the Alexa Developers Youtube channel, or expand below)
For experienced developers, though, the biggest change is likely going to be their new Dialog Model and Dialogue directives, which provide a way to essentially let Alexa auto-pilot the gathering of bits of information, moving the simple back-and-forth of multi-turn dialogues out of your own server side code. This is a neat idea, and we have a ton of questions about how it will play out in practice. (What happens if a user purposely breaks out of the Dialog Model mid-conversation? Can we piggy-back other concepts onto this feature? etc).
For those who liked the old interface, it's still around. The Skill Builder UI is entirely optional at this point. It's worth noting, however, that using the old interface does not provide you with a way to describe a dialog model, so you're handicapped in that way. There are also a couple caveats that come with the dialog model - you may not use the Yes/No built-in intents, and the Literal is strictly verboten - which may slow its adoption a bit.
(For an even deeper dive, check out Paul Cutsinger's 3-part series on the Alexa Developers Youtube channel, or expand below)
Dialog Directives
Beta Testing
And as if that wasn't enough, on that same day as the new UI release, Amazon announced their new beta testing paradigm. This is something that the development community has been asking for since day one, as it was extremely difficult (and risky!) to share unpublished skills with other users previously.
This one is almost strictly a positive. The feature does pretty much exactly what we need. We can provide a list of users to test our skill, which lets them try it out before publishing. We can also provide a deep-link for users to self-register. It also gives us some simple tester management options.
If we had to be super critical of the feature, it's worth noting that there's a immutable 90 day expiration on the beta. We're not quite sure what that accomplishes, and it unfortunately closes down some interesting avenues for private skill development. But all things considered, this is a huge move in the right direction.
This one is almost strictly a positive. The feature does pretty much exactly what we need. We can provide a list of users to test our skill, which lets them try it out before publishing. We can also provide a deep-link for users to self-register. It also gives us some simple tester management options.
If we had to be super critical of the feature, it's worth noting that there's a immutable 90 day expiration on the beta. We're not quite sure what that accomplishes, and it unfortunately closes down some interesting avenues for private skill development. But all things considered, this is a huge move in the right direction.
Location Awareness
Going back a little bit, the team also announced a location awareness feature, sort of out of left field. Among the top requests by both consumers and developers, this opens up a whole range of new skills, especially as Alexa makes inroads into the mobile space.
For a lot of us, the even bigger side-effect of this change was that we finally have a deviceId, allowing us to disambiguate different Alexa-enabled devices on a given account. The great hope here was that this deviceId would be both globally unique and static across skills, but alas that was too much to ask for and it's variable from skill-to-skill. That said, it still opens up a path (albeit a winding, roundabout path) for implementing some of the more immersive concepts developers have been sitting on. We're super excited to see what people come up with.
While the doors are now open for location information, our hope is that Amazon will actually abstract things out a bit further and give us the option to know just a user's timezone. We can obviously discern that most of the time from postal code, but that's a fair bit of work and also a privacy burden that isn't necessary in a lot of cases.
For a lot of us, the even bigger side-effect of this change was that we finally have a deviceId, allowing us to disambiguate different Alexa-enabled devices on a given account. The great hope here was that this deviceId would be both globally unique and static across skills, but alas that was too much to ask for and it's variable from skill-to-skill. That said, it still opens up a path (albeit a winding, roundabout path) for implementing some of the more immersive concepts developers have been sitting on. We're super excited to see what people come up with.
While the doors are now open for location information, our hope is that Amazon will actually abstract things out a bit further and give us the option to know just a user's timezone. We can obviously discern that most of the time from postal code, but that's a fair bit of work and also a privacy burden that isn't necessary in a lot of cases.
New Metrics Dashboard
And in what's seeming like something of a pattern, there was also a second big announcement on the same day as the location awareness update. The metrics dashboard got a huge overhaul, providing all kinds of new information. It did come at the expense of the old dashboard which had a nice overlay of all of your skills, which is a bummer, but it's all in the name of progress. And who knows, maybe the multi-skill overlay will come back soon.
The biggest addition, in our opinion, is the ability to see which intents are getting triggered at a given frequency. We've admittedly been a bit sloppy with paying attention to our intent ratios, and so some of the data that came out of our initial look at the new dashboard was a big wakeup call.
Now that Amazon has come closer to built-in parity with some of the third-party analytics tools, it'll hopefully nudge those groups to step up their game and take the reporting to the next level.
The biggest addition, in our opinion, is the ability to see which intents are getting triggered at a given frequency. We've admittedly been a bit sloppy with paying attention to our intent ratios, and so some of the data that came out of our initial look at the new dashboard was a big wakeup call.
Now that Amazon has come closer to built-in parity with some of the third-party analytics tools, it'll hopefully nudge those groups to step up their game and take the reporting to the next level.
AWS Promo Credits
Going all the way back to March, this may be the biggest announcement of any over the last month. Amazon is directly addressing the question of "why should we pay you for the right to develop free content for your platform?" by saying that most of us shouldn't be paying. Skill developers can now be reimbursed with up to $100 a month of AWS credits to apply to the AWS services they're using to run their skills. While this still does not provide a path to monetization, it does take away the disincentive of monthly fees, and it creates an environment where we are free to experiment with services outside of the ongoing free tier, use resources at levels above the bare minimum, and theoretically add in safety nets like redundancy.
Change to advertising rules
But, while one door opens, another one closes. In a silent change, the certification rules were updated to tighten the constraints around advertising in skills even further. As the Alexa platform has grown, it's been surprising to see the cert rules get stricter and stricter, whereas you'd generally expect a maturing platform to open up to a greater variety of use cases. The old rules allowed for a greater degree of advertising as long as Alexa's voice was not used to run the ads. The new rules around ads can be seen on the policy testing page. This change was made without fanfare, and there's no changelog, so it's not clear exactly when it happened.
Most developers aren't using in-skill advertising anyway, because it's super jarring (although, quick aside: we'd be super pumped if you clicked on our display ads on this blog...), but it's still frustrating to know that the option isn't even there anymore.
Most developers aren't using in-skill advertising anyway, because it's super jarring (although, quick aside: we'd be super pumped if you clicked on our display ads on this blog...), but it's still frustrating to know that the option isn't even there anymore.
All-in-all, it's been a really strong month for Alexa development. The changes aren't perfect, but 5 steps forward and 1 step back still makes for a really good pace. Let us know what you think of the changes, or which of these features you'd like to hear more about.
RSS Feed