Going to production while keeping the lights on

Welcome back to the Enterprise AI Playbook, Issue 8. Here are the successes, cautionary tales and deep dives from this week.

Successful launches - Big tech continues down the AI path

This week Amazon announced that it’s taking it’s Rufus shopping assistant GA. After “Customers have already asked Rufus tens of millions of questions” Amazon has decided that the product was ready for GA release. The initial version was met with some skepticism back in February, but it is interesting to see the e-commerce giant push forward with a full release.

Meanwhile Google has released its new AI video creation toolkit called Google vids for an alpha testing group. The product aims to help users create AI powered videos and we’ll see when it makes to GA after the initial review cycle.

Cautionary Tales - Blame it on the Gen AI, don’t blame it on me

Last week Intuit announced it was “Investing in our future” by laying off 1800 folks to make way for Gen AI progress. This included 1050 that were laid off due to performance issues. This announcement was juxtaposed with their Big Bet #1 which focused on AI:

”Our Gen AI-powered financial assistant, Intuit Assist, coupled with our network of experts, allows us to deliver delightful “done-for-you” experiences with a gateway to human expertise. To deliver these capabilities, we will accelerate investing in data and AI, leveraging GenOS to reimagine our products from traditional workflows to AI-native experiences…”

The push to Gen AI capabilities continues among companies, while showing a lack of human compassion as part of the announcements labeling folks laid off as underperformers. Within the middle of the post there was an emphasis on hiring Gen AI focused engineers and an assurance of growing head count in 2025. We’ll see how long Gen AI announcement continue to overshadow their human counterparts.

Full link

Deep dive - What’s actually worked in production

Last month I spoke at the Unparsed Conference focused on what’s worked in production for conversational AI. The panel was hosted by Karen Kaushansky, with Sebastian Glock and John F. Andersen also joining.

We covered important topics on:

1. Building the right team for going to production

2. Using the right metrics for a use case

3. Unexpected learning

4. Post launch iterations and product philosophy

5. Working with generative models vs traditional NLU

Launching to production is hard but great takeaways from this group in a 30 min panel!

Full video here.

Question to ask your team

How do you run alpha, beta and full production releases?

Until next week,
Denys - Enterprise AI @ Voiceflow