AWS previews AppFabric for productivity – pitched as AI-powered glue between apps
Park user data in Amazon's servers for ML-generated insights and actions – yea or nay for you?
Re:Invent AWS has previewed AppFabric for productivity, which promises to use generative AI to "provide personalized insights and recommended actions generated from context across a user's other applications."
AWS introduced AppFabric in June this year – but what is it? The initial pitch was that it was "a fully managed service that aggregates and normalizes security data across SaaS applications to improve observability and help reduce operational effort."
The rationale is that users typically run a mix of applications from various vendors, such as Salesforce, Slack, Zoom, Microsoft 365 and/or Google Workspace, Atlassian Jira, and so on. Each application stores logs in its own format, making it hard to aggregate them to identify security incidents. AppFabric will ingest these logs and convert them to a common schema, specifically the Open Cybersecurity Schema Framework, so they can be analyzed together.
With the introduction of AppFabric for productivity, AWS has shown an ambition for using AppFabric for more than just security. According to Federico Torreti, head of product for AWS AppFabric, "App developers are coming to us and say … how can we simplify the technical complexity … in a way that ultimately our end users can benefit from richer, more engaging experiences."
The "technical complexity" is that as with the logs, each application has its own data formats and APIs, making integration between them difficult or a faff. To solve that lack of interoperability, AppFabric for productivity hopes to be a more general-purpose glue or bridge between applications, by providing APIs for creating emails and calendar events with Microsoft 365 or Google Workspace, creating tasks in those services as well as in Asana or Smartsheet, and similar jobs with compatible apps.
From what we can tell, these APIs, available now as a developer preview, are aimed at the makers of software-as-a-service applications who want to plug their apps into AWS so that their software suites can interoperate with other applications to meet user needs without said users having to switch between apps to get things done. For example: allowing folks to, from one app, automatically generate meeting agendas from messages in a group conversation in another application and send out invites and create calendar items. AppFabric may ingest and process user data within AWS to provide this functionality.
Slightly more specifically, we're told AppFabric's APIs can be used by applications to surface "actionable insights" from user data and help people prepare for upcoming meetings, all using the output of Anthropic’s Claude model, which Amazon has backed in a big way. If one application asks, on behalf of a user, AppFabric to send an email, arrange a virtual or in-person get-together, or produce a meeting summary, the service will gather relevant artifacts from connected applications and fire off action requests to get that stuff done, and use Claude in the mix as needed.
AppFabric itself has no user interface; it is down to app vendors to create a UI within their applications that ultimately uses the AppFabric APIs to achieve stuff. It also means apps connected via AppFabric can, in theory at least, all provide a consistent chat-bot-like interface, driven by Claude, to users.
So, how realistic is this?
We have to confess to skepticism. The list of supported actions and insights is short; users will inevitably still have to switch between applications to use their full features; and developers have one more thing to pay for and manage. How much value does the service add?
"There are three problems that AppFabric solves for developers," Torreti told The Register, after detailing research showing that users typically use six to eight different applications in their working day.
"It's not just about six connections. Each app needs to be connected to six apps. Now scale this across. On average, companies tend to deploy over 100 apps across their tech stack, so it becomes very unwieldy for any app developers to try and control these many-to-many integration points."
The second problem, Torreti opined, is that all the apps operate and handle data in their own way, which something like AppFabric can bring some consistency to. There is an element of normalization that needs to happen. Third, actually using the data is a challenge.
- AWS plays with Fire TV Cube, turns it into a thin client for cloudy desktops
- Desperately seeking GPUs? AWS will let you reserve instances in advance – no refunds
- Amazon bankrolling industry lobbying against Microsoft Azure should surprise no one
- AWS staffer shows off the workplace that used to be a prison
An attraction of AppFabric for platform vendors including Asana is that they can keep users longer in their application, by surfacing and automating some of the actions people might otherwise want to take elsewhere. Are we therefore going to see different app vendors fighting to be the primary point of engagement for users?
"I would say that it's a bit anachronistic to think through the lens of there is going to be one application to rule them all. Rather there is going to be an opportunity here for app developers to take advantage of this cross-application content and context to drive better engagement in their app,” Torreti explained.
Do business users want more engaging experiences, though; are they not more focused on getting their work done? "That's spot-on," said Torreti, saying that AppFabric can surface "what it is you need to care about across all the apps you are using," as well as, "what I need to know about my upcoming meetings."
The problems and friction of collaboration are real, but AppFabric's current scope is limited; the question is whether AWS can persuade enough enterprises of its value in order to give the service momentum. ®
More from Re:Invent
AppFabric for productivity was announced at this year's AWS Re:Invent conference, which kicked off Monday with about 90 announcements. We've covered the CodeWhisperer news here and the WorkSpaces Thin Client here. Below is some of other stuff Amazon has announced so far:
- AWS has produced a unified Billing and Cost Management console and a cost optimizer for its cloud subscribers, along with cost data exports.
- Amazon CloudWatch now supports "data querying from multiple sources, enabling you to gain visibility across your hybrid and multicloud metrics in a single view," according to Amazon. There's also CloudWatch Logs Infrequent Access and Anomaly Detection.
- AWS re:Post Private, a secure knowledge service, is now generally available.
- Amazon Transcribe has expanded language support and can (in preview) produce voice call summaries.
- AWS HealthScribe, a generative AI for healthcare apps, is now generally available.
- Amazon Simple Queue Service (SQS) has increased its quota for the high-throughput mode for FIFO queues.